Release 0.7.0

Makes time display work for some raw audio formats (*.shn).

.travis.yml

@@ -4,17 +4,23 @@ os:
   - linux
   - osx
 env:
-  - LIBAV=libav-stable
-  - LIBAV=libav-git
-  - LIBAV=ffmpeg-stable
-  - LIBAV=ffmpeg-git
+  matrix:
+    - LIBAV=libav-stable
+    - LIBAV=libav-git
+    - LIBAV=ffmpeg-stable
+    - LIBAV=ffmpeg-git
+  global:
+    # Coverity token
+    - secure: "H21mSRlMhk4BKS0xHZvCFGJxteCP0hRVUxTuNfM2Z9HBsyutuLEYMtViLO86VtM+Tqla3xXPzUdS4ozLwI72Ax/5ZUDXACROj73yW6QhFB5D6rLut12+FjqC7M33Qv2hl0xwgNBmR5dsm1ToP37+Wn+ecJQNvN8fkTXF+HVzOEw="
 compiler:
   - clang
+  - gcc
 
 branches:
   only:
     - master
     - ci
+    - coverity_scan
 
 matrix:
   exclude:
@@ -24,6 +30,12 @@ matrix:
       env: LIBAV=libav-git
     - os: osx
       env: LIBAV=ffmpeg-git
+    - os: osx
+      compiler: gcc
+    - os: linux
+      env: LIBAV=ffmpeg-stable
+    - os: linux
+      compiler: clang
 
 before_install: ./travis-deps libass-stable $LIBAV
 script:
@@ -38,3 +50,13 @@ notifications:
       - "irc.freenode.org#mpv-player-dev"
     on_success: change
     on_failure: always
+
+addons:
+  coverity_scan:
+    project:
+      name: "mpv-player/mpv"
+      description: "Build submitted via Travis CI"
+    notification_email: mpv-team@googlegroups.com
+    build_command_prepend: "./bootstrap.py && ./waf configure"
+    build_command:   "./waf build"
+    branch_pattern: coverity_scan

DOCS/client-api-changes.rst

@@ -25,8 +25,38 @@ API changes
 
 ::
 
- --- mpv 0.5.2 is released ---
+ --- mpv 0.7.0 is released ---
+ 1.10   - deprecate/disable everything directly related to script_dispatch
+          (most likely affects nobody)
+ 1.9    - add enum mpv_end_file_reason for mpv_event_end_file.reason
+        - add MPV_END_FILE_REASON_ERROR and the mpv_event_end_file.error field
+          for slightly better error reporting on playback failure
+        - add --stop-playback-on-init-failure option, and make it the default
+          behavior for libmpv only
+        - add qthelper.hpp set_option_variant()
+        - mark the following events as deprecated:
+            MPV_EVENT_TRACKS_CHANGED
+            MPV_EVENT_TRACK_SWITCHED
+            MPV_EVENT_PAUSE
+            MPV_EVENT_UNPAUSE
+            MPV_EVENT_METADATA_UPDATE
+            MPV_EVENT_CHAPTER_CHANGE
+          They are handled better with mpv_observe_property() as mentioned in
+          the documentation comments. They are not removed and still work.
+ 1.8    - add qthelper.hpp
+ 1.7    - add mpv_command_node(), mpv_command_node_async()
+ 1.6    - modify "core-idle" property behavior
+        - MPV_EVENT_LOG_MESSAGE now always sends complete lines
+        - introduce numeric log levels (mpv_log_level)
+ --- mpv 0.6.0 is released ---
+ 1.5    - change in X11 and "--wid" behavior again. The previous change didn't
+          work as expected, and now the behavior can be explicitly controlled
+          with the "input-x11-keyboard" option. This is only a temporary
+          measure until XEmbed is implemented and confirmed working.
+          Note: in 1.6, "input-x11-keyboard" was renamed to "input-vo-keyboard",
+          although the old option name still works.
  1.4    - subtle change in X11 and "--wid" behavior
+          (this change was added to 0.5.2, and broke some things, see #1090)
  --- mpv 0.5.0 is released ---
  1.3    - add MPV_MAKE_VERSION()
  1.2    - remove "stream-time-pos" property (no replacement)

DOCS/client_api_examples/cocoabasic.m

@@ -0,0 +1,181 @@
+// Plays a video from the command line in a window provided by mpv.
+// You likely want to play the video in your own window instead,
+// but that's not quite ready yet.
+// You may need a basic Info.plist and MainMenu.xib to make this work.
+
+#include "../../libmpv/client.h"
+#include "shared.h"
+
+#import <Cocoa/Cocoa.h>
+
+@interface CocoaWindow : NSWindow
+@end
+
+@implementation CocoaWindow
+- (BOOL)canBecomeMainWindow { return YES; }
+- (BOOL)canBecomeKeyWindow { return YES; }
+@end
+
+@interface AppDelegate : NSObject <NSApplicationDelegate>
+{
+    mpv_handle *mpv;
+    dispatch_queue_t queue;
+    NSWindow *w;
+}
+@end
+
+static void wakeup(void *);
+
+@implementation AppDelegate
+
+- (void)createWindow {
+
+    int mask = NSTitledWindowMask|NSClosableWindowMask|
+               NSMiniaturizableWindowMask|NSResizableWindowMask;
+
+    self->w = [[CocoaWindow alloc]
+        initWithContentRect:NSMakeRect(0,0, 1280, 720)
+                  styleMask:mask
+                    backing:NSBackingStoreBuffered
+                      defer:NO];
+
+    [self->w setTitle:@"cocoabasic example"];
+    [self->w makeMainWindow];
+    [self->w makeKeyAndOrderFront:nil];
+
+    NSMenu *m = [[NSMenu alloc] initWithTitle:@"AMainMenu"];
+    NSMenuItem *item = [m addItemWithTitle:@"Apple" action:nil keyEquivalent:@""];
+    NSMenu *sm = [[NSMenu alloc] initWithTitle:@"Apple"];
+    [m setSubmenu:sm forItem:item];
+    [sm addItemWithTitle: @"Shutdown mpv" action:@selector(shutdown) keyEquivalent:@"s"];
+    [NSApp setMenu:m];
+    [NSApp activateIgnoringOtherApps:YES];
+}
+
+- (void) applicationDidFinishLaunching:(NSNotification *)notification {
+    [NSApp setActivationPolicy:NSApplicationActivationPolicyRegular];
+    atexit_b(^{
+        // Because activation policy has just been set to behave like a real
+        // application, that policy must be reset on exit to prevent, among
+        // other things, the menubar created here from remaining on screen.
+        [NSApp setActivationPolicy:NSApplicationActivationPolicyProhibited];
+    });
+
+    // Read filename
+    NSArray *args = [NSProcessInfo processInfo].arguments;
+    if (args.count < 2) {
+        NSLog(@"Expected filename on command line");
+        exit(1);
+    }
+    NSString *filename = args[1];
+
+    [self createWindow];
+
+    // Deal with MPV in the background.
+    queue = dispatch_queue_create("mpv", DISPATCH_QUEUE_SERIAL);
+    dispatch_async(queue, ^{
+
+        mpv = mpv_create();
+        if (!mpv) {
+            printf("failed creating context\n");
+            exit(1);
+        }
+
+        int64_t wid = (intptr_t) [self->w contentView];
+        check_error(mpv_set_option(mpv, "wid", MPV_FORMAT_INT64, &wid));
+
+        // Maybe set some options here, like default key bindings.
+        // NOTE: Interaction with the window seems to be broken for now.
+        check_error(mpv_set_option_string(mpv, "input-default-bindings", "yes"));
+
+        // for testing!
+        check_error(mpv_set_option_string(mpv, "input-media-keys", "yes"));
+        check_error(mpv_set_option_string(mpv, "input-cursor", "no"));
+        check_error(mpv_set_option_string(mpv, "input-vo-keyboard", "yes"));
+
+        // request important errors
+        check_error(mpv_request_log_messages(mpv, "warn"));
+
+        check_error(mpv_initialize(mpv));
+
+        // Register to be woken up whenever mpv generates new events.
+        mpv_set_wakeup_callback(mpv, wakeup, (__bridge void *) self);
+
+        // Load the indicated file
+        const char *cmd[] = {"loadfile", filename.UTF8String, NULL};
+        check_error(mpv_command(mpv, cmd));
+    });
+}
+
+- (void) handleEvent:(mpv_event *)event
+{
+    switch (event->event_id) {
+    case MPV_EVENT_SHUTDOWN:
+        // Clean up and shut down.
+        mpv_terminate_destroy(mpv);
+        mpv = NULL;
+        dispatch_async(dispatch_get_main_queue(), ^{
+            [[NSApplication sharedApplication] terminate:nil];
+        });
+        break;
+
+    case MPV_EVENT_LOG_MESSAGE: {
+        struct mpv_event_log_message *msg = (struct mpv_event_log_message *)event->data;
+        printf("[%s] %s: %s", msg->prefix, msg->level, msg->text);
+    }
+
+    case MPV_EVENT_VIDEO_RECONFIG:
+        dispatch_async(dispatch_get_main_queue(), ^{
+            [self->w selectNextKeyView:nil];
+        });
+
+    default:
+        printf("event: %s\n", mpv_event_name(event->event_id));
+    }
+}
+
+- (void) readEvents
+{
+    dispatch_async(queue, ^{
+        while (mpv) {
+            mpv_event *event = mpv_wait_event(mpv, 0);
+            if (event->event_id == MPV_EVENT_NONE)
+                break;
+            [self handleEvent:event];
+        }
+    });
+}
+
+static void wakeup(void *context) {
+    AppDelegate *a = (__bridge AppDelegate *) context;
+    [a readEvents];
+}
+
+// Ostensibly, mpv's window would be hooked up to this.
+- (BOOL) windowShouldClose:(id)sender
+{
+    [self shutdown];
+    if (self->w)
+        [self->w release];
+    return YES;
+}
+
+- (void) shutdown
+{
+    if (mpv) {
+        const char *args[] = {"quit", NULL};
+        mpv_command(mpv, args);
+    }
+}
+@end
+
+// Delete this if you already have a main.m.
+int main(int argc, const char * argv[]) {
+    @autoreleasepool {
+        NSApplication *app = [NSApplication sharedApplication];
+        AppDelegate *delegate = [AppDelegate new];
+        app.delegate = delegate;
+        [app run];
+    }
+    return EXIT_SUCCESS;
+}

DOCS/client_api_examples/qtexample.cpp

@@ -2,14 +2,23 @@
 
 // This example can be built with: qmake && make
 
+#include <clocale>
 #include <sstream>
 
+#include <QtGlobal>
 #include <QFileDialog>
 #include <QStatusBar>
 #include <QMenuBar>
 #include <QMenu>
 #include <QGridLayout>
 #include <QApplication>
+#include <QTextEdit>
+
+#if QT_VERSION >= 0x050000
+#include <QJsonDocument>
+#endif
+
+#include <mpv/qthelper.hpp>
 
 #include "qtexample.h"
 
@@ -26,6 +35,9 @@ static void wakeup(void *ctx)
 MainWindow::MainWindow(QWidget *parent) :
     QMainWindow(parent)
 {
+    setWindowTitle("Qt embedding demo");
+    setMinimumSize(640, 480);
+
     QMenu *menu = menuBar()->addMenu(tr("&File"));
     QAction *on_open = new QAction(tr("&Open"), this);
     on_open->setShortcuts(QKeySequence::Open);
@@ -35,16 +47,25 @@ MainWindow::MainWindow(QWidget *parent) :
 
     statusBar();
 
+    QMainWindow *log_window = new QMainWindow(this);
+    log = new QTextEdit(log_window);
+    log->setReadOnly(true);
+    log_window->setCentralWidget(log);
+    log_window->setWindowTitle("mpv log window");
+    log_window->setMinimumSize(500, 50);
+    log_window->show();
+
     mpv = mpv_create();
     if (!mpv)
         throw "can't create mpv instance";
 
     // Create a video child window. Force Qt to create a native window, and
-    // pass the window ID to the mpv wid option. This doesn't work on OSX,
-    // because Cocoa doesn't support this form of embedding.
+    // pass the window ID to the mpv wid option. Works on: X11, win32, Cocoa
     mpv_container = new QWidget(this);
     setCentralWidget(mpv_container);
+    mpv_container->setAttribute(Qt::WA_DontCreateNativeAncestors);
     mpv_container->setAttribute(Qt::WA_NativeWindow);
+    // If you have a HWND, use: int64_t wid = (intptr_t)hwnd;
     int64_t wid = mpv_container->winId();
     mpv_set_option(mpv, "wid", MPV_FORMAT_INT64, &wid);
 
@@ -52,10 +73,21 @@ MainWindow::MainWindow(QWidget *parent) :
     // mpv as backend would implement its own key bindings.
     mpv_set_option_string(mpv, "input-default-bindings", "yes");
 
+    // Enable keyboard input on the X11 window. For the messy details, see
+    // --input-vo-keyboard on the manpage.
+    mpv_set_option_string(mpv, "input-vo-keyboard", "yes");
+
     // Let us receive property change events with MPV_EVENT_PROPERTY_CHANGE if
     // this property changes.
     mpv_observe_property(mpv, 0, "time-pos", MPV_FORMAT_DOUBLE);
 
+    mpv_observe_property(mpv, 0, "track-list", MPV_FORMAT_NODE);
+    mpv_observe_property(mpv, 0, "chapter-list", MPV_FORMAT_NODE);
+
+    // Request log messages with level "info" or higher.
+    // They are received as MPV_EVENT_LOG_MESSAGE.
+    mpv_request_log_messages(mpv, "info");
+
     // From this point on, the wakeup function will be called. The callback
     // can come from any thread, so we use the Qt QEvent mechanism to relay
     // the wakeup in a thread-safe way.
@@ -81,9 +113,47 @@ void MainWindow::handle_mpv_event(mpv_event *event)
                 // was stopped.
                 statusBar()->showMessage("");
             }
+        } else if (strcmp(prop->name, "chapter-list") == 0 ||
+                   strcmp(prop->name, "track-list") == 0)
+        {
+            // Dump the properties as JSON for demo purposes.
+#if QT_VERSION >= 0x050000
+            if (prop->format == MPV_FORMAT_NODE) {
+                QVariant v = mpv::qt::node_to_variant((mpv_node *)prop->data);
+                // Abuse JSON support for easily printing the mpv_node contents.
+                QJsonDocument d = QJsonDocument::fromVariant(v);
+                append_log("Change property " + QString(prop->name) + ":\n");
+                append_log(d.toJson().data());
+            }
+#endif
         }
         break;
     }
+    case MPV_EVENT_VIDEO_RECONFIG: {
+        // Retrieve the new video size.
+        int64_t w, h;
+        if (mpv_get_property(mpv, "dwidth", MPV_FORMAT_INT64, &w) >= 0 &&
+            mpv_get_property(mpv, "dheight", MPV_FORMAT_INT64, &h) >= 0 &&
+            w > 0 && h > 0)
+        {
+            // Note that the MPV_EVENT_VIDEO_RECONFIG event doesn't necessarily
+            // imply a resize, and you should check yourself if the video
+            // dimensions really changed.
+            // mpv itself will scale/letter box the video to the container size
+            // if the video doesn't fit.
+            std::stringstream ss;
+            ss << "Reconfig: " << w << " " << h;
+            statusBar()->showMessage(QString::fromStdString(ss.str()));
+        }
+        break;
+    }
+    case MPV_EVENT_LOG_MESSAGE: {
+        struct mpv_event_log_message *msg = (struct mpv_event_log_message *)event->data;
+        std::stringstream ss;
+        ss << "[" << msg->prefix << "] " << msg->level << ": " << msg->text;
+        append_log(QString::fromStdString(ss.str()));
+        break;
+    }
     case MPV_EVENT_SHUTDOWN: {
         mpv_terminate_destroy(mpv);
         mpv = NULL;
@@ -120,6 +190,14 @@ void MainWindow::on_file_open()
     }
 }
 
+void MainWindow::append_log(const QString &text)
+{
+    QTextCursor cursor = log->textCursor();
+    cursor.movePosition(QTextCursor::End);
+    cursor.insertText(text);
+    log->setTextCursor(cursor);
+}
+
 MainWindow::~MainWindow()
 {
     if (mpv)
@@ -129,6 +207,11 @@ MainWindow::~MainWindow()
 int main(int argc, char *argv[])
 {
     QApplication a(argc, argv);
+
+    // Qt sets the locale in the QApplication constructor, but libmpv requires
+    // the LC_NUMERIC category to be set to "C", so change it back.
+    std::setlocale(LC_NUMERIC, "C");
+
     MainWindow w;
     w.show();
 

DOCS/client_api_examples/qtexample.h

@@ -5,6 +5,8 @@
 
 #include <mpv/client.h>
 
+class QTextEdit;
+
 class MainWindow : public QMainWindow
 {
     Q_OBJECT
@@ -22,6 +24,9 @@ private slots:
 private:
     QWidget *mpv_container;
     mpv_handle *mpv;
+    QTextEdit *log;
+
+    void append_log(const QString &text);
 
     void create_player();
     void handle_mpv_event(mpv_event *event);

DOCS/client_api_examples/qtexample.pro

@@ -5,7 +5,7 @@ greaterThan(QT_MAJOR_VERSION, 4): QT += widgets
 TARGET = qtexample
 TEMPLATE = app
 
-CONFIG += link_pkgconfig
+CONFIG += link_pkgconfig debug
 PKGCONFIG = mpv
 
 SOURCES += qtexample.cpp

DOCS/client_api_examples/simple.c

@@ -21,6 +21,7 @@ int main(int argc, char *argv[])
     // Enable default key bindings, so the user can actually interact with
     // the player (and e.g. close the window).
     check_error(mpv_set_option_string(ctx, "input-default-bindings", "yes"));
+    mpv_set_option_string(ctx, "input-x11-keyboard", "yes");
     int val = 1;
     check_error(mpv_set_option(ctx, "osc", MPV_FORMAT_FLAG, &val));
 

DOCS/crosscompile-mingw.md

@@ -26,6 +26,15 @@ created scripts to help ease the process. These are the two recommended methods:
 Note that MinGW environments included in Linux distributions are often broken,
 outdated and useless, and usually don't use MinGW-w64.
 
+Additional dependencies
+-----------------------
+
+You need a pthread wrapper. It must be interoperable with native Windows
+threads. pthreads-win32 or MinGW pthreads might work.
+
+If you want ``--input-file=...`` to work, you need libwaio. It's available
+from: git://midipix.org/waio
+
 Example with MXE
 ----------------
 

DOCS/edl-mpv.rst

@@ -27,13 +27,13 @@ The general simplified syntax is::
     <filename>,<start in seconds>,<length in seconds>
 
 If the start time is omitted, 0 is used. If the length is omitted, the
-estimated duration of the source file is used.
+estimated remaining duration of the source file is used.
 
 Note::
 
-    mpv can't use ordered chapter files or libquvi-resolved URLs in EDL
-    entries. Usage of relative or absolute paths as well as any protocol
-    prefixes is prevented for security reasons.
+    mpv can't use ordered chapter files in EDL entries. Usage of relative or
+    absolute paths as well as any protocol prefixes is prevented for security
+    reasons.
 
 
 Syntax of mpv EDL files

DOCS/man/af.rst

@@ -77,10 +77,12 @@ Available filters are:
     ``bitrate=<rate>``
         The bitrate use for the AC-3 stream. Set it to 384 to get 384 kbps.
 
+        The default is 640. Some receivers might not be able to handle this.
+
         Valid values: 32, 40, 48, 56, 64, 80, 96, 112, 128,
         160, 192, 224, 256, 320, 384, 448, 512, 576, 640.
 
-        The special value ``default`` selects a default bitrate based on the
+        The special value ``auto`` selects a default bitrate based on the
         input channel number:
 
         :1ch: 96
@@ -92,7 +94,7 @@ Available filters are:
 
     ``minchn=<n>``
         If the input channel number is less than ``<minchn>``, the filter will
-        detach itself (default: 5).
+        detach itself (default: 3).
 
 ``sweep[=speed]``
     Produces a sine sweep.
@@ -263,9 +265,8 @@ Available filters are:
     Filter for internal use only. Converts between 24-bit and 32-bit sample
     formats.
 
-``convertsignendian``
-    Filter for internal use only. Converts between signed/unsigned formats
-    and formats with different endian.
+``convertsign``
+    Filter for internal use only. Converts between signed/unsigned formats.
 
 ``volume[=<volumedb>[:...]]``
     Implements software volume control. Use this filter with caution since it

DOCS/man/ao.rst

@@ -114,25 +114,18 @@ Available audio output drivers are:
         with fixed setups.
 
 ``coreaudio`` (Mac OS X only)
-    Native Mac OS X audio output driver using the AudioUnits and the CoreAudio
+    Native Mac OS X audio output driver using AudioUnits and the CoreAudio
     sound server.
 
-    ``device_id=<id>``
-        ID of output device to use (0 = default device)
-    ``help``
-        List all available output devices with their IDs.
+    Automatically redirects to ``coreaudio_exclusive`` when playing compressed
+    formats.
 
-``coreaudio_device`` (Mac OS X only)
+``coreaudio_exclusive`` (Mac OS X only)
     Native Mac OS X audio output driver using direct device access and
     exclusive mode (bypasses the sound server).
 
     Supports only compressed formats (AC3 and DTS).
 
-    ``device_id=<id>``
-        ID of output device to use (0 = default device)
-    ``help``
-        List all available output devices with their IDs.
-
 ``openal``
     Experimental OpenAL audio output driver
 
@@ -181,6 +174,9 @@ Available audio output drivers are:
         Sets the device number to use. Playing a file with ``-v`` will show a
         list of available devices.
 
+    ``buffersize=<ms>``
+        DirectSound buffer size in milliseconds (default: 200).
+
 ``sdl``
     SDL 1.2+ audio output driver. Should work on any platform supported by SDL
     1.2, but may require the ``SDL_AUDIODRIVER`` environment variable to be set

DOCS/man/changes.rst

@@ -10,104 +10,159 @@ should generally be treated as a completely different program.
 .. note::
     These lists are incomplete.
 
-General Changes for MPlayer to mpv
-----------------------------------
+General Changes from MPlayer to mpv
+-----------------------------------
 
-* Switch to GPLv2+. (Technically speaking, MPlayer as a whole seems to be
-  GPLv2-only, while mplayer2 is GPLv3+ - see ``Copyright`` file for details.)
-* Removal of the internal GUI, MEncoder, OSD menu, video kernel drivers for
-  Linux 2.4 (including VIDIX)
-* Large internal cleanups
-* Removal of support for dead platforms
-* New build system
-* No embedded copy of FFmpeg and other libraries
-* Better pause handling (do not unpause on a command)
-* Better MKV support (such as ordered chapters)
-* vo_vdpau improvements
-* Precise seeking support
-* Native OpenGL backend for OS X
-* General OS X improvements
-* Improvements in audio/video sync handling
-* Cleaned up terminal output
-* Gapless audio support (``--gapless-audio``)
-* Improved responsiveness on user input
-* Support for modifier keys (alt, shift, ctrl) in input.conf
-* OSS4 volume control
-* More correct color reproduction (color matrix generation)
-* Use libass for subtitle rendering by default (better quality)
-* Generally preferring FFmpeg/Libav over internal demuxers and decoders
-* Improvements when playing multiple files (``--fixed-vo``)
-* Screenshot improvements (instant screenshots without 1-frame delay, allow
-  taking screenshots even with hardware decoding)
-* Improved support for PulseAudio
-* Generally improved MS Windows support (dealing with Unicode file names,
-  improved ``--vo=direct3d``, improved window handling)
-* Better OSD rendering (using libass). This has full Unicode support, and
-  languages like Arabic should be better supported.
-* Cleaned up terminal output (nicer status line, less useless noise)
-* Support for playing URLs of popular streaming sites directly
-  (e.g. ``mpv https://www.youtube.com/watch?v=...``)
-* Improved OpenGL output (``--vo=opengl-hq``)
-* Make ``--softvol`` default (**mpv** is not a mixer control panel)
-* Improved support for .cue files
-* Screenshot improvements (can save screenshots as JPEG or PNG, configurable
-  file names, support for taking screenshots with or without subtitles - the
-  ``screenshot`` video filter is not needed anymore, and should not be put
-  into the mpv config file)
-* Removal of teletext support
-* Removal of most built-in demuxers, using libavformat instead
-* Removal of built-in network support, using libavformat instead (also,
-  support https via libavformat)
-* Replace image VOs (``--vo=jpeg`` etc.) with ``--vo=image``
-* Do not lose settings when playing a new file in the same player instance
-* New location for config files, new name for the binary.
-* Slave mode compatibility broken (see below)
-* Encoding functionality (replacement for MEncoder, see ``DOCS/encoding.rst``)
-* Remove ``--vo=gif89a``, ``--vo=md5sum``, ``--vo=yuv4mpeg``, as encoding can
-  handle these use cases. For yuv4mpeg, for example, use:
-  ``mpv input.mkv -o output.y4m --no-audio --oautofps --oneverdrop``.
-* Image subtitles (DVDs etc.) are rendered in color and use more correct
-  positioning (color can be disabled with ``--sub-gray``)
-* Wayland support
-* Support for precise scrolling which scales the parameter of commands. If the
-  input doesn't support precise scrolling the scale factor stays 1.
-* OS X: Cocoa event loop is independent from MPlayer's event loop, so user
-  actions like accessing menus and live resizing do not block the playback.
-* OS X: Apple Remote support.
-* OS X: Media Keys support.
-* Windows: Added WASAPI audio output.
-* New OSD bar with chapter marks and not positioned in the middle of the video
-  (though this can be customized with the ``--osd-bar-align-y`` option).
-* Allow customizing whether a key binding for seeking shows the video time, the
-  OSD bar, or nothing (see section `Input Command Prefixes`_).
-* Display list of chapters and audio/subtitle tracks on OSD (see section
-  `Properties`_).
+This listing is about changes introduced by mplayer2 and mpv relatively to
+MPlayer.
+
+Player
+~~~~~~
+
+* New name for the binary (``mpv``). New location for config files (either
+  ``~/.config/mpv/mpv.conf``, or if you want, ``~/.mpv/config``).
+* Encoding functionality (replacement for MEncoder, see the `ENCODING`_ section).
+* Support for Lua scripting (see the `LUA SCRIPTING`_ section).
+* Better pause handling (e.g. do not unpause on a command).
+* Precise seeking support.
+* Improvements in audio/video sync handling.
+* Do not lose settings when playing a new file in the same player instance.
+* Slave mode compatibility broken (see below).
+* Re-enable screensaver while the player is paused.
+* Allow resuming playback at a later point with ``Shift+q``, also see the
+  ``quit_watch_later`` input command.
 * ``--keep-open`` option to stop the player from closing the window and
   exiting after playback ends.
-* Re-enable screensaver while the player is paused.
-* Matroska edition switching at runtime.
+* A client API, that allows embedding **mpv** into applications
+  (see ``libmpv/client.h`` in the sources).
+
+Input
+~~~~~
+
+* Improved default keybindings. MPlayer bindings are also available (see
+  ``etc/mplayer-input.conf`` in the source tree).
+* Improved responsiveness on user input.
+* Support for modifier keys (alt, shift, ctrl) in input.conf.
+* Allow customizing whether a key binding for seeking shows the video time, the
+  OSD bar, or nothing (see the `Input Command Prefixes`_ section).
+* Support mapping multiple commands to one key.
+
+Audio
+~~~~~
+
+* Support for gapless audio (see the ``--gapless-audio`` option).
+* Support for OSS4 volume control.
+* Improved support for PulseAudio.
+* Make ``--softvol`` default (**mpv** is not a mixer control panel).
+* By default, do pitch correction if playback speed is increased.
+* Improved downmixing and output of surround audio:
+
+  - Instead of using hardcoded pan filters to do remixing, libavresample is used
+  - Channel maps are used to identify the channel layout, so e.g. ``3.0`` and
+    ``2.1`` audio can be distinguished.
+
+Video
+~~~~~
+
+* Wayland support.
+* Native support for VAAPI and VDA. Improved VDPAU video output.
+* Improved OpenGL output (see the ``opengl-hq`` video output).
+* Make hardware decoding work with the ``opengl`` video output.
 * Support for libavfilter (for video->video and audio->audio). This allows
   using most of FFmpeg's filters, which improve greatly on the old MPlayer
   filters in features, performance, and correctness.
-* Improved downmixing and output of surround audio. Instead of using hardcoded
-  pan filters to do remixing, use libavresample. Channel maps are used to
-  identify the channel layout, so e.g. ``3.0`` and ``2.1`` audio can be
-  distinguished.
-* Allow resuming playback at a later point with ``Shift+q``, also see
-  ``quit_watch_later`` input command.
-* Support mapping multiple commands to one key.
+* More correct color reproduction (color matrix generation).
+* Better subtitle rendering using libass by default.
+* Improvements when playing multiple files (``--fixed-vo`` is default, do not
+  reset settings by default when playing a new file).
+* Replace image video outputs (``--vo=jpeg`` etc.) with ``--vo=image``.
+* Removal of ``--vo=gif89a``, ``--vo=md5sum``, ``--vo=yuv4mpeg``, as encoding
+  can handle these use cases. For yuv4mpeg, for example, use::
+
+    mpv input.mkv -o output.y4m --no-audio --oautofps --oneverdrop
+
+* Image subtitles (DVDs etc.) are rendered in color and use more correct
+  positioning (color for image subs can be disabled with ``--sub-gray``).
+
+OSD and terminal
+~~~~~~~~~~~~~~~~
+
+* Cleaned up terminal output: nicer status line, less useless noise.
+* Improved OSD rendering using libass, with full Unicode support.
+* New OSD bar with chapter marks. Not positioned in the middle of the video
+  (this can be customized with the ``--osd-bar-align-y`` option).
+* Display list of chapters and audio/subtitle tracks on OSD (see the
+  `Properties`_ section).
+
+Screenshots
+~~~~~~~~~~~
+
+* Instant screenshots without 1-frame delay.
+* Support for taking screenshots even with hardware decoding.
+* Support for saving screenshots as JPEG or PNG.
+* Support for configurable file names.
+* Support for taking screenshots with or without subtitles.
+
+Note that the ``screenshot`` video filter is not needed anymore, and should not
+be put into the mpv config file.
+
+Miscellaneous
+~~~~~~~~~~~~~
+
+* Better MKV support (e.g. ordered chapters, 3D metadata).
+* Matroska edition switching at runtime.
+* Support for playing URLs of popular streaming sites directly.
+  (e.g. ``mpv https://www.youtube.com/watch?v=...``).
+  Requires a recent version of ``youtube-dl`` to be installed, and
+  ``ytdl=yes`` in the mpv config file.
+* Support for precise scrolling which scales the parameter of commands. If the
+  input doesn't support precise scrolling the scale factor stays 1.
 * Allow changing/adjusting video filters at runtime. (This is also used to make
-  the ``D`` key insert vf_yadif if deinterlacing is not supported otherwise.)
-* Native VAAPI support
-* OS X: VDA support using libavcodec hwaccel API instead of FFmpeg's decoder. Up
+  the ``D`` key insert vf_yadif if deinterlacing is not supported otherwise).
+* Improved support for .cue files.
+
+Mac OS X
+~~~~~~~~
+
+* Native OpenGL backend.
+* Cocoa event loop is independent from MPlayer's event loop, so user
+  actions like accessing menus and live resizing do not block the playback.
+* Apple Remote support.
+* Media Keys support.
+* VDA support using libavcodec hwaccel API instead of FFmpeg's decoder with up
   to 2-2.5x reduction in CPU usage.
-* Make hardware decoding in general work with the ``opengl`` video output.
-* Lua scripting (see `LUA SCRIPTING`_)
-* A client API, that allows embedding **mpv** into applications
-  (see ``libmpv/client.h`` in the sources)
-* General bug fixes and removal of long-standing issues
-* General code cleanups (including refactoring or rewrites of many parts)
-* Many more changes
+
+Windows
+~~~~~~~
+
+* Improved support for Unicode file names.
+* Improved window handling.
+* Do not block playback when moving the window.
+* Improved Direct3D video output.
+* Added WASAPI audio output.
+
+Internal changes
+~~~~~~~~~~~~~~~~
+
+* Switch to GPLv2+ (see ``Copyright`` file for details).
+* Removal of lots of cruft:
+
+  - Internal GUI (replaced by the OSC, see the `ON SCREEN CONTROLLER`_ section).
+  - MEncoder (replaced by native encoding, see the `ENCODING`_ section).
+  - OSD menu.
+  - Kernel video drivers for Linux 2.4 (including VIDIX).
+  - Teletext support.
+  - Support for dead platforms.
+  - Most built-in demuxers have been replaced by their libavformat counterparts.
+  - Built-in network support has been replaced by libavformat's (which also
+    supports https URLs).
+  - Embedded copies of libraries (such as FFmpeg).
+
+* General code cleanups (including refactoring or rewrites of many parts).
+* New build system.
+* Many bug fixes and removal of long-standing issues.
+* Generally preferring FFmpeg/Libav over internal demuxers, decoders, and
+  filters.
 
 Detailed Listing of User-visible Changes
 ----------------------------------------
@@ -334,8 +389,8 @@ Slave mode
 
   (The option was readded in 0.5.1 and sets exactly these options.)
 
-* A JSON RPC protocol giving access to the client API is planned, but nothing
-  has emerged yet.
+* A JSON RPC protocol giving access to the client API is also supported. See
+  `JSON IPC`_ for more information.
 
 * **mpv** also provides a client API, which can be used to embed the player
   by loading it as shared library. (See ``libmpv/client.h`` in the sources.)

DOCS/man/input.rst

@@ -11,8 +11,9 @@ value (upper case if combined with ``Shift``), or a name for special keys. For
 example, ``a`` maps to the ``a`` key without shift, and ``A`` maps to ``a``
 with shift.
 
-The file is normally located in ``~/.config/mpv/input.conf`` (some platforms
-may use different paths). The default bindings are defined here::
+The file is located in the mpv configuration directory (normally at
+``~/.config/mpv/input.conf`` depending on platform). The default bindings are
+defined here::
 
     https://github.com/mpv-player/mpv/blob/master/etc/input.conf
 
@@ -159,7 +160,9 @@ List of Input Commands
         Take a single screenshot.
     <each-frame>
         Take a screenshot each frame. Issue this command again to stop taking
-        screenshots.
+        screenshots. Note that you should disable frame-dropping when using
+        this mode - or you might receive duplicate images in cases when a
+        frame was dropped.
 
 ``screenshot_to_file "<filename>" [subtitles|video|window]``
     Take a screenshot and save it to a given file. The format of the file will
@@ -190,7 +193,7 @@ List of Input Commands
     force
         Terminate playback if the first file is being played.
 
-``loadfile "<file>" [replace|append [options]]``
+``loadfile "<file>" [replace|append|append-play [options]]``
     Load the given file and play it.
 
     Second argument:
@@ -201,6 +204,8 @@ List of Input Commands
         Append the file to the playlist.
     <append-play>
         Append the file, and if nothing is currently playing, start playback.
+        (Always starts with the added file, even if the playlist was not empty
+        before running this command.)
 
     The third argument is a list of options and values which should be set
     while the file is playing. It is of the form ``opt1=value1,opt2=value2,..``.
@@ -257,16 +262,39 @@ List of Input Commands
     will seek to the previous position on start. The (optional) argument is
     exactly as in the ``quit`` command.
 
-``sub_add "<file>"``
-    Load the given subtitle file. It is not selected as current subtitle after
+``sub_add "<file>" [<flags> [<title> [<lang>]]]``
+    Load the given subtitle file. It is selected as current subtitle after
     loading.
 
+    The ``flags`` args is one of the following values:
+
+    <select>
+
+        Select the subtitle immediately.
+
+    <auto>
+
+        Don't select the subtitle. (Or in some special situations, let the
+        default stream selection mechanism decide.)
+
+    <cached>
+
+        Select the subtitle. If a subtitle with the same filename was already
+        added, that one is selected, instead of loading a duplicate entry.
+        (In this case, title/language are ignored, and if the was changed since
+        it was loaded, these changes won't be reflected.)
+
+    The ``title`` argument sets the track title in the UI.
+
+    The ``lang`` argument sets the track language, and can also influence
+    stream selection with ``flags`` set to ``auto``.
+
 ``sub_remove [<id>]``
     Remove the given subtitle track. If the ``id`` argument is missing, remove
     the current track. (Works on external subtitle files only.)
 
 ``sub_reload [<id>]``
-    Reload the given subtitle tracks. If the ``id`` argument is missing, remove
+    Reload the given subtitle tracks. If the ``id`` argument is missing, reload
     the current track. (Works on external subtitle files only.)
 
     This works by unloading and re-adding the subtitle track.
@@ -325,6 +353,12 @@ List of Input Commands
     Write the resume config file that the ``quit_watch_later`` command writes,
     but continue playback normally.
 
+``stop``
+    Stop playback and clear playlist. With default settings, this is
+    essentially like ``quit``. Useful for the client API: playback can be
+    stopped without terminating the player.
+
+
 Input Commands that are Possibly Subject to Change
 --------------------------------------------------
 
@@ -394,7 +428,7 @@ Input Commands that are Possibly Subject to Change
     The special argument ``!reverse`` can be used to cycle the value list in
     reverse. Compared with a command that just lists the value in reverse, this
     command will actually share the internal counter with the forward-cycling
-    key binding.
+    key binding (as long as the rest of the arguments are the same).
 
     Note that there is a static limit of (as of this writing) 10 arguments
     (this limit could be raised on demand).
@@ -448,6 +482,16 @@ Input Commands that are Possibly Subject to Change
     random points. Instead, call ``overlay_add`` again (preferably with a
     different memory region to prevent tearing).
 
+    It is also possible to pass a raw memory address for use as bitmap memory
+    by passing a memory address as integer prefixed with a ``&`` character.
+    Passing the wrong thing here will crash the player. The ``offset`` parameter
+    is not used and must be 0. This mode might be useful for use with libmpv.
+
+    On Windows, currently only raw memory addresses work. File mapping is not
+    implemented because a ``mmap`` compatibility layer is missing, and because
+    this kind of shared memory method would perhaps not be overly useful on
+    Windows.
+
     ``offset`` is the offset of the first pixel in the source file. It is
     passed directly to ``mmap`` and is subject to certain restrictions
     (see ``man mmap`` for details). In particular, this value has to be a
@@ -497,12 +541,46 @@ Input Commands that are Possibly Subject to Change
     ``<target>``. Each client (scripts etc.) has a unique name. For example,
     Lua scripts can get their name via ``mp.get_script_name()``.
 
-    (Scripts use this internally to dispatch key bindings, and this can also
-    be used in input.conf to reassign such bindings.)
+``script_binding "<name>"``
+    Invoke a script-provided key binding. This can be used to remap key
+    bindings provided by external Lua scripts.
 
+    The argument is the name of the binding.
 
-Undocumented commands: ``tv_last_channel`` (TV/DVB only), ``stop`` (questionable
-use), ``get_property`` (?), ``vo_cmdline`` (experimental).
+    It can optionally be prefixed with the name of the script, using ``/`` as
+    separator, e.g. ``script_binding scriptname/bindingname``.
+
+    For completeness, here is how this command works internally. The details
+    could change any time. On any matching key event, ``script_message_to``
+    or ``script_message`` is called (depending on whether the script name is
+    included), where the first argument is the string ``key-binding``, the
+    second argument is the name of the binding, and the third argument is the
+    key state as string. The key state consists of a number of letters. The
+    first letter is one of ``d`` (key was pressed down), ``u`` (was released),
+    ``r`` (key is still down, and was repeated; only if key repeat is enabled
+    for this binding), ``p`` (key was pressed; happens if up/down can't be
+    tracked). The second letter whether the event originates from the mouse,
+    either ``m`` (mouse button) or ``-`` (something else).
+
+``ab_loop``
+    Cycle through A-B loop states. The first command will set the ``A`` point
+    (the ``ab-loop-a`` property); the second the ``B`` point, and the third
+    will clear both points.
+
+``vo_cmdline "<args>"``
+    Reset the sub-option of the current VO. Currently works with ``opengl``
+    (including ``opengl-hq``). The argument is the sub-option string usually
+    passed to the VO on the command line. Not all sub-options can be set, but
+    those which can will be reset even if they don't appear in the argument.
+    This command might be changed or removed in the future.
+
+``drop_buffers``
+    Drop audio/video/demuxer buffers, and restart from fresh. Might help with
+    unseekable streams that are going out of sync.
+    This command might be changed or removed in the future.
+
+Undocumented commands: ``tv_last_channel`` (TV/DVB only), ``get_property`` (?),
+``ao_reload`` (experimental/internal).
 
 Input Command Prefixes
 ----------------------
@@ -527,7 +605,9 @@ prefixes can be specified. They are separated by whitespace.
     Do not expand properties in string arguments. (Like ``"${property-name}"``.)
 ``expand-properties`` (default)
     All string arguments are expanded as described in `Property Expansion`_.
-
+``repeatable``
+    For some commands, keeping a key pressed doesn't run the command repeatedly.
+    This prefix forces enabling key repeat in any case.
 
 All of the osd prefixes are still overridden by the global ``--osd-level``
 settings.
@@ -594,14 +674,23 @@ Property list
     ``${stream-end}``. For ordered chapters and such, the
     size of the currently played segment is returned.)
 
+``estimated-frame-count``
+    Total number of frames in current file.
+
+    .. note:: This is only an estimate. (It's computed from two unreliable
+              quantities: fps and stream length.)
+
+``estimated-frame-number``
+    Number of current frame in current stream.
+
+    .. note:: This is only an estimate. (It's computed from two unreliable
+              quantities: fps and possibly rounded timestamps.)
+
 ``path``
     Full path of the currently played file.
 
 ``media-title``
-    If libquvi is used and libquvi returns a page title for the currently
-    played URL, return the page title.
-
-    Otherwise, if the currently played file has a ``title`` tag, use that.
+    If the currently played file has a ``title`` tag, use that.
 
     Otherwise, if the media type is DVD, return the volume ID of DVD.
 
@@ -635,8 +724,13 @@ Property list
     disabled.
 
 ``drop-frame-count``
-    Frames dropped because they arrived to late. Unavailable if video
-    is disabled
+    Video frames dropped by decoder, because video is too far behind audio (when
+    using ``--framedrop=decoder``). Sometimes, this may be incremented in other
+    situations, e.g. when video packets are damaged, or the decoder doesn't
+    follow the usual rules. Unavailable if video is disabled.
+
+``vo-drop-frame-count``
+    Frames dropped by VO (when using ``--framedrop=vo``).
 
 ``percent-pos`` (RW)
     Position in current file (0-100). The advantage over using this instead of
@@ -644,9 +738,6 @@ Property list
     estimating the playback position from the byte position, if the file
     duration is not known.
 
-``ratio-pos`` (RW)
-    Position in current file (0.0-1.0). higher precision that ``percent-pos``.
-
 ``time-pos`` (RW)
     Position in current file in seconds.
 
@@ -723,6 +814,10 @@ Property list
                 "title"             MPV_FORMAT_STRING
                 "default"           MPV_FORMAT_FLAG
 
+``ab-loop-a``, ``ab-loop-b`` (TW)
+    Set/get A-B loop points. See corresponding options and ``ab_loop`` command.
+    The special value ``no`` on either of these properties disables looping.
+
 ``angle`` (RW)
     Current DVD angle.
 
@@ -796,6 +891,10 @@ Property list
     be different ``pause`` in special situations, such as when the player
     pauses itself due to low network cache.
 
+    This also returns ``yes`` if playback is restarting or if nothing is
+    playing at all. In other words, it's only ``no`` if there's actually
+    video playing. (Behavior since mpv 0.7.0.)
+
 ``cache``
     Network cache fill state (0-100.0).
 
@@ -821,9 +920,22 @@ Property list
     Returns ``yes`` if the cache is idle, which means the cache is filled as
     much as possible, and is currently not reading more data.
 
+``demuxer-cache-duration``
+    Approximate duration of video buffered in the demuxer, in seconds. The
+    guess is very unreliable, and often the property will not be available
+    at all, even if data is buffered.
+
+``demuxer-cache-idle``
+    Returns ``yes`` if the demuxer is idle, which means the demuxer cache is
+    filled to the requested amount, and is currently not reading more data.
+
 ``paused-for-cache``
     Returns ``yes`` when playback is paused because of waiting for the cache.
 
+``cache-buffering-state``
+    Return the percentage (0-100) of the cache fill status until the player
+    will unpause (related to ``paused-for-cache``).
+
 ``eof-reached``
     Returns ``yes`` if end of playback was reached, ``no`` otherwise. Note
     that this is usually interesting only if ``--keep-open`` is enabled,
@@ -885,6 +997,9 @@ Property list
 ``deinterlace`` (RW)
     See ``--deinterlace``.
 
+``field-dominance`` (RW)
+    See ``--field-dominance``
+
 ``colormatrix`` (RW)
     See ``--colormatrix``.
 
@@ -1007,13 +1122,17 @@ Property list
 ``dwidth``, ``dheight``
     Video display size. This is the video size after filters and aspect scaling
     have been applied. The actual video window size can still be different
-    from this.
+    from this, e.g. if the user resized the video window manually.
+
+    These have the same values as ``video-out-params/dw`` and
+    ``video-out-params/dh``.
 
 ``video-out-params``
     Same as ``video-params``, but after video filters have been applied. If
     there are no video filters in use, this will contain the same values as
     ``video-params``. Note that this is still not necessarily what the video
-    window uses, since all real VOs do their own scaling.
+    window uses, since the user can change the window size, and all real VOs
+    do their own scaling independently from the filter chain.
 
     Has the same sub-properties as ``video-params``.
 
@@ -1033,9 +1152,16 @@ Property list
     Window size multiplier. Setting this will resize the video window to the
     values contained in ``dwidth`` and ``dheight`` multiplied with the value
     set with this property. Setting ``1`` will resize to original video size
-    (or to be exactly, the size the video filters output). ``2`` will set the
+    (or to be exact, the size the video filters output). ``2`` will set the
     double size, ``0.5`` halves the size.
 
+``window-minimized``
+    Return whether the video window is minimized or not.
+
+``display-names``
+    Names of the displays that the mpv window covers. On X11, these
+    are the xrandr names (LVDS1, HDMI1, DP1, VGA1, etc.).
+
 ``video-aspect`` (RW)
     Video aspect, see ``--video-aspect``.
 
@@ -1102,8 +1228,8 @@ Property list
     See ``--ass-style-override``.
 
 ``stream-capture`` (RW)
-    A filename, see ``--capture``. Setting this will start capture using the
-    given filename. Setting it to an empty string will stop it.
+    A filename, see ``--stream-capture``. Setting this will start capture using
+    the given filename. Setting it to an empty string will stop it.
 
 ``tv-brightness``, ``tv-contrast``, ``tv-saturation``, ``tv-hue`` (RW)
     TV stuff.
@@ -1128,6 +1254,13 @@ Property list
     ``playlist/N/filename``
         Filename of the Nth entry.
 
+    ``playlist/N/current``, ``playlist/N/playing``
+        ``yes`` if this entry is currently playing (or being loaded).
+        Unavailable or ``no`` otherwise. When changing files, ``current`` and
+        ``playing`` can be different, because the currently playing file hasn't
+        been unloaded yet; in this case, ``current`` refers to the new
+        selection. (Since mpv 0.7.0.)
+
     When querying the property with the client API using ``MPV_FORMAT_NODE``,
     or with Lua ``mp.get_property_native``, this will return a mpv_node with
     the following contents:
@@ -1137,6 +1270,8 @@ Property list
         MPV_FORMAT_NODE_ARRAY
             MPV_FORMAT_NODE_MAP (for each playlist entry)
                 "filename"  MPV_FORMAT_STRING
+                "current"   MPV_FORMAT_FLAG (might be missing; since mpv 0.7.0)
+                "playing"   MPV_FORMAT_FLAG (same)
 
 ``track-list``
     List of audio/video/sub tracks, current entry marked. Currently, the raw
@@ -1234,11 +1369,6 @@ Property list
                 "title" MPV_FORMAT_STRING
                 "time"  MPV_FORMAT_DOUBLE
 
-``quvi-format`` (RW)
-    See ``--quvi-format``. Cycling this property (``cycle``) will attempt to
-    cycle through known format, although currently this feature doesn't work
-    well at all.
-
 ``af`` (RW)
     See ``--af`` and the ``af`` command.
 
@@ -1261,15 +1391,141 @@ Property list
 
     It's also possible to write the property using this format.
 
+``video-rotate`` (RW)
+    See ``--video-rotate`` option.
+
 ``seekable``
     Return whether it's generally possible to seek in the current file.
 
+``playback-abort``
+    Return whether playback is stopped or is to be stopped. (Useful in obscure
+    situations like during ``on_load`` hook processing, when the user can
+    stop playback, but the script has to explicitly end processing.)
+
+``cursor-autohide`` (RW)
+    See ``--cursor-autohide``. Setting this to a new value will always update
+    the cursor, and reset the internal timer.
+
+``osd-sym-cc``
+    Inserts the current OSD symbol as opaque OSD control code (cc). This makes
+    sense only with the ``show_text`` command or options which set OSD messages.
+    The control code is implementation specific and is useless for anything else.
+
+``osd-ass-cc``
+    ``${osd-ass-cc/0}`` disables escaping ASS sequences of text in OSD,
+    ``${osd-ass-cc/1}`` enables it again. By default, ASS sequences are
+    escaped to avoid accidental formatting, and this property can disable
+    this behavior. Note that the properties return an opaque OSD control
+    code, which only makes sense for the ``show_text`` command or options
+    which set OSD messages.
+
+    .. admonition:: Example
+
+        --osd-status-msg='This is ${osd-ass-cc/0}{\\b1}bold text'
+
+    Any ASS override tags as understood by libass can be used.
+
+    Note that you need to escape the ``\`` character, because the string is
+    processed for C escape sequences before passing it to the OSD code.
+
+``vo-configured``
+    Return whether the VO is configured right now. Usually this corresponds to
+    whether the video window is visible. If the ``--force-window`` option is
+    used, this is usually always returns ``yes``.
+
+``audio-device-list``
+    Return the list of discovered audio devices. This is mostly for use with
+    the client API, and reflects what ``--audio-device=help`` with the command
+    line player returns.
+
+    When querying the property with the client API using ``MPV_FORMAT_NODE``,
+    or with Lua ``mp.get_property_native``, this will return a mpv_node with
+    the following contents:
+
+    ::
+
+        MPV_FORMAT_NODE_ARRAY
+            MPV_FORMAT_NODE_MAP (for each device entry)
+                "name"          MPV_FORMAT_STRING
+                "description"   MPV_FORMAT_STRING
+
+    The ``name`` is what is to be passed to the ``--audio-device`` option (and
+    often a rather cryptic audio API-specific ID), while ``description`` is
+    human readable free form text. The description is an empty string if none
+    was received.
+
+``audio-device`` (RW)
+    Set the audio device. This directly reads/writes the ``--audio-device``
+    option, but on write accesses, the audio output will be scheduled for
+    reloading.
+
+    Writing this property while no audio output is active will not automatically
+    enable audio. (This is also true in the case when audio was disabled due to
+    reinitialization failure after a previous write access to ``audio-device``.)
+
+    This property also doesn't tell you which audio device is actually in use.
+
+    How these details are handled may change in the future.
+
+``mpv-version``
+    Return the mpv version/copyright string. Depending on how the binary was
+    built, it might contain either a release version, or just a git hash.
+
+``mpv-configuration``
+    Return the configuration arguments which were passed to the build system
+    (typically the way ``./waf configure ...`` was invoked).
+
 ``options/<name>`` (RW)
     Read-only access to value of option ``--<name>``. Most options can be
     changed at runtime by writing to this property. Note that many options
     require reloading the file for changes to take effect. If there is an
     equivalent property, prefer setting the property instead.
 
+``file-local-options/<name>``
+    Similar to ``options/<name>``, but when setting an option through this
+    property, the option is reset to its old value once the current file has
+    stopped playing. Trying to write an option while no file is playing (or
+    is being loaded) results in an error.
+
+    (Note that if an option is marked as file-local, even ``options/`` will
+    access the local value, and the ``old`` value, which will be restored on
+    end of playback, can not be read or written until end of playback.)
+
+``option-info/<name>``
+    Additional per-option information.
+
+    This has a number of sub-properties. Replace ``<name>`` with the name of
+    a top-level option. No guarantee of stability is given to any of these
+    sub-properties - they may change radically in the feature.
+
+    ``option-info/<name>/name``
+        Returns the name of the option.
+
+    ``option-info/<name>/type``
+        Return the name of the option type, like ``String`` or ``Integer``.
+        For many complex types, this isn't very accurate.
+
+    ``option-info/<name>/set-from-commandline``
+
+        Return ``yes`` if the option was set from the mpv command line,
+        ``no`` otherwise. What this is set to if the option is e.g. changed
+        at runtime is left undefined (meaning it could change in the future).
+
+    ``option-info/<name>/default-value``
+        The default value of the option. May not always be available.
+
+    ``option-info/<name>/min``, ``option-info/<name>/max``
+        Integer minimum and maximum values allowed for the option. Only
+        available if the options are numeric, and the minimum/maximum has been
+        set internally. It's also possible that only one of these is set.
+
+    ``option-info/<name>/choices``
+        If the option is a choice option, the possible choices. Choices that
+        are integers may or may not be included (they can be implied by ``min``
+        and ``max``). Note that options which behave like choice options, but
+        are not actual choice options internally, may not have this info
+        available.
+
 ``property-list``
     Return the list of top-level properties.
 

DOCS/man/ipc.rst

@@ -0,0 +1,183 @@
+JSON IPC
+========
+
+mpv can be controlled by external programs using the JSON-based IPC protocol. It
+can be enabled by specifying the path to a unix socket using the option
+``--input-unix-socket``. Clients can connect to this socket and send commands to
+the player or receive events from it. You can use the ``socat`` tool to send
+commands (and receive reply) from the shell.
+
+.. warning::
+
+    This is not intended to be a secure network protocol. It is explicitly
+    insecure: there is no authentication, no encryption, and the commands
+    themselves are insecure too. For example, the ``run`` command is exposed,
+    which can run arbitrary system commands. The use-case is controlling the
+    player locally. This is not different from the MPlayer slave protocol.
+
+Protocol
+--------
+
+Clients can execute commands on the player by sending JSON messages of the
+following form:
+
+::
+
+    { "command": ["command_name", "param1", "param2", ...] }
+
+where ``command_name`` is the name of the command to be executed, followed by a
+list of parameters. Parameters must be formatted as native JSON values
+(integers, strings, booleans, ...). Every message **must** be terminated with
+``\n``. Additionally, ``\n`` must not appear anywhere inside the message. In
+practice this means that messages should be minified before being sent to mpv.
+
+mpv will then send back a reply indicating whether the command was run
+correctly, and an additional field holding the command-specific return data (it
+can also be null).
+
+::
+
+    { "error": "success", "data": null }
+
+mpv will also send events to clients with JSON messages of the following form:
+
+::
+
+    { "event": "event_name" }
+
+where ``event_name`` is the name of the event. Additional event-specific fields
+can also be present. See `List of events`_ for a list of all supported events.
+
+If the first character (after skipping whitespace) is not ``{``, the command
+will be interpreted as non-JSON text command, as they are used in input.conf
+(or ``mpv_command_string()`` in the client API). Additionally, line starting
+with ``#`` and empty lines are ignored.
+
+Currently, embedded 0 bytes terminate the current line, but you should not
+rely on this.
+
+Commands
+--------
+
+Additionally to  the commands described in `List of Input Commands`_, a few
+extra commands can also be used as part of the protocol:
+
+``client_name``
+    Return the name of the client as string. This is the string ``ipc-N`` with
+    N being an integer number.
+
+``get_time_us``
+    Return the current mpv internal time in microseconds as a number. This is
+    basically the system time, with an arbitrary offset.
+
+``get_property``
+    Return the value of the given property. The value will be sent in the data
+    field of the replay message.
+
+    Example:
+
+    ::
+
+        { "command": ["get_property", "volume"] }
+        { "data": 50.0, "error": "success" }
+
+``get_property_string``
+    Like ``get_property``, but the resulting data will always be a string.
+
+    Example:
+
+    ::
+
+        { "command": ["get_property_string", "volume"] }
+        { "data": "50.000000", "error": "success" }
+
+``set_property``
+    Set the given property to the given value. See `Properties`_ for more
+    information about properties.
+
+    Example:
+
+    ::
+
+        { "command": ["set_property", "pause", true] }
+        { "error": "success" }
+
+``set_property_string``
+    Like ``set_property``, but the argument value must be passed as string.
+
+    Example:
+
+    ::
+
+        { "command": ["set_property_string", "pause", "yes"] }
+        { "error": "success" }
+
+``observe_property``
+    Watch a property for changes. If the given property is changed, then an
+    event of type ``property-change`` will be generated
+
+    Example:
+
+    ::
+
+        { "command": ["observe_property", 1, "volume"] }
+        { "error": "success" }
+        { "event": "property-change", "id": 1, "data": 52.0, "name": "volume" }
+
+``observe_property_string``
+    Like ``observe_property``, but the resulting data will always be a string.
+
+    Example:
+
+    ::
+
+        { "command": ["observe_property_string", 1, "volume"] }
+        { "error": "success" }
+        { "event": "property-change", "id": 1, "data": "52.000000", "name": "volume" }
+
+``unobserve_property``
+    Undo ``observe_property`` or ``observe_property_string``. This requires the
+    numeric id passed to the observe command as argument.
+
+    Example:
+
+    ::
+
+        { "command": ["unobserve_property", 1] }
+        { "error": "success" }
+
+``request_log_messages``
+    Enable output of mpv log messages. They will be received as events. The
+    parameter to this command is the log-level (see ``mpv_request_log_messages``
+    C API function).
+
+    Log message output is meant for humans only (mostly for debugging).
+    Attempting to retrieve information by parsing these messages will just
+    lead to breakages with future mpv releases. Instead, make a feature request,
+    and ask for a proper event that returns the information you need.
+
+``suspend``
+    Suspend the mpv main loop. There is a long-winded explanation of this in
+    the C API function ``mpv_suspend()``. In short, this prevents the player
+    from displaying the next video frame, so that you don't get blocked when
+    trying to access the player.
+
+``resume``
+    Undo one ``suspend`` call. ``suspend`` increments an internal counter, and
+    ``resume`` decrements it. When 0 is reached, the player is actually resumed.
+
+``get_version``
+    Returns the client API version the C API of the remote mpv instance
+    provides. (Also see ``DOCS/client-api-changes.rst``.)
+
+UTF-8
+-----
+
+Normally, all strings are in UTF-8. Sometimes it can happen that strings are
+in some broken encoding (often happens with file tags and such, and filenames
+on many Unixes are not required to be in UTF-8 either). This means that mpv
+sometimes sends invalid JSON. If that is a problem for the client application's
+parser, it should filter the raw data for invalid UTF-8 sequences and perform
+the desired replacement, before feeding the data to its JSON parser.
+
+mpv will not attempt to construct invalid UTF-8 with broken escape sequences.

DOCS/man/lua.rst

@@ -1,10 +1,16 @@
 LUA SCRIPTING
 =============
 
-mpv can load Lua scripts. Scripts in ``~/.mpv/lua/`` will be loaded on program
-start, or if passed to ``--lua``. mpv provides the built-in module ``mp``, which
-provides functions to send commands to the mpv core and to retrieve information
-about playback state, user settings, file information, and so on.
+mpv can load Lua scripts. Scripts passed to the ``--lua`` option, or found in
+the ``lua`` subdirectory of the mpv configuration directory (usually
+``~/.config/mpv/lua/``) will be loaded on program start. mpv also appends the
+``lua`` subdirectory to the end of Lua's path so you can import scripts from
+there too. Since it's added to the end, don't name scripts you want to import
+the same as Lua libraries because they will be overshadowed by them.
+
+mpv provides the built-in module ``mp``, which provides functions to send
+commands to the mpv core and to retrieve information about playback state, user
+settings, file information, and so on.
 
 These scripts can be used to control mpv in a similar way to slave mode.
 Technically, the Lua code uses the client API internally.
@@ -27,12 +33,13 @@ allow users implement features which are not going to be added to the mpv core.
 Mode of operation
 -----------------
 
-Your script will be loaded by the player at program start from ``~/.mpv/lua/``,
-or ``--lua``, or in some cases, internally (like ``--osc``). Each script runs
-in its own thread. Your script is first run "as is", and once that is done,
-the event loop is entered. This event loop will dispatch events received by mpv
-and call your own event handlers which you have registered with
-``mp.register_event``, or timers added with ``mp.add_timeout`` or similar.
+Your script will be loaded by the player at program start from the ``lua``
+configuration subdirectory, from a path specified with the ``--lua`` option, or
+in some cases, internally (like ``--osc``). Each script runs in its own
+thread. Your script is first run "as is", and once that is done, the event loop
+is entered. This event loop will dispatch events received by mpv and call your
+own event handlers which you have registered with ``mp.register_event``, or
+timers added with ``mp.add_timeout`` or similar.
 
 When the player quits, all scripts will be asked to terminate. This happens via
 a ``shutdown`` event, which by default will make the event loop return. If your
@@ -47,6 +54,15 @@ The event loop will wait for events and dispatch events registered with
 ``mp.register_event``. It will also handle timers added with ``mp.add_timeout``
 and similar (by waiting with a timeout).
 
+Since mpv 0.6.0, the player will wait until the script is fully loaded before
+continuing normal operation. The player considers a script as fully loaded as
+soon as it starts waiting for mpv events (or it exits). In practice this means
+the player will more or less hang until the script returns from the main chunk
+(and ``mp_event_loop`` is called), or the script calls ``mp_event_loop`` or
+``mp.dispatch_events`` directly. This is done to make it possible for a script
+to fully setup event handlers etc. before playback actually starts. In older
+mpv versions, this happened asynchronously.
+
 mp functions
 ------------
 
@@ -78,6 +94,15 @@ The ``mp`` module is preloaded, although it can be loaded manually with
     the ``expand-properties`` prefix, or the ``mp.get_property`` family of
     functions.
 
+``mp.command_native(table [,def])``
+    Similar to ``mp.commandv``, but pass the argument list as table. This has
+    the advantage that in at least some cases, arguments can be passed as
+    native types.
+
+    Returns a result table on success (usually empty), or ``def, error`` on
+    error. ``def`` is the second parameter provided to the function, and is
+    nil if it's missing.
+
 ``mp.get_property(name [,def])``
     Return the value of the given property as string. These are the same
     properties as used in input.conf. See `Properties`_ for a list of
@@ -156,7 +181,7 @@ The ``mp`` module is preloaded, although it can be loaded manually with
     Return the current mpv internal time in seconds as a number. This is
     basically the system time, with an arbitrary offset.
 
-``mp.add_key_binding(key, name|fn [,fn])``
+``mp.add_key_binding(key, name|fn [,fn [,flags]])``
     Register callback to be run on a key binding. The binding will be mapped to
     the given ``key``, which is a string describing the physical key. This uses
     the same key names as in input.conf, and also allows combinations
@@ -173,8 +198,23 @@ The ``mp`` module is preloaded, although it can be loaded manually with
     overwritten. You can omit the name, in which case a random name is generated
     internally.
 
-    Internally, key bindings are dispatched via the ``script_message_to`` input
-    command and ``mp.register_script_message``.
+    The last argument is used for optional flags. This is a table, which can
+    have the following entries:
+
+        ``repeatable``
+            If set to ``true``, enables key repeat for this specific binding.
+
+        ``complex``
+            If set to ``true``, then ``fn`` is called on both key up and down
+            events (as well as key repeat, if enabled), with the first
+            argument being a table. This table has an ``event`` entry, which
+            is set to one of the strings ``down``, ``repeat``, ``up`` or
+            ``press`` (the latter if key up/down can't be tracked). It further
+            has an ``is_mouse`` entry, which tells whether the event was caused
+            by a mouse button.
+
+    Internally, key bindings are dispatched via the ``script_message_to`` or
+    ``script_binding`` input commands and ``mp.register_script_message``.
 
     Trying to map multiple commands to a key will essentially prefer a random
     binding, while the other bindings are not called. It is guaranteed that
@@ -197,7 +237,7 @@ The ``mp`` module is preloaded, although it can be loaded manually with
 
     ::
 
-        y script_message something
+        y script_binding something
 
 
     This will print the message when the key ``y`` is pressed. (``x`` will
@@ -208,7 +248,7 @@ The ``mp`` module is preloaded, although it can be loaded manually with
 
     ::
 
-        y script_message_to fooscript something
+        y script_binding fooscript.something
 
 ``mp.add_forced_key_binding(...)``
     This works almost the same as ``mp.add_key_binding``, but registers the
@@ -486,6 +526,13 @@ This built-in module provides generic helper functions for Lua, and have
 strictly speaking nothing to do with mpv or video/audio playback. They are
 provided for convenience. Most compensate for Lua's scarce standard library.
 
+Be warned that any of these functions might disappear any time. They are not
+strictly part of the guaranteed API.
+
+``utils.getcwd()``
+    Returns the directory that mpv was launched from. On error, ``nil, error``
+    is returned.
+
 ``utils.readdir(path [, filter])``
     Enumerate all entries at the given path on the filesystem, and return them
     as array. Each entry is a directory entry (without the path).
@@ -522,6 +569,60 @@ provided for convenience. Most compensate for Lua's scarce standard library.
     Return the concatenation of the 2 paths. Tries to be clever. For example,
     if ```p2`` is an absolute path, p2 is returned without change.
 
+``utils.subprocess(t)``
+    Runs an external process and waits until it exits. Returns process status
+    and the captured output.
+
+    The paramater ``t`` is a table. The function reads the following entries:
+
+        ``args``
+            Array of strings. The first array entry is the executable. This
+            can be either an absolute path, or a filename with no path
+            components, in which case the ``PATH`` environment variable is
+            used to resolve the executable. The other array elements are
+            passed as command line arguments.
+
+        ``cancellable``
+            Optional. If set to ``true`` (default), then if the user stops
+            playback or goes to the next file while the process is running,
+            the process will be killed.
+
+        ``max_size``
+            Optional. The maximum size in bytes of the data that can be captured
+            from stdout. (Default: 16 MB.)
+
+    The function returns a table as result with the following entries:
+
+        ``status``
+            The raw exit status of the process. It will be negative on error.
+
+        ``stdout``
+            Captured output stream as string, limited to ``max_size``.
+
+        ``error``
+            ``nil`` on success. The string ``killed`` if the process was
+            terminated in an unusual way. The string ``init`` if the process
+            could not be started.
+
+            On Windows, ``killed`` is only returned when the process has been
+            killed by mpv as a result of ``cancellable`` being set to ``true``.
+
+    In all cases, ``mp.resume_all()`` is implicitly called.
+
+``utils.parse_json(str [, trail])``
+    Parses the given string argument as JSON, and returns it as a Lua table. On
+    error, returns ``nil, error``. (Currently, ``error`` is just a string
+    reading ``error``, because there is no fine-grained error reporting of any
+    kind.)
+
+    The returned value uses similar conventions as ``mp.get_property_native()``
+    to distinguish empty objects and arrays.
+
+    If the ``trail`` parameter is ``true`` (or any value equal to ``true``),
+    then trailing non-whitespace text is tolerated by the function, and the
+    trailing text is returned as 3rd return value. (The 3rd return value is
+    always there, but with ``trail`` set, no error is raised.)
+
 Events
 ------
 
@@ -563,28 +664,11 @@ List of events
 ``playback-restart``
     Start of playback after seek or after file was loaded.
 
-``tracks-changed``
-    The list of video/audio/sub tracks was updated. (This happens on playback
-    start, and very rarely during playback.)
-
-``track-switched``
-    A video/audio/subtitle track was switched on or off. This usually happens
-    when the user (or a script) changes the subtitle track and so on.
-
 ``idle``
     Idle mode is entered. This happens when playback ended, and the player was
     started with ``--idle`` or ``--force-window``. This mode is implicitly ended
     when the ``start-file`` or ``shutdown`` events happen.
 
-``pause``
-    Playback was paused. This also happens when for example the player is
-    paused on low network cache. Then the event type indicates the pause state
-    (like the property "pause" as opposed to the "core-idle" property), and you
-    might receive multiple ``pause`` events in a row.
-
-``unpause``
-    Playback was unpaused. See above for details.
-
 ``tick``
     Called after a video frame was displayed. This is a hack, and you should
     avoid using it. Use timers instead and maybe watch pausing/unpausing events
@@ -644,8 +728,32 @@ List of events
 ``audio-reconfig``
     Happens on audio output or filter reconfig.
 
-``metadata-update``
-    Metadata (like file tags) was updated.
+The following events also happen, but are deprecated: ``tracks-changed``,
+``track-switched``, ``pause``, ``unpause``, ``metadata-update``,
+``chapter-change``. Use ``mp.observe_property()`` instead.
 
-``chapter-change``
-    The current chapter possibly changed.
+Extras
+------
+
+This documents experimental features, or features that are "too special" and
+we don't guarantee a stable interface to it.
+
+``mp.add_hook(type, priority, fn)``
+    Add a hook callback for ``type`` (a string identifying a certain kind of
+    hook). These hooks allow the player to call script functions and wait for
+    their result (normally, the Lua scripting interface is asynchronous from
+    the point of view of the player core). ``priority`` is an arbitrary integer
+    that allows ordering among hooks of the same kind. Using the value 50 is
+    recommended as neutral default value. ``fn`` is the function that will be
+    called during execution of the hook.
+
+    Currently existing hooks:
+
+    ``on_load``
+        Called when a file is to be opened, before anything is actually done.
+        For example, you could read and write the ``stream-open-filename``
+        property to redirect an URL to something else (consider support for
+        streaming sites which rarely give the user a direct media URL), or
+        you could set per-file options with by setting the property
+        ``file-local-options/<option name>``. The player will wait until all
+        hooks are run.

DOCS/man/mpv.rst

@@ -12,8 +12,7 @@ a media player
 SYNOPSIS
 ========
 
-| **mpv** [options] [file|URL|-]
-| **mpv** [options] --playlist=PLAYLIST
+| **mpv** [options] [file|URL|PLAYLIST|-]
 | **mpv** [options] files
 
 DESCRIPTION
@@ -47,6 +46,10 @@ UP and DOWN
     Seek forward/backward 1 minute. Shift+arrow does a 5 second exact seek (see
     ``--hr-seek``).
 
+Ctrl+LEFT and Ctrl+RIGHT
+    Seek to the previous/next subtitle. Subject to some restrictions and
+    might not work always; see ``sub_seek`` command.
+
 [ and ]
     Decrease/increase current playback speed by 10%.
 
@@ -60,7 +63,7 @@ BACKSPACE
     Go backward/forward in the playlist.
 
 ENTER
-    Go forward in the playlist, even over the end.
+    Go forward in the playlist.
 
 p / SPACE
     Pause (pressing again unpauses).
@@ -73,7 +76,7 @@ p / SPACE
     Step backward. Pressing once will pause, every consecutive press will
     play one frame in reverse and then go into pause mode again.
 
-q / ESC
+q
     Stop playing and quit.
 
 Q
@@ -98,6 +101,9 @@ m
 f
     Toggle fullscreen (see also ``--fs``).
 
+ESC
+    Exit fullscreen mode.
+
 T
     Toggle stay-on-top (see also ``--ontop``).
 
@@ -123,6 +129,12 @@ j and J
 x and z
     Adjust subtitle delay by +/- 0.1 seconds.
 
+l
+    Set/clear A-B loop points. See ``ab_loop`` command for details.
+
+Ctrl + and Ctrl -
+    Adjust audio delay by +/- 0.1 seconds.
+
 V
     Toggle subtitle VSFilter aspect compatibility mode. See
     ``--ass-vsfilter-aspect-compat`` for more info.
@@ -145,6 +157,10 @@ PGUP and PGDWN
     "previous" will actually go to the beginning of the current chapter; see
     ``--chapter-seek-threshold``.
 
+Shift+PGUP and Shift+PGDWN
+    Seek backward or forward by 10 minutes. (This used to be mapped to
+    PGUP/PGDWN without Shift.)
+
 D
     Activate/deactivate deinterlacer.
 
@@ -255,7 +271,7 @@ need to escape special characters. To work this around, the path can be
 additionally wrapped in the ``%n%string_of_length_n`` syntax (see above).
 
 Some mpv options interpret paths starting with ``~``. Currently, the prefix
-``~~/`` expands to the mpv configuration directory (usually ``~/.mpv/``).
+``~~/`` expands to the mpv configuration directory (usually ``~/.config/mpv/``).
 ``~/`` expands to the user's home directory. (The trailing ``/`` is always
 required.)
 
@@ -345,11 +361,11 @@ File-specific Configuration Files
 You can also write file-specific configuration files. If you wish to have a
 configuration file for a file called 'video.avi', create a file named
 'video.avi.conf' with the file-specific options in it and put it in
-``~/.mpv/``. You can also put the configuration file in the same directory
-as the file to be played, as long as you give the ``--use-filedir-conf``
+``~/.config/mpv/``. You can also put the configuration file in the same directory
+as the file to be played. Both require you to set the ``--use-filedir-conf``
 option (either on the command line or in your global config file). If a
 file-specific configuration file is found in the same directory, no
-file-specific configuration is loaded from ``~/.mpv``. In addition, the
+file-specific configuration is loaded from ``~/.config/mpv``. In addition, the
 ``--use-filedir-conf`` option enables directory-specific configuration files.
 For this, mpv first tries to load a mpv.conf from the same directory
 as the file played and then tries to load any file-specific configuration.
@@ -404,6 +420,52 @@ taking screenshots is handled by the VOs, and adding the screenshot filter will
 break hardware decoding. (The filter may still be useful for taking screenshots
 at a certain point within the video chain when using multiple video filters.)
 
+TERMINAL STATUS LINE
+====================
+
+During playback, mpv shows the playback status on the terminal. It looks like
+something like this:
+
+    ``AV: 00:03:12 / 00:24:25 (13%) A-V: -0.000``
+
+The status line can be overridden with the ``--term-status-msg`` option.
+
+The following is a list of things that can show up in the status line. Input
+properties, that can be used to get the same information manually, are also
+listed.
+
+- ``AV:`` or ``V:`` (video only) or ``A:`` (audio only)
+- The current time position in ``HH:MM:SS`` format (``playback-time`` property)
+- The total file duration (or ``00:00:00`` if unknown) (``length`` property)
+- Playback speed, e.g. `` x2.0``. Only visible if the speed is not normal. This
+  is the user-requested speed, and not the actual speed  (usually they should
+  be the same, unless playback is too slow). (``speed`` property.)
+- Playback percentage, e.g. ``(13%)``. How much of the file has been played.
+  Normally calculated out of playback position and duration, but can fallback
+  to other methods (like byte position) if these are not available.
+  (``percent-pos`` property.)
+- The audio/video sync as ``A-V:  0.000``. This is the difference between
+  audio and video time. Normally it should be 0 or close to 0. If it's growing,
+  it might indicate a playback problem. (``avsync`` property.)
+- Total A/V sync change, e.g. ``ct: -0.417``. Normally invisible. Can show up
+  if there is audio "missing", or not enough frames can be dropped. Usually
+  this will indicate a problem. (``total-avsync-change`` property.)
+- Encoding state in ``{...}``, only shown in encoding mode.
+- Dropped frames, e.g. ``Dropped: 4``. Shows up only if the count is not 0. Can
+  grow if the video framerate is higher than that of the display, or if video
+  rendering is too slow. Also can be incremented on "hiccups" and when the video
+  frame couldn't be displayed on time. (``vo-drop-frame-count`` property.)
+  If the decoder drops frames, the number of decoder-dropped frames is appended
+  to the display as well, e.g.: ``Dropped: 4/34``. This should almost never
+  happen, unless decoder-framedropping is enabled with one of the
+  ``--framedrop`` options, the stream contains errors, or a weird codec is in
+  use. (``drop-frame-count`` property.)
+- Cache state, e.g. ``Cache:  2s+134KB``. Visible if the stream cache is enabled.
+  The first value shows the amount of video buffered in the demuxer in seconds,
+  the second value shows *additional* data buffered in the stream cache in
+  kilobytes. (``demuxer-cache-duration`` and ``cached-used`` properties.)
+
+
 PROTOCOLS
 =========
 
@@ -412,6 +474,10 @@ PROTOCOLS
     be specified. mpv will never attempt to guess whether a filename is
     actually a network address. A protocol prefix is always required.
 
+    Note that not all prefixes are documented here. Undocumented prefixes are
+    either aliases to documented protocols, or are just redirections to
+    protocols implemented and documented in FFmpeg.
+
 ``-``
     Play data from stdin.
 
@@ -501,6 +567,8 @@ PROTOCOLS
 
 .. include:: lua.rst
 
+.. include:: ipc.rst
+
 .. include:: changes.rst
 
 ENVIRONMENT VARIABLES
@@ -656,6 +724,39 @@ override the standard directory ``~/.config/mpv/``.
 Also, the old config location at ``~/.mpv/`` is still read, and if the XDG
 variant does not exist, will still be preferred.
 
+FILES ON WINDOWS
+================
+
+On win32 (if compiled with MinGW, but not Cygwin), the default config file
+locations are different:
+
+``%APPDATA%/mpv/mpv.conf``
+    Preferred mpv config file. This maps to a system and user-specific path,
+    for example ``C:\users\USERNAME\Application Data\mpv\mpv.conf``. You can
+    find the exact path by running ``echo %APPDATA%\mpv\mpv.conf`` in cmd.exe.
+
+``%APPDATA%/mpv/input.conf``
+    key bindings (see `INPUT.CONF`_ section)
+
+``%APPDATA%/mpv/lua/``
+    equivalent of ``~/.config/mpv/lua/`` on Unix.
+
+The environment variable ``$MPV_HOME`` completely overrides these, like on
+UNIX.
+
+Config files located in the same directory as ``mpv.exe`` are loaded with
+lower priority. Some config files are loaded only once, which means that
+e.g. of 2 ``input.conf`` files located in two config directories, only the
+one from the directory with higher priority will be loaded.
+
+A third config directory with lowest priority is the directory named ``mpv``
+in the same directory as ``mpv.exe``. This used to be the directory with
+highest priority, but is now discouraged to use and might be removed in the
+future.
+
+Note that mpv likes to mix ``/`` and ``\`` path separators for simplicity.
+kernel32.dll accepts this, but cmd.exe does not.
+
 EXAMPLES OF MPV USAGE
 =====================
 

DOCS/man/options.rst

@@ -45,6 +45,15 @@ Track Selection
 ``--vid=<ID|auto|no>``
     Select video channel. ``auto`` selects the default, ``no`` disables video.
 
+``--ff-aid=<ID|auto|no>``, ``--ff-sid=<ID|auto|no>``, ``--ff-vid=<ID|auto|no>``
+    Select audio/subtitle/video streams by the FFmpeg stream index. The FFmpeg
+    stream index is relatively arbitrary, but useful when interacting with
+    other software using FFmpeg (consider ``ffprobe``).
+
+    Note that with external tracks (added with ``--sub-file`` and similar
+    options) will have duplicate IDs. In that case, the first stream in order
+    is selected.
+
 ``--edition=<ID|auto>``
     (Matroska files only)
     Specify the edition (set of chapters) to use, where 0 is the first. If set
@@ -95,6 +104,9 @@ Playback Control
 ``--speed=<0.01-100>``
     Slow down or speed up playback by the factor given as parameter.
 
+    If ``--audio-pitch-correction`` is used, playing with a speed higher than
+    normal automatically inserts the ``scaletempo`` audio filter.
+
 ``--loop=<N|inf|no>``
     Loops playback ``N`` times. A value of ``1`` plays it one time (default),
     ``2`` two times, etc. ``inf`` means forever. ``no`` is the same as ``1`` and
@@ -116,21 +128,23 @@ Playback Control
     no format is detected, t will be treated as list of files, separated by
     newline characters. Note that XML playlist formats are not supported.)
 
+    You can play playlists directly and without this option, however, this
+    option disables any security mechanisms that might be in place. You may
+    also need this option to load plaintext files as playlist.
+
     .. warning::
 
-        The way mpv uses playlist files is not safe against maliciously
-        constructed files. Such files may trigger harmful actions.
+        The way mpv uses playlist files via ``--playlist`` is not safe against
+        maliciously constructed files. Such files may trigger harmful actions.
         This has been the case for all mpv and MPlayer versions, but
         unfortunately this fact was not well documented earlier, and some people
         have even misguidedly recommended use of ``--playlist`` with untrusted
         sources. Do NOT use ``--playlist`` with random internet sources or files
         you do not trust!
 
-        The main problem is that playlists can point to arbitrary network
-        addresses (including local addresses inside of your LAN), and thus
-        can't be considered secure. Playlists also can contain entries using
-        other protocols, such as local files, or (most severely), special
-        protocols like ``avdevice://``, which are inherently unsafe.
+        Playlist can contain entries using other protocols, such as local files,
+        or (most severely), special protocols like ``avdevice://``, which are
+        inherently unsafe.
 
 ``--chapter-merge-threshold=<number>``
     Threshold for merging almost consecutive ordered chapter parts in
@@ -198,26 +212,28 @@ Playback Control
         (i.e. not with stdin, pipe, etc).
 
 ``--load-unsafe-playlists``
-    Normally, something like ``mpv playlist.m3u`` won't load the playlist. This
-    is because the playlist code is unsafe. (This is the same in all other
-    variations of MPlayer.)
+    Load URLs from playlists which are considered unsafe (default: no). This
+    includes special protocols and anything that doesn't refer to normal files.
+    Local files and HTTP links on the other hand are always considered safe.
 
-    See ``--playlist`` for details.
+    Note that ``--playlist`` always loads all entries, so you use that instead
+    if you really have the need for this functionality.
 
-    Note: this option will allow opening playlists using the ``playlist``
-    special demuxer. The ``--playlist`` uses different code, and supports more
-    playlist formats than the playlist demuxer. This means that for now, the
-    ``--playlist`` option should always be used if you intend to open playlists.
-    Background: the special demuxer contains newly written code, while the
-    ``--playlist`` option uses the old MPlayer code. Adding support for more
-    playlist formats to the special demuxer is work in progress, and eventually
-    the old code should disappear.
+``--loop-file=<N|inf|no>``
+    Loop a single file N times. ``inf`` means forever, ``no`` means normal
+    playback. For compatibility, ``--loop-file`` and ``--loop-file=yes`` are
+    also accepted, and are the same as ``--loop-file=inf``.
 
-``--loop-file``
-    Loop a single file. The difference to ``--loop=inf`` is that this doesn't
-    loop the playlist, just the file itself. If the playlist contains only a
-    single file, the difference between the two option is that this option
-    performs a seek on loop, instead of reloading the file.
+    The difference to ``--loop`` is that this doesn't loop the playlist, just
+    the file itself. If the playlist contains only a single file, the difference
+    between the two option is that this option performs a seek on loop, instead
+    of reloading the file.
+
+``--ab-loop-a=<time>``, ``--ab-loop-b=<time>``
+    Set loop points. If playback passes the ``b`` timestamp, it will seek to
+    the ``a`` timestamp. Seeking past the ``b`` point doesn't loop (this is
+    intentional). The loop-points can be adjusted at runtime with the
+    corresponding properties. See also ``ab_loop`` command.
 
 ``--ordered-chapters``, ``--no-ordered-chapters``
     Enabled by default.
@@ -237,6 +253,10 @@ Playback Control
     Note: a playlist can be as simple as a text file containing filenames
     separated by newlines.
 
+``--chapters-file=<filename>``
+    Load chapters from this file, instead of using the chapter metadata found
+    in the main file.
+
 ``--sstep=<sec>``
     Skip <sec> seconds after every frame.
 
@@ -244,6 +264,12 @@ Playback Control
 
         Without ``--hr-seek``, skipping will snap to keyframes.
 
+``--stop-playback-on-init-failure=<yes|no>``
+    Stop playback if either audio or video fails to initialize. Currently,
+    the default behavior is ``no`` for the command line player, but ``yes``
+    for libmpv. With ``no``, playback will continue in video-only or audio-only
+    mode if one of them fails. This doesn't affect playback of audio-only or
+    video-only files.
 
 Program Behavior
 ----------------
@@ -259,9 +285,8 @@ Program Behavior
     Print version string and exit.
 
 ``--no-config``
-    Do not load default configuration files. This prevents loading of
-    ``~/.config/mpv/mpv.conf`` and ``~/.config/mpv/input.conf``, as well as
-    loading the same files from system wide configuration directories. Other
+    Do not load default configuration files. This prevents loading of both the
+    user-level and system-wide ``mpv.conf`` and ``input.conf`` files. Other
     configuration files are blocked as well, such as resume playback files.
 
     .. note::
@@ -317,7 +342,8 @@ Program Behavior
     Specify configuration file to be parsed after the default ones.
 
 ``--load-scripts=<yes|no>``
-    If set to ``no``, don't auto-load scripts from ``~/.mpv/lua/``.
+    If set to ``no``, don't auto-load scripts from the ``lua`` configuration
+    subdirectory (usually ``~/.config/mpv/lua/``).
     (Default: ``yes``)
 
 ``--lua=<filename>``
@@ -332,14 +358,15 @@ Program Behavior
 ``--merge-files``
     Pretend that all files passed to mpv are concatenated into a single, big
     file. This uses timeline/EDL support internally. Note that this won't work
-    for ordered chapter files or quvi-resolved URLs (such as YouTube links).
+    for ordered chapter files.
 
     This option is interpreted at program start, and doesn't affect for
     example files or playlists loaded with the ``loadfile`` or ``loadlist``
     commands.
 
 ``--no-resume-playback``
-    Do not restore playback position from ``~/.mpv/watch_later/``.
+    Do not restore playback position from the ``watch_later`` configuration
+    subdirectory (usually ``~/.config/mpv/watch_later/``).
     See ``quit_watch_later`` input command.
 
 ``--profile=<profile1,profile2,...>``
@@ -395,6 +422,24 @@ Program Behavior
 
         May be dangerous if playing from untrusted media.
 
+``--ytdl``, ``--no-ytdl``
+    Enable the youtube-dl hook-script. It will look at the input URL, and will
+    play the video located on the website. This works with many streaming sites,
+    not just the one the scripts are named after. This requires a recent version
+    of youtube-dl to be installed on the system.
+
+    If the script can't do anything with an URL, it will do nothing.
+
+    Currently disabled by default, because youtube-dl's generic extractor can
+    get stuck on some URL, preventing playback.
+
+    (Note: this is the replacement for the now removed libquvi support.)
+
+``--ytdl-format=<best|worst|mp4|webm|...>``
+    Video format/quality that is directly passed to youtube-dl. The possible
+    values are specific to the website and the video, for a given url the
+    available formats can be found with the command
+    ``youtube-dl --list-formats URL``.
 
 Video
 -----
@@ -430,16 +475,45 @@ Video
     Do not sleep when outputting video frames. Useful for benchmarks when used
     with ``--no-audio.``
 
-``--framedrop=<no|yes>``
-    Skip displaying some frames to maintain A/V sync on slow systems. Video
-    filters are not applied to such frames. For B-frames even decoding is
-    skipped completely. May produce unwatchably choppy output.
+``--framedrop=<mode>``
+    Skip displaying some frames to maintain A/V sync on slow systems, or
+    playing high framerate video on video outputs that have an upper framerate
+    limit.
 
-    The ``--vd-lavc-framedrop`` option controls what frames to drop.
+    The argument selects the drop methods, and can be one of the following:
+
+    <no>
+        Disable any framedropping.
+    <vo>
+        Drop late frames on video output (default). This still decodes and
+        filters all frames, but doesn't render them on the VO. It tries to query
+        the display FPS (X11 only, not correct on multi-monitor systems), or
+        assumes infinite display FPS if that fails. Drops are indicated in
+        the terminal status line as ``D:`` field. If the decoder is too slow,
+        in theory all frames would have to be dropped (because all frames are
+        too late) - to avoid this, frame dropping stops if the effective
+        framerate is below 10 FPS.
+    <decoder>
+        Old, decoder-based framedrop mode. (This is the same as ``--framedrop=yes``
+        in mpv 0.5.x and before.) This tells the decoder to skip frames (unless
+        they are needed to decode future frames). May help with slow systems,
+        but can produce unwatchably choppy output, or even freeze the display
+        complete. Not recommended.
+        The ``--vd-lavc-framedrop`` option controls what frames to drop.
+    <decoder+vo>
+        Enable both modes. Not recommended.
 
     .. note::
 
-        Practical use of this feature is questionable. Disabled by default.
+        ``--vo=vdpau`` has its own code for the ``vo`` framedrop mode. Slight
+        differences to other VOs are possible.
+
+``--display-fps=<fps>``
+    Set the maximum assumed display FPS used with ``--framedrop``. By default
+    a detected value is used (X11 only, not correct on multi-monitor systems),
+    or infinite display FPS if that fails. Infinite FPS means only frames too
+    late are dropped. If a correct FPS is provided, frames that are predicted
+    to be too late are dropped too.
 
 ``--hwdec=<api>``
     Specify the hardware video decoding API that should be used if possible.
@@ -453,16 +527,19 @@ Video
     :vdpau:     requires ``--vo=vdpau`` or ``--vo=opengl`` (Linux only)
     :vaapi:     requires ``--vo=opengl`` or ``--vo=vaapi`` (Linux with Intel GPUs only)
     :vaapi-copy: copies video back into system RAM (Linux with Intel GPUs only)
-    :vda:       requires ``--vo=opengl`` or ``--vo=corevideo`` (OS X only)
+    :vda:       requires ``--vo=opengl`` (OS X only)
+    :dxva2-copy: copies video back to system RAM (Windows only) (experimental)
 
     ``auto`` tries to automatically enable hardware decoding using the first
     available method. This still depends what VO you are using. For example,
-    if you are not using ``--vo=vdpau``, vdpau decoding will never be enabled.
-    Also note that if the first found method doesn't actually work, it will
-    always fall back to software decoding, instead of trying the next method.
+    if you are not using ``--vo=vdpau`` or ``--vo=opengl``, vdpau decoding will
+    never be enabled. Also note that if the first found method doesn't actually
+    work, it will always fall back to software decoding, instead of trying the
+    next method (might matter on some Linux systems).
 
-    The ``vaapi-copy`` function allows you to use vaapi with any VO. Because
-    this copies the decoded video back to system RAM, it's quite inefficient.
+    The ``vaapi-copy`` mode allows you to use vaapi with any VO. Because
+    this copies the decoded video back to system RAM, it's likely less efficient
+    than the ``vaapi`` mode.
 
     .. note::
 
@@ -531,6 +608,22 @@ Video
     which means the value ``0`` would rotate the video according to the
     rotation metadata.)
 
+``--video-stereo-mode=<mode>``
+    Set the stereo 3D output mode (default: ``mono``). This is done by inserting
+    the ``stereo3d`` conversion filter.
+
+    The mode ``mono`` is an alias to ``ml``, which refers to the left frame in
+    2D. This is the default, which means mpv will try to show 3D movies in 2D,
+    instead of the mangled 3D image not intended for consumption (such as
+    showing the left and right frame side by side, etc.).
+
+    The pseudo-mode ``none`` disables automatic conversion completely.
+
+    Use ``--video-stereo-mode=help`` to list all available modes. Check with
+    the ``stereo3d`` filter documentation to see what the names mean. Note that
+    some names refer to modes not supported by ``stereo3d`` - these modes can
+    appear in files, but can't be handled properly by mpv.
+
 ``--video-zoom=<value>``
     Adjust the video display scale factor by the given value. The unit is in
     fractions of the (scaled) window video size.
@@ -621,31 +714,6 @@ Video
         ``mpv --hwdec=vdpau --vo=vdpau --hwdec-codecs=h264,mpeg2video``
             Enable vdpau decoding for h264 and mpeg2 only.
 
-``--quvi-format=<best|default|...>``
-    Video format/quality that is directly passed to libquvi (default: ``best``).
-    This is used when opening links to streaming sites like YouTube. The
-    interpretation of this value is highly specific to the streaming site and
-    the video.
-
-    libquvi 0.4.x:
-
-        The only well-defined values that work on all sites are ``best``
-        (best quality/highest bandwidth, default), and ``default`` (lowest
-        quality).
-
-        The quvi command line tool can be used to find out which formats are
-        supported for a given URL: ``quvi --query-formats URL``.
-
-    libquvi 0.9.x:
-
-        The following explanations are relevant:
-        `<http://quvi.sourceforge.net/r/api/0.9/glossary_termino.html#m_stream_id>`_
-
-    The ``quvi-format`` property can be used at runtime to cycle through the
-    list of formats. Unfortunately, this is slow. On libquvi 0.4.x, this
-    functionality is limited to switching between ``best`` and ``default`` if
-    the ``cycle`` input command is used.
-
 ``--vd-lavc-check-hw-profile=<yes|no>``
     Check hardware decoder profile (default: yes). If ``no`` is set, the
     highest profile of the hardware decoder is unconditionally selected, and
@@ -719,6 +787,30 @@ Video
 Audio
 -----
 
+``--audio-pitch-correction=<yes|no>``
+    If this is enabled (default), playing with a speed higher than normal
+    automatically inserts the ``scaletempo`` audio filter. For details, see
+    audio filter section.
+
+``--audio-device=<name>``
+    Use the given audio device. This consists of the audio output name, e.g.
+    ``alsa``, followed by ``/``, followed by the audio output specific device
+    name.
+
+    You can list audio devices with ``--audio-device=help``. This outputs the
+    device name in quotes, followed by a description. The device name is what
+    you have to pass to the ``--audio-device`` option.
+
+    The default value for this option is ``auto``, which tries every audio
+    output in preference order with the default device.
+
+    Note that many AOs have a ``device`` sub-option, which overrides the
+    device selection of this option (but not the audio output selection).
+    Likewise, forcing an AO with ``--ao`` will override the audio output
+    selection of ``--audio-device`` (but not the device selecton).
+
+    Currently not implemented for most AOs.
+
 ``--ao=<driver1[:suboption1[=value]:...],driver2,...[,]>``
     Specify a priority list of audio output drivers to be used. For
     interactive use one would normally specify a single one to use, but in
@@ -847,8 +939,9 @@ Audio
     lists speaker names, which can be used to express arbitrary channel
     layouts (e.g. ``fl-fr-lfe`` is 2.1).
 
-    You can use ``--audio-channels=empty`` to disable this. In this case, the AO
-    use the channel layout as the audio filter chain indicates.
+    You can use ``--audio-channels=auto`` to disable this. In this case, the AO
+    use the channel layout as the audio filter chain indicates. (``empty`` is
+    an accepted obsolete for ``auto``.)
 
     This will also request the channel layout from the decoder. If the decoder
     does not support the layout, it will fall back to its native channel layout.
@@ -885,7 +978,7 @@ Audio
     of the current media, the lavrresample audio filter will be inserted into
     the audio filter layer to compensate for the difference.
 
-``--gapless-audio=<no|yes|weak``
+``--gapless-audio=<no|yes|weak>``
     Try to play consecutive audio files with no silence or disruption at the
     point of file change. Default: ``weak``.
 
@@ -940,9 +1033,10 @@ Audio
         maximum amplification, i.e. amplify by 200%. The default volume (no
         change in volume) will be ``50`` in this case.
 
-``--volstep=<0-100>``
-    Set the step size of mixer volume changes in percent of the full range
-    (default: 3).
+``--audio-client-name=<name>``
+    The application name the player reports to the audio API. Can be useful
+    if you want to force a different audio profile (e.g. with PulseAudio),
+    or to set your own application name when using libmpv.
 
 ``--volume-restore-data=<string>``
     Used internally for use by playback resume (e.g. with ``quit_watch_later``).
@@ -956,6 +1050,20 @@ Audio
 
     Do not use.
 
+``--audio-buffer=<seconds>``
+    Set the audio output minimum buffer. The audio device might actually create
+    a larger buffer if it pleases. If the device creates a smaller buffer,
+    additional audio is buffered in an additional software buffer.
+
+    Making this larger will make soft-volume and other filters react slower,
+    introduce additional issues on playback speed change, and block the
+    player on audio format changes. A smaller buffer might lead to audio
+    dropouts.
+
+    This option should be used for testing only. If a non-default value helps
+    significantly, the mpv developers should be contacted.
+
+    Default: 0.2 (200 ms).
 
 Subtitles
 ---------
@@ -1014,7 +1122,7 @@ Subtitles
         This affects ASS subtitles as well, and may lead to incorrect subtitle
         rendering. Use with care, or use ``--sub-text-font-size`` instead.
 
-``--sub-scale-with-window=yes|no``
+``--sub-scale-with-window=<yes|no>``
     Make the subtitle font size relative to the window, instead of the video.
     This is useful if you always want the same font size, even if the video
     doesn't covert the window fully, e.g. because screen aspect and window
@@ -1057,7 +1165,7 @@ Subtitles
 
         Using this option may lead to incorrect subtitle rendering.
 
-``--ass-hinting=none|light|normal|native``
+``--ass-hinting=<none|light|normal|native>``
     Set font hinting type. <type> can be:
 
     :none:       no hinting (default)
@@ -1075,7 +1183,7 @@ Subtitles
 ``--ass-line-spacing=<value>``
     Set line spacing value for SSA/ASS renderer.
 
-``--ass-shaper=simple|complex``
+``--ass-shaper=<simple|complex>``
     Set the text layout engine used by libass.
 
     :simple:   uses Fribidi only, fast, doesn't render some languages correctly
@@ -1099,6 +1207,7 @@ Subtitles
     :yes:   Apply all the ``--ass-*`` style override options. Changing the default
             for any of these options can lead to incorrect subtitle rendering
             (default).
+    :signfs: like ``yes``, but apply ``--sub-scale`` only to signs
     :no:    Render subtitles as forced by subtitle scripts.
     :force: Try to force the font style as defined by the ``--sub-text-*``
             options. Requires a modified libass, can break rendering easily.
@@ -1159,15 +1268,6 @@ Subtitles
     a subtitle script with another video file. The ``--ass-style-override``
     option doesn't affect how this option is interpreted.
 
-``--quvi-fetch-subtitles=<yes|no>``
-    Toggles fetching of subtitles from streaming sites with libquvi. Disabled
-    by default, because it's unreliable and slow. Note that when enabled,
-    subtitles will always be fetched, even if subtitles are explicitly
-    disabled with ``--no-sub`` (because you might want to enable subtitles
-    at runtime).
-
-    Supported when using libquvi 0.9.x.
-
 ``--stretch-dvd-subs=<yes|no>``
     Stretch DVD subtitles when playing anamorphic videos for better looking
     fonts on badly mastered DVDs. This switch has no effect when the
@@ -1316,12 +1416,19 @@ Subtitles
         - ``/path/to/video/sub/``
         - ``/path/to/video/subtitles/``
         - ``/tmp/subs/``
-        - ``~/.mpv/sub/``
+        -  the ``sub`` configuration subdirectory (usually ``~/.config/mpv/sub/``)
 
 ``--sub-visibility``, ``--no-sub-visibility``
     Can be used to disable display of subtitles, but still select and decode
     them.
 
+``--sub-clear-on-seek``
+    (Obscure, rarely useful.) Can be used to play broken mkv files with
+    duplicate ReadOrder fields. ReadOrder is the first field in a
+    Matroska-style ASS subtitle packets. It should be unique, and libass
+    uses it for fast elimination of duplicates. This option disables caching
+    of subtitles across seeks, so after a seek libass can't eliminate subtitle
+    packets with the same ReadOrder as earlier packets.
 
 Window
 ------
@@ -1372,7 +1479,8 @@ Window
     See also ``--screen``.
 
 ``--keep-open``
-    Do not terminate when playing or seeking beyond the end of the file.
+    Do not terminate when playing or seeking beyond the end of the file, and
+    there is not next file to be played (and ``--loop`` is not used).
     Instead, pause the player. When trying to seek beyond end of the file, the
     player will pause at an arbitrary playback position (or, in corner cases,
     not redraw the window at all).
@@ -1385,6 +1493,12 @@ Window
         chapter will terminate playback as well, even if ``--keep-open`` is
         given.
 
+    Since mpv 0.6.0, this doesn't pause if there is a next file in the playlist,
+    or the playlist is looped. Approximately, this will pause when the player
+    would normally exit, but in practice there are corner cases in which this
+    is not the case (e.g. ``mpv --keep-open file.mkv /dev/null`` will play
+    file.mkv normally, then fail to open ``/dev/null``, then exit).
+
 ``--force-window``
     Create a video output window even if there is no video. This can be useful
     when pretending that mpv is a GUI application. Currently, the window
@@ -1549,13 +1663,6 @@ Window
     there is a change in video parameters, video stream or file. This used to
     be the default behavior. Currently only affects X11 VOs.
 
-``--fs-missioncontrol``
-    (OS X only)
-    Use OS X Mission Control's fullscreen feature instead of the custom one
-    provided by mpv. This can potentially break a lot of stuff like
-    ``--geometry`` and is disabled by default. On the other hand it provides
-    a more 'OS X-like' user experience.
-
 ``--heartbeat-cmd=<command>``
     Command that is executed every 30 seconds during playback via *system()* -
     i.e. using the shell. The time between the commands can be customized with
@@ -1591,12 +1698,19 @@ Window
     .. note::
 
         This does not affect the normal screensaver operation in any way.
-        
+
 ``--no-keepaspect``, ``--keepaspect``
     ``--no-keepaspect`` will always stretch the video to window size, and will
     disable the window manager hints that force the window aspect ratio.
     (Ignored in fullscreen mode.)
 
+``--no-keepaspect-window``, ``--keepaspect-window``
+    ``--keepaspect-window`` (the default) will lock the window size to the
+    video aspect. ``--no-keepaspect-window`` disables this behavior, and will
+    instead add black bars if window aspect and video aspect mismatch. Whether
+    this actually works depends on the VO backend.
+    (Ignored in fullscreen mode.)
+
 ``--monitoraspect=<ratio>``
     Set the aspect ratio of your monitor or TV screen. A value of 0 disables a
     previous setting (e.g. in the config file). Overrides the
@@ -1624,14 +1738,25 @@ Window
     to to work this around using ``--heartbeat-cmd`` instead.
 
 ``--wid=<ID>``
-    (X11 and Windows only)
-    This tells mpv to attach to an existing window. The ID is interpreted as
-    "Window" on X11, and as HWND on Windows. If a VO is selected that supports
-    this option, a new window will be created and the given window will be set
-    as parent. The window will always be resized to cover the parent window
-    fully, and will add black bars to compensate for the video aspect ratio.
+    This tells mpv to attach to an existing window. If a VO is selected that
+    supports this option, it will use that window for video output. mpv will
+    scale the video to the size of this window, and will add black bars to
+    compensate if the aspect ratio of the video is different.
 
-    See also ``--slave-broken``.
+    On X11, the ID is interpreted as a ``Window`` on X11. Unlike
+    MPlayer/mplayer2, mpv always creates its own window, and sets the wid
+    window as parent. The window will always be resized to cover the parent
+    window fully. The value ``0`` is interpreted specially, and mpv will
+    draw directly on the root window.
+
+    On win32, the ID is interpreted as ``HWND``. Pass it as value cast to
+    ``intptr_t``. mpv will create its own window, and set the wid window as
+    parent, like with X11.
+
+    On OSX/Cocoa. the ID is interpreted as ``NSView*``. Pass it as value cast
+    to ``intptr_t``. mpv will creates its own sub-view. Because OSX does not
+    support window embedding of foreign processes, this works only with libmpv,
+    and will crash when used from the command line.
 
 ``--no-window-dragging``
     Don't move the window when clicking on it and moving the mouse pointer.
@@ -1639,16 +1764,26 @@ Window
 ``--x11-name``
     Set the window class name for X11-based video output methods.
 
-``--x11-netwm=no``
+``--x11-netwm=<yes|no|auto>``
     (X11 only)
-    Disable use of the NetWM protocol when switching to or from fullscreen.
+    Control the use of NetWM protocol features.
+
     This may or may not help with broken window managers. This provides some
     functionality that was implemented by the now removed ``--fstype`` option.
     Actually, it is not known to the developers to which degree this option
     was needed, so feedback is welcome.
 
-    By default, NetWM support is autodetected, and using this option forces
-    autodetection to fail.
+    Specifically, ``yes`` will force use of NetWM fullscreen support, even if
+    not advertised by the WM. This can be useful for WMs that are broken on
+    purpose, like XMonad. (XMonad supposedly doesn't advertise fullscreen
+    support, because Flash uses it. Apparently, applications which want to
+    use fullscreen anyway are supposed to either ignore the NetWM support hints,
+    or provide a workaround. Shame on XMonad for deliberately breaking X
+    protocols (as if X isn't bad enough already).
+
+    By default, NetWM support is autodetected (``auto``).
+
+    This option might be removed in the future.
 
 
 Disc Devices
@@ -1708,7 +1843,7 @@ Disc Devices
     Add ``<value>`` sectors to the values reported when addressing tracks.
     May be negative.
 
-``--cdda-skip=<es|no``
+``--cdda-skip=<yes|no>``
     (Never) accept imperfect data reconstruction.
 
 ``--dvd-speed=<speed>``
@@ -1927,6 +2062,22 @@ Demuxer
     and hr-seeks, and this option changes behavior with relative or imprecise
     seeks only.
 
+    You can use the ``--demuxer-mkv-subtitle-preroll-secs`` option to specify
+    how mach data the demuxer should pre-read at most in order to find subtitle
+    packets that may overlap. Setting this to 0 will effectively disable this
+    preroll mechanism. Setting a very large value can make seeking very slow,
+    and an extremely large value would completely reread the entire file from
+    start to seek target on every seek - seeking can become slower towards the
+    end of the file. The details are messy, and the value is actually rounded
+    down to the cluster with the previous video keyframe.
+
+    Some files, especially files muxed with newer mkvmerge versions, have
+    information embedded that can be used to determine what subtitle packets
+    overlap with a seek target. In these cases, mpv will reduce the amount
+    of data read to a minimum. (Although it will still read *all* data between
+    the cluster that contains the first wanted subtitle packet, and the seek
+    target.)
+
     See also ``--hr-seek-demuxer-offset`` option. This option can achieve a
     similar effect, but only if hr-seek is active. It works with any demuxer,
     but makes seeking much slower, as it has to decode audio and video data
@@ -1934,12 +2085,23 @@ Demuxer
 
     ``--mkv-subtitle-preroll`` is a deprecated alias.
 
+``--demuxer-mkv-subtitle-preroll-secs=<value>``
+    See ``--demuxer-mkv-subtitle-preroll``.
+
+``--demuxer-mkv-probe-video-duration``
+    When opening the file, seek to the end of it, and check what timestamp the
+    last video packet has, and report that as file duration. This is strictly
+    for compatibility with Haali only. In this mode, it's possible that opening
+    will be slower (especially when playing over http), or that behavior with
+    broken files is much worse. So don't use this option.
+
 ``--demuxer-rawaudio-channels=<value>``
     Number of channels (or channel layout) if ``--demuxer=rawaudio`` is used
     (default: stereo).
 
 ``--demuxer-rawaudio-format=<value>``
     Sample format for ``--demuxer=rawaudio`` (default: s16le).
+    Use ``--demuxer-rawaudio-format=help`` to get a list of all formats.
 
 ``--demuxer-rawaudio-rate=<value>``
     Sample rate for ``--demuxer=rawaudio`` (default: 44 kHz).
@@ -1975,20 +2137,32 @@ Demuxer
 
 ``--demuxer-thread=<yes|no>``
     Run the demuxer in a separate thread, and let it prefetch a certain amount
-    of packets (default: no).
+    of packets (default: yes). Having this enabled may lead to smoother
+    playback, but on the other hand can add delays to seeking or track
+    switching.
 
-``--demuxer-readahead-packets=N``
+``--demuxer-readahead-secs=<seconds>``
     If ``--demuxer-thread`` is enabled, this controls how much the demuxer
-    should buffer ahead. If the number of packets in the packet queue exceeds
-    ``--demuxer-readahead-packets``, or the total number of bytes exceeds
-    ``--demuxer-readahead-bytes``, the thread stops reading ahead.
+    should buffer ahead in seconds (default: 0.2). As long as no packet has
+    a timestamp difference higher than the readahead amount relative to the
+    last packet returned to the decoder, the demuxer keeps reading.
+
+    (This value tends to be fuzzy, because many file formats don't store linear
+    timestamps.)
+
+``--demuxer-readahead-packets=<packets>``
+    If ``--demuxer-thread`` is enabled, this controls how much the demuxer
+    should buffer ahead. As long as the number of packets in the packet queue
+    doesn't exceed ``--demuxer-readahead-packets``, and the total number of
+    bytes doesn't exceed ``--demuxer-readahead-bytes``, the thread keeps
+    reading ahead.
 
     Note that if you set these options near the maximum, you might get a
     packet queue overflow warning.
 
     See ``--list-options`` for defaults and value range.
 
-``--demuxer-readahead-bytes=N``
+``--demuxer-readahead-bytes=<bytes>``
     See ``--demuxer-readahead-packets``.
 
 
@@ -2008,8 +2182,8 @@ Input
     Number of key presses to generate per second on autorepeat.
 
 ``--input-conf=<filename>``
-    Specify input configuration file other than the default
-    ``~/.config/mpv/input.conf``.
+    Specify input configuration file other than the default location in the mpv
+    configuration directory (usually ``~/.config/mpv/input.conf``).
 
 ``--no-input-default-bindings``
     Disable mpv default (built-in) key bindings.
@@ -2039,7 +2213,11 @@ Input
     like any other binding). See `INPUT.CONF`_.
 
 ``--input-file=<filename>``
-    Read commands from the given file. Mostly useful with a FIFO.
+    Read commands from the given file. Mostly useful with a FIFO. Since
+    mpv 0.7.0 also understands JSON commands (see `JSON IPC`_), but you can't
+    get replies or events. Use ``--input-unix-socket`` for something
+    bi-directional. On MS Windows, JSON commands are not available.
+
     See also ``--slave-broken``.
 
     .. note::
@@ -2056,8 +2234,16 @@ Input
     or intend to read from stdin later on via the loadfile or loadlist slave
     commands.
 
-``--input-appleremote``, ``--no-input-appleremote``
-    Enable/disable AppleIR remote support. Enabled by default.
+``--input-unix-socket=<filename>``
+    Enable the IPC support and create the listening socket at the given path.
+
+    See `JSON IPC`_ for details.
+
+    Not available on MS Windows.
+
+``--input-appleremote=<yes|no>``
+    (OS X only)
+    Enable/disable Apple Remote support. Enabled by default (except for libmpv).
 
 ``--input-cursor``, ``--no-input-cursor``
     Permit mpv to receive pointer events reported by the video output
@@ -2077,14 +2263,41 @@ Input
     (LIRC only)
     Specifies a configuration file for LIRC (default: ``~/.lircrc``).
 
-``--input-media-keys``, ``--no-input-media-keys``
-    OS X only: Enabled by default. Enables/disable media keys support.
+``--input-media-keys=<yes|no>``
+    (OS X only)
+    Enable/disable media keys support. Enabled by default (except for libmpv).
 
 ``--input-right-alt-gr``, ``--no-input-right-alt-gr``
     (Cocoa and Windows only)
     Use the right Alt key as Alt Gr to produce special characters. If disabled,
     count the right Alt as an Alt modifier key. Enabled by default.
 
+``--input-vo-keyboard=<yes|no>``
+    Disable all keyboard input on for VOs which can't participate in proper
+    keyboard input dispatching. This currently affects X11. Generally useful for
+    embedding only.
+
+    On X11, a sub-window with input enabled grabs all keyboard input as long
+    as it is 1. a child of a focused window, and 2. the mouse is inside of
+    the sub-window. The can steal away all keyboard input from the
+    application embedding the mpv window, and on the other hand, the mpv
+    window will receive no input if the mouse is outside of the mpv window,
+    even though mpv has focus. Modern toolkits work around this weird X11
+    behavior, but naively embedding foreign windows breaks it.
+
+    The only way to handle this reasonably is using the XEmbed protocol, which
+    was designed to solve these problems. But Qt has questionable support, and
+    mpv doesn't implement it yet.
+
+    As a workaround, this option is disabled by default in libmpv. (Note that
+    ``input-default-bindings`` is disabled by default in libmpv as well.)
+
+    (This option was renamed from ``--input-x11-keyboard``.)
+
+``--input-app-events=<yes|no>``
+    (OS X only)
+    Enable/disable application wide keyboard events so that keyboard shortcuts
+    can be processed without a window. Enabled by default (except for libmpv).
 
 OSD
 ---
@@ -2124,7 +2337,27 @@ OSD
     height: if the window height is larger or smaller than 720, the actual size
     of the text increases or decreases as well.
 
-    Default: 45.
+    Default: 55.
+
+``--osd-msg1=<string>``
+    Show this string as message on OSD with OSD level 1 (visible by default).
+    The message will be visible by default, and as long no other message
+    covers it, and the OSD level isn't changed (see ``--osd-level``).
+    Expands properties; see `Property Expansion`_.
+
+``--osd-msg2=<string>``
+    Similar as ``--osd-msg1``, but for OSD level 2. If this is an empty string
+    (default), then the playback time is shown.
+
+``--osd-msg3=<string>``
+    Similar as ``--osd-msg1``, but for OSD level 3. If this is an empty string
+    (default), then the playback time, duration, and some more information is
+    shown.
+
+    This is also used for the ``show_progress`` command (by default mapped to
+    ``P``), or in some non-default cases when seeking.
+
+    ``--osd-status-msg`` is a legacy equivalent (but with a minor difference).
 
 ``--osd-status-msg=<string>``
     Show a custom string during playback instead of the standard status text.
@@ -2133,6 +2366,17 @@ OSD
     non-default cases when seeking. Expands properties. See
     `Property Expansion`_.
 
+    This option has been replaced with ``--osd-msg3``. The only difference is
+    that this option implicitly includes ``${osd-sym-cc}``. This option is
+    ignored if ``--osd-msg3`` is not empty.
+
+``--osd-playing-msg=<string>``
+    Show a message on OSD when playback starts. The string is expanded for
+    properties, e.g. ``--osd-playing-msg='file: ${filename}'`` will show the
+    message ``file:`` followed by a space and the currently played filename.
+
+    See `Property Expansion`_.
+
 ``--osd-bar-align-x=<-1-1>``
     Position of the OSD bar. -1 is far left, 0 is centered, 1 is far right.
     Fractional values (like 0.5) are allowed.
@@ -2167,7 +2411,7 @@ OSD
     Size of the OSD/sub font border in scaled pixels (see ``--osd-font-size``
     for details). A value of 0 disables borders.
 
-    Default: 2.5.
+    Default: 3.
 
 ``--osd-color=<color>``, ``--sub-text-color=<color>``
     Specify the color used for OSD/unstyled text subtitles.
@@ -2225,12 +2469,12 @@ OSD
     for unstyled text subtitles. If you just want to raise the vertical
     subtitle position, use ``--sub-pos``.
 
-    Default: 10.
+    Default: 22.
 
 ``--osd-scale=<factor>``
     OSD font size multiplier, multiplied with ``--osd-font-size`` value.
 
-``--osd-scale-by-window=yes|no``
+``--osd-scale-by-window=<yes|no>``
     Whether to scale the OSD with the window size (default: yes). If this is
     disabled, ``--osd-font-size`` and other OSD options that use scaled pixels
     are always in actual pixels. The effect is that changing the window size
@@ -2252,6 +2496,16 @@ OSD
 
     Default: 0.
 
+``--use-text-osd=<yes|no>``
+    Disable text OSD rendering completely. (This includes the complete OSC as
+    well.) This is mostly useful for avoiding loading fontconfig in situations
+    where fontconfig does not behave well, and OSD is unused - this could for
+    example allow GUI programs using libmpv to workaround fontconfig issues.
+
+    Note that selecting subtitles of any kind still initializes fontconfig.
+
+    Default: ``no``.
+
 
 Screenshot
 ----------
@@ -2697,20 +2951,6 @@ Cache
     will not automatically enable the cache e.g. when playing from a network
     stream. Note that using ``--cache`` will always override this option.
 
-``--cache-pause-below=<kBytes|no>``
-    If the cache size goes below the specified value (in KB), pause and wait
-    until the size set by ``--cache-pause-restart`` is reached, then  resume
-    playback (default: 50). If ``no`` is specified, this behavior is disabled.
-
-    When the player is paused this way, the status line shows ``Buffering``
-    instead of ``Paused``, and the OSD uses a clock symbol instead of the
-    normal paused symbol.
-
-``--cache-pause-restart=<kBytes>``
-    If the cache is paused due to the ``--cache-pause-below`` functionality,
-    then the player unpauses as soon as the cache has this much data (in KB).
-    (Default: 100)
-
 ``--cache-initial=<kBytes>``
     Playback will start when the cache has been filled up with this many
     kilobytes of data (default: 0).
@@ -2727,22 +2967,40 @@ Cache
     on the situation, either of these might be slower than the other method.
     This option allows control over this.
 
-``--cache-file=<path>``
-    Create a cache file on the filesystem with the given name. The file is
-    always overwritten. When the general cache is enabled, this file cache
-    will be used to store whatever is read from the source stream.
+``--cache-file=<TMP|path>``
+    Create a cache file on the filesystem.
 
-    This will always overwrite the cache file, and you can't use an existing
-    cache file to resume playback of a stream. (Technically, mpv wouldn't
-    even know which blocks in the file are valid and which not.)
+    There are two ways of using this:
 
-    The resulting file will not necessarily contain all data of the source
-    stream. For example, if you seek, the parts that were skipped over are
-    never read and consequently are not written to the cache. The skipped over
-    parts are filled with zeros. This means that the cache file doesn't
-    necessarily correspond to a full download of the source stream.
+    1. Passing a path (a filename). The file will always be overwritten. When
+       the general cache is enabled, this file cache will be used to store
+       whatever is read from the source stream.
 
-    Both of these issues could be improved if there is any user interest.
+       This will always overwrite the cache file, and you can't use an existing
+       cache file to resume playback of a stream. (Technically, mpv wouldn't
+       even know which blocks in the file are valid and which not.)
+
+       The resulting file will not necessarily contain all data of the source
+       stream. For example, if you seek, the parts that were skipped over are
+       never read and consequently are not written to the cache. The skipped over
+       parts are filled with zeros. This means that the cache file doesn't
+       necessarily correspond to a full download of the source stream.
+
+       Both of these issues could be improved if there is any user interest.
+
+       .. warning:: Causes random corruption when used with ordered chapters or
+                    with ``--audio-file``.
+
+    2. Passing the string ``TMP``. This will not be interpreted as filename.
+       Instead, an invisible temporary file is created. It depends on your
+       C library where this file is created (usually ``/tmp/``), and whether
+       filename is visible (the ``tmpfile()`` function is used). On some
+       systems, automatic deletion of the cache file might not be guaranteed.
+
+       If you want to use a file cache, this mode is recommended, because it
+       doesn't break ordered chapters or ``--audio-file``. These modes open
+       multiple cache streams, and using the same file for them obviously
+       clashes.
 
     Also see ``--cache-file-size``.
 
@@ -2750,11 +3008,24 @@ Cache
     Maximum size of the file created with ``--cache-file``. For read accesses
     above this size, the cache is simply not used.
 
+    Keep in mind that some use-cases, like playing ordered chapters with cache
+    enabled, will actually create multiple cache files, each of which will
+    use up to this much disk space.
+
     (Default: 1048576, 1 GB.)
 
 ``--no-cache``
     Turn off input stream caching. See ``--cache``.
 
+``--cache-secs=<seconds>``
+    How many seconds of audio/video to prefetch if the cache is active. This
+    overrides the ``--demuxer-readahead-secs`` option if and only if the cache
+    is enabled. (Default: 2.)
+
+``--cache-pause``, ``--no-cache-pause``
+    Whether the player should automatically pause when the cache runs low,
+    and unpause once more data is available ("buffering").
+
 
 Network
 -------
@@ -2763,11 +3034,9 @@ Network
     Use ``<string>`` as user agent for HTTP streaming.
 
 ``--cookies``, ``--no-cookies``
-    (network only)
     Support cookies when making HTTP requests. Disabled by default.
 
 ``--cookies-file=<filename>``
-    (network only)
     Read HTTP cookies from <filename>. The file is assumed to be in Netscape
     format.
 
@@ -2807,6 +3076,17 @@ Network
     network transport when playing ``rtsp://...`` URLs. The value ``lavf``
     leaves the decision to libavformat.
 
+``--hls-bitrate=<no|min|max>``
+    If HLS streams are played, this option controls what streams are selected
+    by default. The option allows the following parameters:
+
+    :no:        Don't do anything special. Typically, this will simply pick the
+                first audio/video streams it can find. (Default.)
+    :min:       Pick the streams with the lowest bitrate.
+    :max:       Same, but highest bitrate.
+
+    The bitrate as used is sent by the server, and there's no guarantee it's
+    actually meaningful.
 
 DVB
 ---
@@ -2815,9 +3095,10 @@ DVB
     Specifies using card number 1-4 (default: 1).
 
 ``--dvbin-file=<filename>``
-    Instructs mpv to read the channels list from ``<filename>``. Default is
-    ``~/.mpv/channels.conf.{sat,ter,cbl,atsc}`` (based on your card type) or
-    ``~/.mpv/channels.conf`` as a last resort.
+    Instructs mpv to read the channels list from ``<filename>``. The default is
+    in the mpv configuration directory (usually ``~/.config/mpv``) with the
+    filename ``channels.conf.{sat,ter,cbl,atsc}`` (based on your card type) or
+    ``channels.conf`` as a last resort.
 
 ``--dvbin-timeout=<1-30>``
     Maximum number of seconds to wait when trying to tune a frequency before
@@ -2941,6 +3222,11 @@ Miscellaneous
     You can also try to use ``--no-correct-pts`` for files with completely
     broken timestamps.
 
+``--media-title=<string>``
+    Force the contents of the ``media-title`` property to this value. Useful
+    for scripts which want to set a title, without overriding the user's
+    setting in ``--title``.
+
 ``--slave-broken``
     Switches on the old slave mode. This is for testing only, and incompatible
     to the removed ``--slave`` switch.
@@ -2958,10 +3244,3 @@ Miscellaneous
         if there is enough interest).
 
         This affects most third-party GUI frontends.
-
-``--softsleep``
-    Time frames by repeatedly checking the current time instead of asking
-    the kernel to wake up mpv at the correct time. Useful if your kernel
-    timing is imprecise and you cannot use the RTC either. Comes at the
-    price of higher CPU consumption.
-

DOCS/man/osc.rst

@@ -161,7 +161,7 @@ To avoid collisions with other scripts, all options need to be prefixed with
 
 Example::
 
-    --lua-opts=osc-optionA=value1:osc-optionB=value2
+    --lua-opts=osc-optionA=value1,osc-optionB=value2
 
 
 Configurable Options
@@ -229,6 +229,10 @@ Configurable Options
     | Default: yes
     | Display a tooltip over the seekbar indicating time at mouse position.
 
+``seekbarstyle``
+    | Default: slider
+    | Sets the style of the seekbar, slider (diamond marker) or bar (fill)
+
 Script Commands
 ~~~~~~~~~~~~~~~
 

DOCS/man/vf.rst

@@ -231,12 +231,23 @@ Available filters are:
     ``<fmt>``
         Format name, e.g. rgb15, bgr24, 420p, etc. (default: 420p).
 
-``pp[=[filter1[:option1[:option2...]]/[-]filter2...]]``
+``pp=[filter string]``
     Enables the specified chain of post-processing subfilters. Subfilters must
     be separated by '/' and can be disabled by prepending a '-'. Each
     subfilter and some options have a short and a long name that can be used
-    interchangeably, i.e. ``dr``/``dering`` are the same. All subfilters share
-    common options to determine their scope:
+    interchangeably, i.e. ``dr``/``dering`` are the same.
+
+    .. note::
+
+        Unlike in MPlayer or in earlier versions, you must quote the pp string
+        if it contains ``:`` characters, e.g. ``'--vf=pp=[...]'``.
+
+    .. admonition:: Warning
+
+        This filter is most likely useless on modern HD video. It might be
+        helpful with old low-resolution files.
+
+    All subfilters share common options to determine their scope:
 
     ``a/autoq``
         Automatically switch the subfilter off if the CPU is too slow.
@@ -251,11 +262,6 @@ Available filters are:
 
         ``--vf=pp:help`` shows a list of available subfilters.
 
-    .. note::
-
-        Unlike in MPlayer or in earlier versions, you must quote the pp string
-        if it contains ``:`` characters, e.g. ``'--vf=pp=[...]'``.
-
     Available subfilters are:
 
     ``hb/hdeblock[:difference[:flatness]]``
@@ -904,6 +910,31 @@ Available filters are:
         making it higher than the number of cores can actually make it
         slower.
 
+``vapoursynth-lazy``
+    The same as ``vapoursynth``, but doesn't load Python scripts. Instead, a
+    custom backend using Lua and the raw VapourSynth API is used. The syntax
+    is completely different, and absolutely no conveniencve features are
+    provided. There's no type checking either, and you can trigger crashes.
+
+    .. admonition:: Example:
+
+        ::
+
+            video_out = invoke("morpho", "Open", {clip = video_in})
+
+    The special variable ``video_in`` is the mpv video source, while the
+    special variable ``video_out`` is used to read video from. The 1st argument
+    is the plugin (queried with ``getPluginByNs``), the 2nd is the filter name,
+    and the 3rd argument is a table with the arguments. Positional arguments
+    are not supported. The types must match exactly. Since Lua is terrible and
+    can't distinguish integers and floats, integer arguments must be prefixed
+    with ``i_``, in which case the prefix is removed and the argument is cast
+    to an integer. Should the argument's name start with ``i_``, you're out of
+    luck.
+
+    Clips (VSNodeRef) are passed as light userdata, so trying to pass any
+    other userdata type will result in hard crashes.
+
 ``vavpp``
     VA-AP-API video post processing. Works with ``--vo=vaapi`` and ``--vo=opengl``
     only. Currently deinterlaces. This filter is automatically inserted if

DOCS/man/vo.rst

@@ -233,18 +233,29 @@ Available video output drivers are:
         Always force textures to power of 2, even if the device reports
         non-power-of-2 texture sizes as supported.
 
-    ``texture-memory=N``
+    ``texture-memory=<mode>``
         Only affects operation with shaders/texturing enabled, and (E)OSD.
-        Values for N:
+        Possible values:
 
-            0
-                default, will often use an additional shadow texture + copy
-            1
-                use ``D3DPOOL_MANAGED``
-            2
-                use ``D3DPOOL_DEFAULT``
-            3
-                use ``D3DPOOL_SYSTEMMEM``, but without shadow texture
+        ``default`` (default)
+            Use ``D3DPOOL_DEFAULT``, with a ``D3DPOOL_SYSTEMMEM`` texture for
+            locking. If the driver supports ``D3DDEVCAPS_TEXTURESYSTEMMEMORY``,
+            ``D3DPOOL_SYSTEMMEM`` is used directly.
+
+        ``default-pool``
+            Use ``D3DPOOL_DEFAULT``. (Like ``default``, but never use a
+            shadow-texture.)
+
+        ``default-pool-shadow``
+            Use ``D3DPOOL_DEFAULT``, with a ``D3DPOOL_SYSTEMMEM`` texture for
+            locking. (Like ``default``, but always force the shadow-texture.)
+
+        ``managed``
+            Use ``D3DPOOL_MANAGED``.
+
+        ``scratch``
+            Use ``D3DPOOL_SCRATCH``, with a ``D3DPOOL_SYSTEMMEM`` texture for
+            locking.
 
     ``swap-discard``
         Use ``D3DSWAPEFFECT_DISCARD``, which might be faster.
@@ -259,13 +270,6 @@ Available video output drivers are:
 
     .. note:: This driver is for compatibility with old systems.
 
-``corevideo`` (Mac OS X 10.6 and later)
-    Mac OS X CoreVideo video output driver. Uses the CoreVideo APIs to fill
-    PixelBuffers and generate OpenGL textures from them (useful as a fallback
-    for ``opengl``).
-
-    .. note:: This driver is for compatibility with old systems.
-
 ``opengl``
     OpenGL video output driver. It supports extended scaling methods, dithering
     and color management.
@@ -297,6 +301,9 @@ Available video output drivers are:
         ``lanczos3``
             Lanczos with radius=3.
 
+        ``lanczos``
+            Generic Lanczos scaling filter. Set radius with ``lradius``.
+
         ``spline36``
             This is the default when using ``opengl-hq``.
 
@@ -316,6 +323,14 @@ Available video output drivers are:
             Mitchell-Netravali. The ``b`` and ``c`` parameters can be set with
             ``lparam1`` and ``lparam2``. Both are set to 1/3 by default.
 
+        ``gaussian``
+            Gaussian filter with a parameter ``p`` for sharpness control.
+            ``p`` can be set to float number between 1(blurry) and 100(sharp)
+            and has a default value of about 28.8 (see ``lparam1``).
+
+            Note that for extremely small value of ``p``, a large filter radius
+            might be required to avoid unintended artifacts (see ``lradius``).
+
 
         There are some more filters. For a complete list, pass ``help`` as
         value, e.g.::
@@ -329,6 +344,16 @@ Available video output drivers are:
     ``lparam2=<value>``
         See ``lparam1``.
 
+    ``lradius=<r>``
+        Set radius for filters listed below, must be a float number between 1.0
+        and 8.0. Defaults to be 2.0 if not specified.
+
+            ``sinc``, ``lanczos``, ``blackman``, ``gaussian``
+
+        Note that depending on filter implementation details and video scaling
+        ratio, the radius that actually being used might be different
+        (most likely being increased a bit).
+
     ``scaler-resizes-only``
         Disable the scaler if the video image is not resized. In that case,
         ``bilinear`` is used instead whatever is set with ``lscale``. Bilinear
@@ -424,12 +449,22 @@ Available video output drivers are:
         this options will make rendering a single operation.
         Note that chroma scalers are always done as 1-pass filters.
 
-    ``cscale=<n>``
+    ``cscale=<filter>``
         As ``lscale``, but for chroma (2x slower with little visible effect).
         Note that with some scaling filters, upscaling is always done in
         RGB. If chroma is not subsampled, this option is ignored, and the
         luma scaler is used instead. Setting this option is often useless.
 
+    ``lscale-down=<filter>``, ``cscale-down=<filter>``
+        Like ``lscale`` and ``cscale``, but apply these filters on downscaling
+        instead. If these options are unset, the filter implied by ``lscale``
+        (and ``cscale``, respectively) will be applied.
+
+    ``cparam1``, ``cparam2``, ``cradius``
+        Set filter parameters and radius for ``cscale``.
+
+        See ``lparam1``, ``lparam2`` and ``lradius``.
+
     ``fancy-downscaling``
         When using convolution based filters, extend the filter size
         when downscaling. Trades quality for reduced downscaling performance.
@@ -439,7 +474,16 @@ Available video output drivers are:
         Borders will be distorted due to filtering.
 
     ``glfinish``
-        Call ``glFinish()`` before swapping buffers
+        Call ``glFinish()`` before and after swapping buffers (default: disabled).
+        Slower, but might help getting better results when doing framedropping.
+        The details depend entirely on the OpenGL driver.
+
+    ``waitvsync``
+        Call ``glXWaitVideoSyncSGI`` after each buffer swap (default: disabled).
+        This may or may not help with video timing accuracy and frame drop. It's
+        possible that this makes video output slower, or has no effect at all.
+
+        X11 only.
 
     ``sw``
         Continue even if a software renderer is detected.

DOCS/tech-overview.txt

@@ -1,4 +1,4 @@
-This file intends to give a big picture of how mpv is structured.
+This file intends to give a big picture overview of how mpv is structured.
 
 player/*.c:
     Essentially makes up the player applications, including the main() function
@@ -31,15 +31,28 @@ player/*.c:
             * loop, or exit if no next file or quit is requested
               (see enum stop_play_reason)
         * call exit_player_with_rc()
+    * run_playloop():
+        * calls fill_audio_out_buffers()
+            This checks whether new audio needs to be decoded, and pushes it
+            to the AO.
+        * calls write_video()
+            Decode new video, and push it to the VO.
+        * determines whether playback of the current file has ended
+        * determines when to start playback after seeks
+        * and calls a whole lot of other stuff
+            (Really, this function does everything.)
 
     Things worth saying about the playback core:
-    - the currently played tracks are in sh_video and sh_audio
+    - most state is in MPContext (core.h), which is not available to the
+      subsystems
+    - the currently played tracks are in mpctx->current_tracks, and decoder
+      state in d_video/d_audio/d_sub
     - the timeline stuff is used only with MKV ordered chapters (and some other
       minor features: cue, edl)
-    - most state is in MPContext (mp_core.h), which is not available to the
-      subsystems
     - the other subsystems rarely call back into the frontend, and the frontend
       polls them instead (probably a good thing)
+    - one exceptions are wakeup callbacks, which notify a "higher" component
+      of a changed situation in a subsystem
 
     I like to call the player/*.c files the "frontend".
 
@@ -103,7 +116,7 @@ options/options.h, options/options.c
     options.c. Most default values for options and MPOpts are in
     mp_default_opts at the end of options.c.
 
-    MPOpts is unfortunarely quite monolithic, and virtually accessed by
+    MPOpts is unfortunately quite monolithic, and virtually accessed by
     everything.But some components (like video outputs and video filters) have
     their own sub-option tables separate from MPOpts.
 
@@ -111,7 +124,7 @@ options/options.h, options/options.c
     parser-mpcmd.c, and uses the option table in options.c.
 
 input/input.c:
-    This translates keyboard input comming from libvo and other sources (such
+    This translates keyboard input comming from VOs and other sources (such
     as remote control devices like Apple IR or client API commands) to the
     key bindings listed in the user's (or the builtin) input.conf and turns
     them into items of type struct mp_cmd. These commands are queued, and read
@@ -122,7 +135,7 @@ input/input.c:
     of input commands somewhere else.
 
 common/msg.h:
-    All terminal output should go through mp_msg().
+    All terminal output must go through mp_msg().
 
 stream/*:
     File input is implemented here. stream.h/.c provides a simple stream based
@@ -181,7 +194,7 @@ video/out/:
     vo_opengl can pick a windowing system at runtime, e.g. the same binary can
     provide both X11 and Cocoa support on OSX.
 
-    VOs can be reconfigured at runtime. A config() call can change the video
+    VOs can be reconfigured at runtime. A vo_config() call can change the video
     resolution and format, without destroying the window.
 
     vo_vdpau and vo_opengl should be taken as reference.
@@ -203,9 +216,12 @@ audio/filter/:
 audio/out/:
     Audio outputs.
 
-    Unlike VOs, AOs can't be reconfigured on a format change. Without
-    --gapless-audio, even playing a new file will close and re-open the audio
-    device.
+    Unlike VOs, AOs can't be reconfigured on a format change. On audio format
+    changes, the AO will simply be closed and re-opened.
+
+    There are wrappers to support for two types of audio APIs: push.c and
+    pull.c. ao.c calls into one of these. They contain generic code to deal
+    with the data flow these APIs impose.
 
     Note that mpv synchronizes the video to the audio. That's the reason
     why buggy audio drivers can have a bad influence on playback quality.
@@ -218,12 +234,14 @@ sub/:
     the OSD text renderer (which uses libass, and takes care of all the tricky
     fontconfig/freetype API usage and text layouting).
 
-    Subtitle loading is now in demux/ instead. demux_libass.c wraps loading
-    .ass subtitles via libass. demux_lavf.c loads most subtitle types via
-    FFmpeg. demux_subreader.c is the old MPlayer code. It's used as last
-    fallback, or to handle some text subtitle types on Libav. (It also can
-    load UTF-16 encoded subtitles without requiring the use of -sub-codepage.)
-    demux_subreader.c should eventually go away (maybe).
+    The VOs call osd.c to render OSD and subtitle (via e.g. osd_draw()). osd.c
+    in turn asks dec_sub.c for subtitle overlay bitmaps, which relays the
+    request to one of the sd_*.c subtitle decoders/renderers.
+
+    Subtitle loading is in demux/. The MPlayer subreader.c is mostly gone - parts
+    of it survive in demux_subreader.c. It's used as last fallback, or to handle
+    some text subtitle types on Libav. It should go away eventually. Normally,
+    subtitles are loaded via demux_lavf.c.
 
     The subtitles are passed to dec_sub.c and the subtitle decoders in sd_*.c
     as they are demuxed. All text subtitles are rendered by sd_ass.c. If text

README.md

@@ -3,6 +3,8 @@ mpv
 
 [![Build Status](https://api.travis-ci.org/mpv-player/mpv.png)](https://travis-ci.org/mpv-player/mpv)
 
+[![Coverity Scan Build Status](https://scan.coverity.com/projects/1232/badge.svg)](https://scan.coverity.com/projects/1232)
+
 Overview
 --------
 
@@ -48,7 +50,7 @@ Essential dependencies (incomplete list):
 - FFmpeg libraries (libavutil libavcodec libavformat libswscale libavresample)
   At least FFmpeg 2.1.4 or Libav 10 is required.
 - libjpeg (for screenshots)
-- libquvi if you want to play Youtube videos directly
+- Lua (strictly speaking optional, but required for the OSC pseudo-GUI)
 
 Libass dependencies:
 
@@ -117,17 +119,17 @@ make code review easier for both parties later on.
 Contact
 -------
 
-These forms of contact are meant to ask questions about mpv usage, give
-feedback on mpv and discuss it's development.
+Most activity happens on the IRC channel and the github issue tracker. The
+mailing lists are mostly unused.
 
-If possible, please avoid posting bugs here and use the [issue tracker][issue-tracker]
-instead.
-
- - **Users IRC Channel**: `#mpv-player` on `irc.freenode.net`
+ - **Github issue tracker**: [issue tracker][issue-tracker] (report bugs here)
+ - **User IRC Channel**: `#mpv-player` on `irc.freenode.net`
+ - **Developer IRC Channel**: `#mpv-player-dev` on `irc.freenode.net`
  - **Users Mailing List**: `mpv-users@googlegroups.com` ([Archive / Subscribe][mpv-users]).
  - **Devel Mailing List**: `mpv-devel@googlegroups.com` ([Archive / Subscribe][mpv-devel])
 
-To contact the `mpv` team in private write to `mpv-team@googlegroups.com`.
+To contact the `mpv` team in private write to `mpv-team@googlegroups.com`. Use
+only if discretion is required.
 
 [changes]: https://github.com/mpv-player/mpv/blob/master/DOCS/man/changes.rst
 [mpv-build]: https://github.com/mpv-player/mpv-build

RELEASE_NOTES

@@ -1,146 +1,67 @@
-Release 0.5.2
+Release 0.7.0
 =============
 
 Changes
 -------
 
-- Inherit event flags from parent X11 window when using --wid.
-- Properly resume playlists loaded at runtime (this makes mpv try to resume a
-  playlist when loaded like "mpv playlist.m3u").
-- Don't honor $XDG_CONFIG_DIRS anymore for global paths (this fixes a regression
-  on systems that set the variable by default to somewhere else then /etc,
-  making mpv skip previously created config files).
-- Allow up to 256 MB of extradata to make broken files work.
-
-Bug fixes
----------
-
-- Restore old speed change behavior (this fixes a temporary lack of sound after
-  restoring normal speed e.g. when pressing Backspace).
-- Fix quit handling when seeking past EOF while paused.
-- Prevent burning CPU when seeking while paused.
-- Don't add the playlist base path twice when using --playlist.
-- Fix a crash when the initialization of the EGL context fails on Wayland.
-- Make EOF handling properly event-based (this fixes a bug that made mpv hang
-  when changing from one file to the next with --gapless-audio=no).
-- Assume icy title data is terminated with ';' (this fixes handling of streams
-  with metadata containing apostrophes).
-
-This listing is not complete. There are many more bug fixes and changes. The
-complete change log can be viewed by running ``git log 0f1d6c0..e3d3a47`` in
-the git repository.
-
-Release 0.5.1
-=============
-
-Changes
--------
-
-- Add audio/flac to the MIME type list in mpv.desktop.
-- The --slave-broken option has been reintroduced and is equivalent to using
-  --input-file=/dev/stdin and --input-terminal=no.
-
-Bug fixes
----------
-
-- Fix screenshots with anamorphic videos in the vdpau vo.
-- Fix a crash happening when mpv fails to create a vdpau mixer in the vdpau vo.
-- Don't quit when the --loop-file option is used.
-- Always load the cursor theme in the wayland vo (this fixes a crash caused by
-  moving the mouse cursor over the video).
-- Properly return non-zero from the zsh completion script when no matches are
-  found (this fixes the use of user-defined zsh completion matchers).
-- Fix the description of the opengl-hq vo default options in the manpage.
-- Remove duplicated content from the manpage that was accidentally added in the
-  manpage reorganization from the previous release.
-- Avoid clobbering playback position on video end (when both audio and video
-  streams reached the end, the playback poisition was briefly set at 0).
-- Avoid overwriting user input in the coreaudio_device ao.
-
-This listing is not complete. There are many more bug fixes and changes. The
-complete change log can be viewed by running ``git log 6729e39..fdd648b`` in
-the git repository.
-
-Release 0.5.0
-=============
-
-Changes
--------
-
-- Support setting the start time relative to the start PTS with --start.
-- Use /dev/sr0 as default cdrom/dvd device on Linux.
-- Change "cache" property representation to float.
-- Split the ao_coreaudio audio output into ao_coreaudio (which uses the
-  AudioUnits and the CoreAudio sound server) and ao_coreaudio_device (which uses
-  direct device access).
-- Complete supported URL schemes in the zsh completion script.
-- Enable compiler optimizations (-O2) during build.
-- Allow precise seeking with percent-based seeks too.
-- Do precise seeks when clicking on the OSC seekbar.
-- Remove support for older VDA hwaccel versions on Mac OS X (requires recent
-  versions of ffmpeg/libav).
-- Remove the last remains of slave mode (--slave-broken option). Support for
-  reading input commands from stdin can be emulated using: --input-terminal=no
-  --input-file=/dev/stdin.
-- Some key binding were remapped or removed:
-  + "LEFT" and "RIGHT" have been remapped to seek 5 seconds instead of 10.
-  + "c" for changing the YUV color space was unmapped.
-  + OSD level cycling has been remapped from "o" to "O".
-  + "o" is now mapped to the show_progress command (like "P").
-  + "PGUP" and "PGDWN" have been remapped to chapter-based seeks (the old
-    bindings "@" and "!" are still available). The old behvaiour has been
-    remapped to "Shift+PGUP" and "Shift+PGDWM".
-  + Increasing/decreasing audio delay have been remapped from "+" and "-" to
-    "Ctrl++" and "Ctrl+-".
-  + "(" and ")" for adjusting audio balance have been unmapped.
-  + "F" for toggling the "sub-forced-only" option has been unmapped.
-  + "TAB" for cycling TS programs has been unmapped.
-  + "A" for cycling DVD angle has been unmapped.
-  + "U" for stopping playback has been unmapped.
-  + "l" and "L" for cycling quvi format have been unmapped.
-- Make it possible to build the client API (libmpv) on Mac OS X.
-- Do necessary checks for nested Wayland compositors which only have limited
-  interfaces.
-- The options in the manpage have been re-organized into sections and ordered by
-  usage to make it easier to navigate.
-- The "hard" framedrop mode has been removed (this behaviour is still available
-  via the --vd-lavc-framedrop option).
-- Show the basename of a subtitle file only (without the directory part).
-
-Bug fixes
----------
-
-- Use a default color space for files with no video (this fixes color managed
-  OSD messages when no video channel is present).
-- Fix key equivalent dispatching for Cocoa on Mac OS X.
-- Disable non-interleaved audio formats by default with ao_alsa, since some
-  ALSA plugins don't cope well with them (non-interleaved formats can still be
-  enabled using the "non-interleaved" ao_alsa option).
-- Fix deadlock in the client API when calling mpv_terminate_destroy() before
-  mpv_initialize().
-- Fix flushing of filter chain (this fixes a bug where the last frames would get
-  dropped on EOF, if buffered by filters).
-- Actually use the channels.conf configuration file for DVB.
-- Restore the --dvbin-file option.
-- Fix off by one error in the command parser which caused a segfault when using
-  the mp.commandv() Lua function with 10 or more arguments.
+- Buffer partial log messages in the client API (the client API will now only
+  pass full log messages to clients).
+- Remove ncurses/terminfo/termcap support (it was disabled by default and
+  replaced by new code since v0.6.0).
+- Enable cdda:// support by default again (it was disabled since v0.6.0).
+- Cascade-load input.conf (if there are several input.confs in the set of valid
+  config paths, load them all).
+- Draw the OSD twice in 3D mode (this fixes subtitles display in 3D mode).
+- Make wasapi the default AO on Windows again since many of its problems have
+  been solved.
+- Use "site-functions" subdir to install the zsh completion script instead of
+  the Debian-specific "vendor-completions" (also provide the --zshdir waf
+  configure option for changing this value).
+- Improve synchronization between the Cocoa GUI and the player (this fixes some
+  long standing deadlock issues on Mac OS X).
+- Remove --fs-missioncontrol option (only relevant to Mac OS X).
 
 New features
 ------------
 
-- Search for config files in $XDG_CONFIG_HOME and $XDG_CONFIG_DIRS in addition
-  to ~/.mpv. The config file has also been renamed to mpv.conf (although the
-  old file still works and will be loaded even if the new one exists).
-- Add "playback-time" property that returns the current playback time always
-  starting at 0. This is now used to show the playback time in the OSD and the
-  status line, fixing the shown playback time for transport streams that start
-  at a completely random timestamp.
-- Add cache-related "cache-free", "cache-used" and "cache-idle" properties.
-- Add option to list supported protocols (--list-protocols).
-- Allow setting libavformat stream options with the --stream-lavf-o option.
-- Support running the demuxer in a separate thread and prefetch packets (this
-  can be enabled using the --demuxer-thread option).
+- Enable pitch correction by default when playing at higher speeds (this can be
+  controlled with the --audio-pitch-correction option).
+- Open stream and demuxer asynchronously (this should avoid having the player
+  get blocked on network streams).
+- Add cache-buffering-state property for querying the cache fill status until
+  the player unpauses.
+- Add support for listing and selecting the audio device (note that it is not
+  implemented for all AOs, see the --audio-device option for more information).
+- Add support for a JSON-based IPC mechanism (note that this is not currently
+  supported on Windows, see the JSON IPC section in the manpage for more
+  information).
+- Add Lua utility function for starting processes (see utils.subprocess() in the
+  manpage).
+- Add Lua utility function for parsing JSON (see utils.parse_json() in the
+  manpage).
+- Add field-dominance property (see --field-dominance option).
+- Add video-rotate property (see --video-rotate option).
+- Add playback-abort property for querying whether playback is stopped or is to
+  be stopped.
+- Add cursor-autohide property (see --cursor-autohide option).
+- Add vo-configured property for querying whether a window is created.
+- Add support for dxva2 hardware acceleration on Windows.
+- Drop libquvi support (this has been replaced by a built-in Lua script that
+  invokes the youtube-dl tool, which needs to be installed, see the --ytdl
+  option).
+- Add support for loading chapters from an external file (see the
+  --chapters-file option).
+- Add window-minimized property for querying whether the window is minimized
+  (works for X11 only).
+- Make it possible to configure the OSC seekbar style (see the "seekbarstyle"
+  OSC option).
+- Add support for libmpv on Mac OS X (it used to be broken, now it's fully
+  functional and also provides support for embedding the mpv window inside a
+  Cocoa/Qt application).
+- Try to use the audio channel map reported by ALSA in the alsa AO.
+- Add option to disable text OSD rendering completely (useful for working around
+  certain fontconfig issues, see the --use-text-osd option).
 
 This listing is not complete. There are many more bug fixes and changes. The
-complete change log can be viewed by running ``git log cb250d4..52f2f67`` in
+complete change log can be viewed by running ``git log 7759c18..9479daa`` in
 the git repository.

TOOLS/gen-x11-icon.sh

@@ -5,8 +5,8 @@
 
 conv() {
     echo
-    identify -format "icon: %w %h" $1
-    convert $1 -depth 8 rgba:-
+    identify -format "icon: %w %h" "$1"
+    convert "$1" -depth 8 rgba:-
 }
 
 (echo "# File generated by gen-x11-icon.sh" ;

TOOLS/idet.sh

@@ -1,23 +1,15 @@
 #!/bin/sh
 
-case "$0" in
-    */*)
-        MYDIR=${0%/*}
-        ;;
-    *)
-        MYDIR=.
-        ;;
-esac
+: "${MPV:=mpv}"
+: "${ILDETECT_MPV:=$MPV}"
+: "${ILDETECT_MPVFLAGS:=--start=35% --length=35}"
+: "${ILDETECT_DRY_RUN:=}"
+: "${ILDETECT_QUIET:=}"
+: "${ILDETECT_RUN_INTERLACED_ONLY:=}"
+: "${ILDETECT_FORCE_RUN:=}"
 
-: ${MPV:=mpv}
-: ${ILDETECT_MPV:=$MPV}
-: ${ILDETECT_MPV:=$MPV}
-: ${ILDETECT_MPVFLAGS:=--start=35% --length=35}
-: ${ILDETECT_DRY_RUN:=}
-: ${ILDETECT_QUIET:=}
-: ${ILDETECT_RUN_INTERLACED_ONLY:=}
-: ${ILDETECT_FORCE_RUN:=}
-: ${MAKE:=make}
+# This script uses ffmpeg's "idet" filter for interlace detection. In the
+# long run this should replace ildetect.sh+ildetect.so.
 
 # exit status:
 # 0 progressive
@@ -34,57 +26,62 @@ testfun()
         --o= --vo=null --no-audio --untimed \
         $ILDETECT_MPVFLAGS \
         | { if [ -n "$ILDETECT_QUIET" ]; then cat; else tee /dev/stderr; fi } \
-        | grep "Parsed_idet_0: Multi frame detection: " | tail -n 1
+        | grep "Parsed_idet_0: Multi frame detection: "
 }
 
 judge()
 {
-    out=`testfun "$@"`
-
-    tff=${out##* TFF:}; tff=${tff%% *}
-    bff=${out##* BFF:}; bff=${bff%% *}
-    progressive=${out##* Progressive:}; progressive=${progressive%% *}
-    undetermined=${out##* Undetermined:}; undetermined=${undetermined%% *}
-
-    case "$tff$bff$progressive$undetermined" in
-        *[!0-9]*)
-            echo >&2 "ERROR: Unrecognized idet output: $out"
-            exit 16
-            ;;
-    esac
+    tff=0
+    bff=0
+    progressive=0
+    undetermined=0
+    while read -r _ _ _ _ _ _ tff1 _ bff1 _ progressive1 _ undetermined1 _; do
+        case "$tff1$bff1$progressive1$undetermined1" in
+            *[!0-9]*)
+                printf >&2 'ERROR: Unrecognized idet output: %s\n' "$out"
+                exit 16
+                ;;
+        esac
+        tff=$((tff + tff1))
+        bff=$((bff + bff1))
+        progressive=$((progressive + progressive1))
+        undetermined=$((undetermined + undetermined1))
+    done <<EOF
+$(testfun "$@")
+EOF
 
     interlaced=$((bff + tff))
     determined=$((interlaced + progressive))
 
-    if [ $undetermined -gt $determined ] || [ $determined -lt 250 ]; then
+    if [ "$undetermined" -gt "$determined" ] || [ "$determined" -lt 250 ]; then
         echo >&2 "ERROR: Less than 50% or 250 frames are determined."
         [ -n "$ILDETECT_FORCE_RUN" ] || exit 8
         echo >&2 "Assuming interlacing."
-        if [ $tff -gt $((bff * 10)) ]; then
-            verdict=interlaced-tff
-        elif [ $bff -gt $((tff * 10)) ]; then
-            verdict=interlaced-bff
+        if [ "$tff" -gt $((bff * 10)) ]; then
+            verdict="interlaced-tff"
+        elif [ "$bff" -gt $((tff * 10)) ]; then
+            verdict="interlaced-bff"
         else
-            verdict=interlaced
+            verdict="interlaced"
         fi
-    elif [ $((interlaced * 20)) -gt $progressive ]; then
+    elif [ $((interlaced * 20)) -gt "$progressive" ]; then
         # At least 5% of the frames are interlaced!
-        if [ $tff -gt $((bff * 10)) ]; then
-            verdict=interlaced-tff
-        elif [ $bff -gt $((tff * 10)) ]; then
-            verdict=interlaced-bff
+        if [ "$tff" -gt $((bff * 10)) ]; then
+            verdict="interlaced-tff"
+        elif [ "$bff" -gt $((tff * 10)) ]; then
+            verdict="interlaced-bff"
         else
             echo >&2 "ERROR: Content is interlaced, but can't determine field order."
             [ -n "$ILDETECT_FORCE_RUN" ] || exit 8
             echo >&2 "Assuming interlacing with default field order."
-            verdict=interlaced
+            verdict="interlaced"
         fi
     else
-        # Likely progrssive.
-        verdict=progressive
+        # Likely progressive
+        verdict="progressive"
     fi
 
-    echo "$verdict"
+    printf '%s\n' "$verdict"
 }
 
 judge "$@" --vf-clr
@@ -94,7 +91,7 @@ case "$verdict" in
             [ -n "$ILDETECT_RUN_INTERLACED_ONLY" ] || \
             $ILDETECT_MPV "$@"
         r=$?
-        [ $r -eq 0 ] || exit $(($r | 16))
+        [ $r -eq 0 ] || exit $((r | 16))
         exit 0
         ;;
     interlaced-tff)
@@ -104,14 +101,14 @@ case "$verdict" in
                 [ -n "$ILDETECT_DRY_RUN" ] || \
                     $ILDETECT_MPV "$@" --vf-pre=pullup --field-dominance=top
                 r=$?
-                [ $r -eq 0 ] || exit $(($r | 16))
+                [ $r -eq 0 ] || exit $((r | 16))
                 exit 1
                 ;;
             *)
                 [ -n "$ILDETECT_DRY_RUN" ] || \
                     $ILDETECT_MPV "$@" --vf-pre=yadif --field-dominance=top
                 r=$?
-                [ $r -eq 0 ] || exit $(($r | 16))
+                [ $r -eq 0 ] || exit $((r | 16))
                 exit 2
                 ;;
         esac
@@ -123,14 +120,14 @@ case "$verdict" in
                 [ -n "$ILDETECT_DRY_RUN" ] || \
                     $ILDETECT_MPV "$@" --vf-pre=pullup --field-dominance=bottom
                 r=$?
-                [ $r -eq 0 ] || exit $(($r | 16))
+                [ $r -eq 0 ] || exit $((r | 16))
                 exit 1
                 ;;
             *)
                 [ -n "$ILDETECT_DRY_RUN" ] || \
                     $ILDETECT_MPV "$@" --vf-pre=yadif --field-dominance=bottom
                 r=$?
-                [ $r -eq 0 ] || exit $(($r | 16))
+                [ $r -eq 0 ] || exit $((r | 16))
                 exit 2
                 ;;
         esac
@@ -142,14 +139,14 @@ case "$verdict" in
                 [ -n "$ILDETECT_DRY_RUN" ] || \
                     $ILDETECT_MPV "$@" --vf-pre=pullup
                 r=$?
-                [ $r -eq 0 ] || exit $(($r | 16))
+                [ $r -eq 0 ] || exit $((r | 16))
                 exit 1
                 ;;
             *)
                 [ -n "$ILDETECT_DRY_RUN" ] || \
                     $ILDETECT_MPV "$@" --vf-pre=yadif
                 r=$?
-                [ $r -eq 0 ] || exit $(($r | 16))
+                [ $r -eq 0 ] || exit $((r | 16))
                 exit 2
                 ;;
         esac

TOOLS/lib/Parse/Matroska/Definitions.pm

@@ -232,6 +232,7 @@ sub define_matroska {
                     elem('Block',           'a1', 'binary'),
                     elem('BlockDuration',   '9b', 'uint'),
                     elem('ReferenceBlock*', 'fb', 'sint'),
+                    elem('DiscardPadding',  '75A2', 'sint'),
                 }),
                 elem('SimpleBlock*', 'a3', 'binary'),
             }),
@@ -256,6 +257,8 @@ sub define_matroska {
                     elem('CodecPrivate',         '63a2', 'binary'),
                     elem('CodecName',          '258688', 'str'),
                     elem('CodecDecodeAll',         'aa', 'uint'),
+                    elem('CodecDelay',           '56AA', 'uint'),
+                    elem('SeekPreRoll',          '56BB', 'uint'),
                     elem('Video', 'e0', {
                         elem('FlagInterlaced',  '9a', 'uint'),
                         elem('PixelWidth',      'b0', 'uint'),
@@ -265,6 +268,7 @@ sub define_matroska {
                         elem('DisplayUnit',   '54b2', 'uint'),
                         elem('FrameRate',   '2383e3', 'float'),
                         elem('ColourSpace', '2eb524', 'binary'),
+                        elem('StereoMode',    '53b8', 'uint'),
                     }),
                     elem('Audio', 'e1', {
                         elem('SamplingFrequency',         'b5', 'float'),
@@ -292,6 +296,8 @@ sub define_matroska {
                     elem('CueTrackPositions*', 'b7', {
                         elem('CueTrack',           'f7', 'uint'),
                         elem('CueClusterPosition', 'f1', 'uint'),
+                        elem('CueRelativePosition','f0', 'uint'),
+                        elem('CueDuration',        'b2', 'uint'),
                     }),
                 }),
             }),

TOOLS/lua/README.md

@@ -8,12 +8,5 @@ adding the option
 
 to mpv's command line.
 
-Unless otherwise specified, they are also suitable for inclusion in
-the `~/.mpv/lua` directory where they will be loaded every time mpv
-starts, obviating the need to load them with the above `--lua=...`
-argument. This is acceptable as they do only basic argument parsing
-and key-binding registration, until those bound keys are actually
-pressed.  They should therefore not interfere with normal playback
-(unless you have a conflicting user-defined key-binding, in which
-case, you may want to modify either the `mp.add_key_binding()` calls
-in the scripts, or your keybinding).
+Where appropriate, they may also be placed in ~/.config/mpv/lua from where
+they will be automatically loaded when mpv starts.

TOOLS/lua/autocrop.lua

@@ -42,11 +42,11 @@ function del_filter_if_present(label)
     -- error if the filter doesn't exist
     local vfs = mp.get_property_native("vf")
     for i,vf in pairs(vfs) do
-	if vf["label"] == label then
-	    table.remove(vfs, i)
-	    mp.set_property_native("vf", vfs)
-	    return true
-	end
+        if vf["label"] == label then
+            table.remove(vfs, i)
+            mp.set_property_native("vf", vfs)
+            return true
+        end
     end
     return false
 end
@@ -65,10 +65,10 @@ function autocrop_start()
 
     -- insert the cropdetect filter
     ret=mp.command(
-	string.format(
-	    'vf add @%s:lavfi=graph="cropdetect=limit=24:round=2:reset=0"',
-	    cropdetect_label
-	)
+        string.format(
+            'vf add @%s:lavfi=graph="cropdetect=limit=24:round=2:reset=0"',
+            cropdetect_label
+        )
     )
     -- wait to gather data
     timer=mp.add_timeout(detect_seconds, do_crop)
@@ -77,15 +77,15 @@ end
 function do_crop()
     -- get the metadata
     local cropdetect_metadata = mp.get_property_native(
-	string.format("vf-metadata/%s", cropdetect_label)
+        string.format("vf-metadata/%s", cropdetect_label)
     )
     -- use it to crop if its valid
     if cropdetect_metadata then
-	if cropdetect_metadata["lavfi.cropdetect.w"]
-	    and cropdetect_metadata["lavfi.cropdetect.h"]
-	    and cropdetect_metadata["lavfi.cropdetect.x"]
-	    and cropdetect_metadata["lavfi.cropdetect.y"]
-	then
+        if cropdetect_metadata["lavfi.cropdetect.w"]
+            and cropdetect_metadata["lavfi.cropdetect.h"]
+            and cropdetect_metadata["lavfi.cropdetect.x"]
+            and cropdetect_metadata["lavfi.cropdetect.y"]
+        then
             mp.command(string.format("vf add @%s:crop=%s:%s:%s:%s",
                                      crop_label,
                                      cropdetect_metadata["lavfi.cropdetect.w"],
@@ -94,9 +94,9 @@ function do_crop()
                                      cropdetect_metadata["lavfi.cropdetect.y"]))
         else
             mp.msg.error(
-		"Got empty crop data. You might need to increase detect_seconds."
-	    )
-	end
+                "Got empty crop data. You might need to increase detect_seconds."
+            )
+        end
     else
         mp.msg.error(
             "No crop data. Was the cropdetect filter successfully inserted?"

TOOLS/lua/autodeint.lua

@@ -0,0 +1,164 @@
+-- This script uses the lavfi idet filter to automatically insert the
+-- appropriate deinterlacing filter based on a short section of the
+-- currently playing video.
+--
+-- It registers the key-binding ctrl+d, which when pressed, inserts
+-- the filter vf=lavfi=idet. After 4 second, it examines the results
+-- to determine whether the content progressive or interlaced and the
+-- interlacing field dominance. It immediately sets the latter within
+-- mpv.
+--
+-- If the content is judged not to be progressive, it may be either
+-- interlaced or telecined. To determine which one, the video is
+-- seeked back to the detection starting point with the "pullup"
+-- filter inserted. After another four seconds, if idet thinks that
+-- the pulled up viedo is "progressive", then pullup is the correct
+-- filter and this script leaves it in places while removing the idet
+-- filter. Otherwise, the content plain interlaced and mpv's deinterlace
+-- property is set. This will usually insert the yadif deinterlacing filter.
+--
+-- The default detection time may be overridden by adding
+--
+-- --lua-opts=autodeint.detect_seconds=<number of seconds>
+--
+-- to mpv's arguments. This may be desirable to allow idet more
+-- time to collect data.
+--
+-- To see counts of the various types of frames for each detection phase,
+-- the verbosity can be increased with
+--
+-- --msg-level autodeint=v
+--
+-- This script requires a recent version of ffmpeg for which the idet
+-- filter adds the required metadata.
+
+require "mp.msg"
+
+script_name = mp.get_script_name()
+detect_label = string.format("%s-detect", script_name)
+pullup_label = string.format("%s", script_name)
+
+-- number of seconds to gather cropdetect data
+detect_seconds = tonumber(mp.get_opt(string.format("%s.detect_seconds", script_name)))
+if not detect_seconds then
+    detect_seconds = 4
+end
+
+function del_filter_if_present(label)
+    -- necessary because mp.command('vf del @label:filter') raises an
+    -- error if the filter doesn't exist
+    local vfs = mp.get_property_native("vf")
+
+    for i,vf in pairs(vfs) do
+        if vf["label"] == label then
+            table.remove(vfs, i)
+            mp.set_property_native("vf", vfs)
+            return true
+        end
+    end
+    return false
+end
+
+function start_detect()
+    -- exit if detection is already in progress
+    if timer then
+        mp.msg.warn("already detecting!")
+        return
+    end
+
+    mp.set_property("deinterlace","no")
+    del_filter_if_present(pullup_label)
+
+    percent_pos = mp.get_property("percent-pos")
+
+    -- insert the detection filter
+    local cmd = string.format('vf add @%s:lavfi=graph="idet"', detect_label)
+    if not mp.command(cmd) then
+        mp.msg.error("failed to insert detection filter")
+        return
+    end
+
+    -- wait to gather data
+    timer = mp.add_timeout(detect_seconds, judge_field_dominance)
+end
+
+function stop_detect()
+    del_filter_if_present(detect_label)
+    timer = nil
+    mp.set_property("playback-pos", percent_pos)
+end
+
+progressive, interlaced_tff, interlaced_bff, interlaced = 0, 1, 2, 3, 4
+
+function judge()
+    -- get the metadata
+    local result = mp.get_property_native(string.format("vf-metadata/%s", detect_label))
+    num_tff          = tonumber(result["lavfi.idet.multiple.tff"])
+    num_bff          = tonumber(result["lavfi.idet.multiple.bff"])
+    num_progressive  = tonumber(result["lavfi.idet.multiple.progressive"])
+    num_undetermined = tonumber(result["lavfi.idet.multiple.undetermined"])
+    num_interlaced   = num_tff + num_bff
+    num_determined   = num_interlaced + num_progressive
+
+    mp.msg.verbose("progressive    = "..num_progressive)
+    mp.msg.verbose("interlaced-tff = "..num_tff)
+    mp.msg.verbose("interlaced-bff = "..num_bff) 
+    mp.msg.verbose("undetermined   = "..num_undetermined)
+
+    if num_determined < num_undetermined then
+        mp.msg.warn("majority undetermined frames")
+    end
+    if num_progressive > 20*num_interlaced then
+        return progressive
+    elseif num_tff > 10*num_bff then
+        return interlaced_tff
+    elseif num_bff > 10*num_tff then
+        return interlaced_bff
+    else
+        return interlaced
+    end
+end
+
+function judge_field_dominance()
+    verdict = judge()
+    if verdict == progressive then
+        stop_detect()
+        mp.msg.info("progressive: doing nothing")
+        return
+    elseif verdict == interlaced_tff then
+        mp.set_property("field-dominance", "top")
+    elseif verdict == interlaced_bff then
+        mp.set_property("field-dominance", "bottom")
+    elseif verdict == interlaced then
+        mp.set_property("field-dominance", "auto")
+    end
+
+    local cmd = string.format("vf pre @%s:pullup", pullup_label)
+    if not mp.command(cmd) then
+        mp.msg.error("failed to insert pullup filter")
+        stop_detect()
+        return
+    end
+
+    -- redo the detection with the pullup filter to see if it's telecined
+    mp.set_property("percent-pos", percent_pos)
+    timer = mp.add_timeout(detect_seconds, judge_ivtc)
+end
+
+function judge_ivtc()
+    verdict = judge()
+
+    if verdict == progressive then
+        mp.msg.info(string.format("telecinied with %s field dominance: using pullup", mp.get_property("field-dominance")))
+        stop_detect()
+    else
+        mp.msg.info(string.format("interlaced with %s field dominance: setting deinterlace property", mp.get_property("field-dominance")))
+        del_filter_if_present(pullup_label)
+        mp.set_property("deinterlace","yes")
+        stop_detect()
+    end
+
+    return
+end
+
+mp.add_key_binding("ctrl+d", script_name, start_detect)

TOOLS/lua/autoload.lua

@@ -26,6 +26,9 @@ function find_and_add_entries()
     end
     local files = mputils.readdir(dir, "files")
     table.sort(files)
+    if dir == "." then
+        dir = ""
+    end
     local pl = mp.get_property_native("playlist", {})
     local pl_current = mp.get_property_number("playlist-pos", 0) + 1
     -- Find the current pl entry (dir+"/"+filename) in the sorted dir list
@@ -40,27 +43,28 @@ function find_and_add_entries()
         return
     end
     local append = {[-1] = {}, [1] = {}}
-    for dir = -1, 1, 2 do -- 2 iterations, with dir = -1 and +1
+    for direction = -1, 1, 2 do -- 2 iterations, with direction = -1 and +1
         for i = 1, MAXENTRIES do
-            local file = files[current + i * dir]
-            local pl_e = pl[pl_current + i * dir]
+            local file = files[current + i * direction]
+            local pl_e = pl[pl_current + i * direction]
             if file == nil or file[1] == "." then
                 break
             end
+            local filepath = dir .. file
             if pl_e then
                 -- If there's a playlist entry, and it's the same file, stop.
-                if pl_e.filename == file then
+                if pl_e.filename == filepath then
                     break
                 end
             end
-            if dir == -1 then
+            if direction == -1 then
                 if pl_current == 1 then -- never add additional entries in the middle
                     mp.msg.info("Prepending " .. file)
-                    table.insert(append[-1], 1, file)
+                    table.insert(append[-1], 1, filepath)
                 end
             else
                 mp.msg.info("Adding " .. file)
-                table.insert(append[1], file)
+                table.insert(append[1], filepath)
             end
         end
     end

TOOLS/lua/cycle-deinterlace-pullup.lua

@@ -20,9 +20,9 @@ pullup_label = string.format("%s-pullup", script_name)
 
 function pullup_on()
     for i,vf in pairs(mp.get_property_native('vf')) do
-	if vf['label'] == pullup_label then
-	    return "yes"
-	end
+        if vf['label'] == pullup_label then
+            return "yes"
+        end
     end
     return "no"
 end

TOOLS/lua/drc-control.lua

@@ -21,30 +21,30 @@ script_name = mp.get_script_name()
 
 function print_state(params)
     if params then
-	mp.osd_message(script_name..":\n"
-			   .."method = "..params["method"].."\n"
-			   .."target = "..params["target"])
+        mp.osd_message(script_name..":\n"
+                           .."method = "..params["method"].."\n"
+                           .."target = "..params["target"])
     else
-	mp.osd_message(script_name..":\noff")
+        mp.osd_message(script_name..":\noff")
     end
 end
 
 function get_index_of_drc(afs)
     for i,af in pairs(afs) do
-	if af["label"] == script_name then
-	    return i
-	end
+        if af["label"] == script_name then
+            return i
+        end
     end
 end
 
 function append_drc(afs)
     afs[#afs+1] = {
-	name   = "drc",
-	label  = script_name,
-	params = {
-	    method = "1",
-	    target = "0.25"
-	}
+        name   = "drc",
+        label  = script_name,
+        params = {
+            method = "1",
+            target = "0.25"
+        }
     }
     print_state(afs[#afs]["params"])
 end
@@ -53,34 +53,34 @@ function modify_or_create_af(fun)
     afs = mp.get_property_native("af")
     i = get_index_of_drc(afs)
     if not i then
-	append_drc(afs)
+        append_drc(afs)
     else
-	fun(afs, i)
+        fun(afs, i)
     end
     mp.set_property_native("af", afs)
 end
 
 function drc_toggle_method_handler()
     modify_or_create_af(
-	function (afs, i)
-	    new_method=(afs[i]["params"]["method"]+1)%3
-	    if new_method == 0 then
-		table.remove(afs, i)
-		print_state(nil)
-	    else
-		afs[i]["params"]["method"] = tostring((afs[i]["params"]["method"])%2+1)
-		print_state(afs[i]["params"])
-	    end
-	end
+        function (afs, i)
+            new_method=(afs[i]["params"]["method"]+1)%3
+            if new_method == 0 then
+                table.remove(afs, i)
+                print_state(nil)
+            else
+                afs[i]["params"]["method"] = tostring((afs[i]["params"]["method"])%2+1)
+                print_state(afs[i]["params"])
+            end
+        end
     )
 end
 
 function drc_scale_target(factor)
     modify_or_create_af(
-	function (afs)
-	    afs[i]["params"]["target"] = tostring(afs[i]["params"]["target"]*factor)
-	    print_state(afs[i]["params"])
-	end
+        function (afs)
+            afs[i]["params"]["target"] = tostring(afs[i]["params"]["target"]*factor)
+            print_state(afs[i]["params"])
+        end
     )
 end
 

TOOLS/mpv_identify.sh

@@ -17,157 +17,134 @@
 # When multiple files were specified, their info will be put into FOO_* for the
 # first file, FOO_1_* for the second file, FOO_2_* for the third file, etc.
 
-case "$0" in
-    mpv_identify.sh|*/mpv_identify.sh)
-        # we are NOT being sourced
-        case "$1" in
-            '')
+__midentify__main() {
+
+    case "$0" in
+        mpv_identify.sh|*/mpv_identify.sh)
+            # we are NOT being sourced
+            [ -n "$1" ] && set -- '' "$@"
+            ;;
+    esac
+
+    if [ "$#" -lt 2 ]; then
+        cat >&2 <<EOF
+Usage 1 (for humans only): $0 filename.mkv
+will print all property values.
+Note that this output really shouldn't be parsed, as the
+format is subject to change.
+
+Usage 2 (for use by scripts): see top of this file
+
+NOTE: for mkv with ordered chapters, this may
+not always identify the specified file, but the
+file providing the first chapter. Specify
+--no-ordered-chapters to prevent this.
+EOF
+        return 2
+    fi
+
+    local LF="
+"
+
+    local nextprefix="$1"
+    shift
+
+    if [ -n "$nextprefix" ]; then
+        # in case of error, we always want this unset
+        unset "${nextprefix}path"
+    fi
+
+    local allprops="
+        filename
+        path
+        stream-start
+        stream-end
+        stream-length
+
+        demuxer
+
+        length
+        chapters
+        editions
+        titles
+
+        audio
+        audio-bitrate
+        audio-codec
+        audio-format
+        channels
+        samplerate
+
+        video
+        angle
+        video-bitrate
+        video-codec
+        video-format
+        video-aspect
+        fps
+        width
+        height
+        dwidth
+        dheight
+
+        sub
+    "
+    # TODO add metadata support once mpv can do it
+
+    local propstr="X-MIDENTIFY-START:$LF"
+    local key
+    for key in $allprops; do
+        propstr="${propstr}X-MIDENTIFY: $key \${=$key}$LF"
+        key="$(printf '%s\n' "$key" | tr - _)"
+        unset "$nextprefix$key"
+    done
+
+    local fileindex=0
+    local prefix=
+    local line
+    while IFS= read -r line; do
+        case "$line" in
+            X-MIDENTIFY-START:)
+                if [ -n "$nextprefix" ]; then
+                    prefix="$nextprefix"
+                    if [ "$fileindex" -gt 0 ]; then
+                        nextprefix="${prefix%${fileindex}_}"
+                    fi
+                    fileindex="$((fileindex+1))"
+                    nextprefix="${nextprefix}${fileindex}_"
+                    for key in $allprops; do
+                        key="$(printf '%s\n' "$key" | tr - _)"
+                        unset "$nextprefix$key"
+                    done
+                else
+                    if [ "$fileindex" -gt 0 ]; then
+                        printf '\n'
+                    fi
+                    fileindex="$((fileindex+1))"
+                fi
                 ;;
-            *)
-                set -- '' "$@"
+            X-MIDENTIFY:\ *)
+                local key="${line#X-MIDENTIFY: }"
+                local value="${key#* }"
+                key="${key%% *}"
+                key="$(printf '%s\n' "$key" | tr - _)"
+                if [ -n "$nextprefix" ]; then
+                    if [ -z "$prefix" ]; then
+                        echo >&2 "Got X-MIDENTIFY: without X-MIDENTIFY-START:"
+                    elif [ -n "$value" ]; then
+                        eval "$prefix$key"='"$value"'
+                    fi
+                else
+                    if [ -n "$value" ]; then
+                        printf '%s=%s\n' "$key" "$value"
+                    fi
+                fi
                 ;;
         esac
-        ;;
-esac
+    done <<EOF
+$(${MPV:-mpv} --term-playing-msg="$propstr" --vo=null --ao=null \
+              --frames=1 --quiet --no-cache --no-config -- "$@")
+EOF
+}
 
-if [ $# -lt 2 ]; then
-    echo >&2 "Usage 1 (for humans only): $0 filename.mkv"
-    echo >&2 "will print all property values."
-    echo >&2 "Note that this output really shouldn't be parsed, as the"
-    echo >&2 "format is subject to change."
-    echo >&2
-    echo >&2 "Usage 2 (for use by scripts): see top of this file"
-    echo >&2
-    echo >&2 "NOTE: for mkv with ordered chapters, this may"
-    echo >&2 "not always identify the specified file, but the"
-    echo >&2 "file providing the first chapter. Specify"
-    echo >&2 "--no-ordered-chapters to prevent this."
-    exit 1
-fi
-
-if [ -z "$MPV" ]; then
-    MPV="mpv"
-fi
-
-__midentify__LF="
-"
-
-__midentify__nextprefix=$1
-shift
-
-if [ -n "$__midentify__nextprefix" ]; then
-    # in case of error, we always want this unset
-    eval unset $__midentify__nextprefix'path'
-fi
-
-__midentify__allprops="
-    filename
-    path
-    stream-start
-    stream-end
-    stream-length
-
-    demuxer
-
-    length
-    chapters
-    editions
-    titles
-
-    audio
-    audio-bitrate
-    audio-codec
-    audio-format
-    channels
-    samplerate
-
-    video
-    angle
-    video-bitrate
-    video-codec
-    video-format
-    video-aspect
-    fps
-    width
-    height
-    dwidth
-    dheight
-
-    sub
-"
-# TODO add metadata support once mpv can do it
-
-__midentify__propstr="X-MIDENTIFY-START:$__midentify__LF"
-for __midentify__key in $__midentify__allprops; do
-    __midentify__propstr=$__midentify__propstr"X-MIDENTIFY: $__midentify__key \${=$__midentify__key}$__midentify__LF"
-    __midentify__key=`echo "$__midentify__key" | tr - _`
-    eval unset $__midentify__nextprefix$__midentify__key
-done
-
-__midentify__output=`$MPV --term-playing-msg="$__midentify__propstr" --vo=null --ao=null --frames=1 --quiet --no-cache --no-config "$@"`
-__midentify__fileindex=0
-__midentify__prefix=
-while :; do
-    case "$__midentify__output" in
-        '')
-            break
-            ;;
-        *$__midentify__LF*)
-            __midentify__line=${__midentify__output%%$__midentify__LF*}
-            __midentify__output=${__midentify__output#*$__midentify__LF}
-            ;;
-        *)
-            __midentify__line=$__midentify__output
-            __midentify__output=
-            ;;
-    esac
-    case "$__midentify__line" in
-        X-MIDENTIFY-START:)
-            if [ -n "$__midentify__nextprefix" ]; then
-                __midentify__prefix=$__midentify__nextprefix
-                if [ $__midentify__fileindex -gt 0 ]; then
-                    __midentify__nextprefix=${__midentify__prefix%$__midentify__fileindex\_}
-                fi
-                __midentify__fileindex=$(($__midentify__fileindex+1))
-                __midentify__nextprefix=$__midentify__nextprefix$__midentify__fileindex\_
-                for __midentify__key in $__midentify__allprops; do
-                    __midentify__key=`echo "$__midentify__key" | tr - _`
-                    eval unset $__midentify__nextprefix$__midentify__key
-                done
-            else
-                if [ $__midentify__fileindex -gt 0 ]; then
-                    echo
-                fi
-                __midentify__fileindex=$(($__midentify__fileindex+1))
-            fi
-            ;;
-        X-MIDENTIFY:\ *)
-            __midentify__key=${__midentify__line#X-MIDENTIFY:\ }
-            __midentify__value=${__midentify__key#* }
-            __midentify__key=${__midentify__key%% *}
-            __midentify__key=`echo "$__midentify__key" | tr - _`
-            if [ -n "$__midentify__nextprefix" ]; then
-                if [ -z "$__midentify__prefix" ]; then
-                    echo >&2 "Got X-MIDENTIFY: without X-MIDENTIFY-START:"
-                elif [ -n "$__midentify__value" ]; then
-                    eval $__midentify__prefix$__midentify__key=\$__midentify__value
-                fi
-            else
-                if [ -n "$__midentify__value" ]; then
-                    echo "$__midentify__key=$__midentify__value"
-                fi
-            fi
-            ;;
-    esac
-done
-
-unset __midentify__fileindex
-unset __midentify__allprops
-unset __midentify__key
-unset __midentify__LF
-unset __midentify__line
-unset __midentify__output
-unset __midentify__nextprefix
-unset __midentify__prefix
-unset __midentify__propstr
-unset __midentify__value
+__midentify__main "$@"

TOOLS/stats-conv.py

@@ -18,10 +18,11 @@ e.g.:
 
 Currently, the following event types are supported:
 
-    'start' <name>          start of the named event
-    'end' <name>            end of the named event
-    'value' <float> <name>  a normal value (as opposed to event)
-    <event>                 singular event
+    'start' <name>              start of the named event
+    'end' <name>                end of the named event
+    'value' <float> <name>      a normal value (as opposed to event)
+    'event-timed' <ts> <name>   singular event at the given timestamp
+    <name>                      singular event
 
 """
 
@@ -29,19 +30,32 @@ class G:
     events = {}
     sevents = []  # events, deterministically sorted
     start = None
+    # http://matplotlib.org/api/markers_api.html#module-matplotlib.markers
+    markers = ["o", "8", "s", "p", "*", "h", "+", "x", "D"]
+
+def find_marker():
+    if len(G.markers) == 0:
+        return "o"
+    m = G.markers[0]
+    G.markers = G.markers[1:]
+    return m
 
 class Event:
     pass
 
-def get_event(event):
+def get_event(event, evtype):
     if event not in G.events:
         e = Event()
         G.events[event] = e
         e.name = event
         e.vals = []
-        e.type = "unknown"
+        e.type = evtype
+        e.marker = "o"
+        e.numid = len(G.events)
         G.sevents = list(G.events.values())
         G.sevents.sort(key=lambda x: x.name)
+        if e.type == "event-signal":
+            e.marker = find_marker()
     return G.events[event]
 
 for line in [line.split("#")[0].strip() for line in open(filename, "r")]:
@@ -54,30 +68,29 @@ for line in [line.split("#")[0].strip() for line in open(filename, "r")]:
         G.start = ts
     ts = ts - G.start
     if event.startswith("start "):
-        e = get_event(event[6:])
-        e.type = "event"
+        e = get_event(event[6:], "event")
         e.vals.append((ts, 0))
         e.vals.append((ts, 1))
     elif event.startswith("end "):
-        e = get_event(event[4:])
-        e.type = "event"
+        e = get_event(event[4:], "event")
         e.vals.append((ts, 1))
         e.vals.append((ts, 0))
     elif event.startswith("value "):
         _, val, name = event.split(" ", 2)
         val = float(val)
-        e = get_event(name)
-        e.type = "value"
+        e = get_event(name, "value")
         e.vals.append((ts, val))
+    elif event.startswith("event-timed "):
+        _, val, name = event.split(" ", 2)
+        val = int(val) / 1000 - G.start
+        e = get_event(name, "event-signal")
+        e.vals.append((val, 1))
     else:
-        e = get_event(event)
-        e.type = "event-signal"
+        e = get_event(event, "event-signal")
         e.vals.append((ts, 1))
 
-ao_events = ["ao fill", "audio wait"]
 for e in G.sevents:
-    if e.name in ao_events:
-        e.vals = [(x, y * 0.5) for (x, y) in e.vals]
+    e.vals = [(x, y * e.numid / len(G.events)) for (x, y) in e.vals]
 
 plot.hold(True)
 mainpl = plot.subplot(2, 1, 1)
@@ -89,7 +102,7 @@ for e in G.sevents:
         plot.subplot(2, 1, 1)
     pl, = plot.plot([x for x,y in e.vals], [y for x,y in e.vals], label=e.name)
     if e.type == "event-signal":
-        plot.setp(pl, marker = "o", linestyle = "None")
+        plot.setp(pl, marker = e.marker, linestyle = "None")
     legend.append(pl)
 plot.subplot(2, 1, 1)
 plot.legend(legend, [pl.get_label() for pl in legend])

TOOLS/umpv

@@ -21,20 +21,13 @@ script after that will start a new mpv instance.
 
 Note that you can control the mpv instance by writing to the command fifo:
 
-    echo "cycle fullscreen" > /tmp/umpv-fifo-*
+    echo "cycle fullscreen" > ~/.umpv-fifo
 
-Note: you can supply custom options with the UMPV_OPTIONS environment variable.
-      The environment variable will be split on whitespace, and each item is
-      passed as option to mpv _if_ the script starts mpv. If mpv is not started
-      by the script (i.e. mpv is already running), the options will be ignored.
-
-Warning:
-
-The script attempts to make sure the FIFO is safely created (i.e. not world-
-writable), and checks that it's really a FIFO. This is important for security,
-because the FIFO allows anyone with write access to run arbitrary commands
-in mpv's context using the "run" input command. If you are worried about
-security, you should verify that the code handles these concerns correctly.
+Note: you can supply custom mpv path and options with the MPV environment
+      variable. The environment variable will be split on whitespace, and the
+      first item is used as path to mpv binary and the rest is passed as options
+      _if_ the script starts mpv. If mpv is not started by the script (i.e. mpv
+      is already running), this will be ignored.
 """
 
 import sys
@@ -45,21 +38,17 @@ import fcntl
 import stat
 import string
 
-if len(sys.argv) < 1:
-    sys.exit(1)
 files = sys.argv[1:]
 
-# this is about the same method mpv uses to decide this
+# this is the same method mpv uses to decide this
 def is_url(filename):
-    parts = filename.split(":", 1)
+    parts = filename.split("://", 1)
     if len(parts) < 2:
         return False
     # protocol prefix has no special characters => it's an URL
+    allowed_symbols = string.ascii_letters + string.digits + '_'
     prefix = parts[0]
-    for c in prefix:
-        if string.ascii_letters.find(c) < 0:
-            return False
-    return True
+    return all(map(lambda c: c in allowed_symbols, prefix))
 
 # make them absolute; also makes them safe against interpretation as options
 def make_abs(filename):
@@ -68,10 +57,7 @@ def make_abs(filename):
     return filename
 files = [make_abs(f) for f in files]
 
-opts_env = os.getenv("UMPV_OPTIONS")
-opts_env = opts_env.split() if opts_env else []
-
-FIFO = "/tmp/umpv-fifo-" + os.getenv("USER")
+FIFO = os.path.join(os.getenv("HOME"), ".umpv_fifo")
 
 fifo_fd = -1
 try:
@@ -84,14 +70,6 @@ except OSError as e:
     else:
         raise e
 
-if fifo_fd >= 0:
-    st = os.fstat(fifo_fd)
-    if (((st.st_mode & (stat.S_IRWXG | stat.S_IRWXO)) != 0) or
-        (not stat.S_ISFIFO(st.st_mode)) or
-        (st.st_uid != os.getuid())):
-        print("error: command FIFO is not a FIFO or has bogus permissions")
-        sys.exit(1)
-
 if fifo_fd >= 0:
     # Unhandled race condition: what if mpv is terminating right now?
     fcntl.fcntl(fifo_fd, fcntl.F_SETFL, 0) # set blocking mode
@@ -108,15 +86,12 @@ else:
     try:
         os.unlink(FIFO)
     except OSError as e:
-        if e.errno == errno.ENOENT:
-            pass
-        else:
-            raise e
-    os.mkfifo(FIFO, int("0600", 8))
+        pass
+    os.mkfifo(FIFO, 0o600)
 
-    opts = ["mpv", "--no-terminal", "--force-window", "--input-file=" + FIFO]
-    opts.extend(opts_env)
-    opts.append("--")
+    opts = (os.getenv("MPV") or "mpv").split()
+    opts.extend(["--no-terminal", "--force-window", "--input-file=" + FIFO,
+                 "--"])
     opts.extend(files)
 
     subprocess.check_call(opts)

TOOLS/vf_dlopen/framestep.c

@@ -90,8 +90,7 @@ int vf_dlopen_getcontext(struct vf_dlopen_context *ctx, int argc, const char **a
     if (argc != 1 && argc != 2)
         return -1;
 
-    framestep_data_t *framestep = malloc(sizeof(framestep_data_t));
-    memset(framestep, 0, sizeof(*framestep));
+    framestep_data_t *framestep = calloc(1,sizeof(framestep_data_t));
 
     framestep->step = atoi(argv[0]);
     framestep->pos = (argc >= 2) ? atoi(argv[1]) : 0;

TOOLS/vf_dlopen/ildetect.c

@@ -267,8 +267,7 @@ int vf_dlopen_getcontext(struct vf_dlopen_context *ctx, int argc, const char **a
     (void) argc;
     (void) argv;
 
-    ildetect_data_t *il = malloc(sizeof(ildetect_data_t));
-    memset(il, 0, sizeof(*il));
+    ildetect_data_t *il = calloc(1,sizeof(ildetect_data_t));
 
 #define A(i,d) ((argc>(i) && *argv[i]) ? atof(argv[i]) : (d))
     il->method = A(0, 0);

TOOLS/vf_dlopen/ildetect.sh

@@ -11,7 +11,6 @@ esac
 
 : ${MPV:=mpv}
 : ${ILDETECT_MPV:=$MPV}
-: ${ILDETECT_MPV:=$MPV}
 : ${ILDETECT_MPVFLAGS:=--start=35% --length=35}
 : ${ILDETECT_DRY_RUN:=}
 : ${ILDETECT_QUIET:=}

TOOLS/vf_dlopen/rectangle.c

@@ -348,8 +348,7 @@ int vf_dlopen_getcontext(struct vf_dlopen_context *ctx, int argc, const char **a
         { "yuv420p", "yuv420p" },
         { NULL, NULL }
     };
-    privdata *priv = malloc(sizeof(privdata));
-    memset(priv, 0, sizeof(*priv));
+    privdata *priv = calloc(1,sizeof(privdata));
     priv->step = 8;
     if(argc >= 1)
         priv->w = atoi(argv[0]);

TOOLS/vf_dlopen/telecine.c

@@ -237,8 +237,7 @@ int vf_dlopen_getcontext(struct vf_dlopen_context *ctx, int argc, const char **a
     if (!a0[0] || a0[1] || !a1[0] || argc > 2)
         return -1;
 
-    tc_data_t *tc = malloc(sizeof(tc_data_t));
-    memset(tc, 0, sizeof(*tc));
+    tc_data_t *tc = calloc(1,sizeof(tc_data_t));
 
     if (a0[0] == 't')
         tc->firstfield = 0;

TOOLS/vf_dlopen/tile.c

@@ -150,8 +150,7 @@ int vf_dlopen_getcontext(struct vf_dlopen_context *ctx, int argc, const char **a
     if (argc != 2)
         return -1;
 
-    tile_data_t *tile = malloc(sizeof(tile_data_t));
-    memset(tile, 0, sizeof(*tile));
+    tile_data_t *tile = calloc(1,sizeof(tile_data_t));
 
     tile->cols = atoi(argv[0]);
     tile->rows = atoi(argv[1]);

TOOLS/youtube-dl_mpv.sh

@@ -10,14 +10,13 @@ video_url="$(youtube-dl \
     --user-agent="$user_agent" \
     --cookies="$cookies_file" \
     --get-url \
-    "$1")"
+    "$@")"
 
-shift
-
-mpv \
+set -f
+${MPV:-mpv} \
     --cookies \
     --cookies-file="$cookies_file" \
     --user-agent="$user_agent" \
-    "$@" -- $video_url
+    -- $video_url
 
 rm -rf "$cookies_dir"

TOOLS/zsh.pl

@@ -37,6 +37,36 @@ chomp $vf_str;
 $protos_str .= qq{$_ } foreach (@protos);
 chomp $protos_str;
 
+my $profile_comp = <<'EOS';
+      local -a profiles
+      local current
+      for current in "${(@f)$($words[1] --profile=help)}"; do
+        current=${current//\*/\\\*}
+        current=${current//\:/\\\:}
+        current=${current//\[/\\\[}
+        current=${current//\]/\\\]}
+        if [[ $current =~ $'\t'([^$'\t']*)$'\t'(.*) ]]; then
+          if [[ -n $match[2] ]]; then
+            current="$match[1][$match[2]]"
+          else
+            current="$match[1]"
+          fi
+          profiles=($profiles $current)
+        fi
+      done
+      if [[ $state == profile ]]; then
+        # For --show-profile, only one allowed
+        if (( ${#profiles} > 0 )); then
+          _values 'profile' $profiles && rc=0
+        fi
+      else
+        # For --profile, multiple allowed
+        profiles=($profiles 'help[list profiles]')
+        _values -s , 'profile(s)' $profiles && rc=0
+      fi
+EOS
+chomp $profile_comp;
+
 my $tmpl = <<"EOS";
 #compdef mpv
 
@@ -88,6 +118,16 @@ $vf_str
     _describe -t values 'video filters' values && rc=0
   ;;
 
+  profile|profiles)
+$profile_comp
+  ;;
+
+  files)
+    compset -P '*,'
+    compset -S ',*'
+    _files -r ',/ \\t\\n\\-' && rc=0
+  ;;
+
   mfiles)
     local expl
     _tags files urls
@@ -141,6 +181,9 @@ sub parse_opts {
             $entry .= '->vo' if ($1 eq '--vo');
             $entry .= '->af' if ($1 eq '--af');
             $entry .= '->vf' if ($1 eq '--vf');
+            $entry .= '->profiles' if ($1 eq '--profile');
+            $entry .= '->profile' if ($1 eq '--show-profile');
+            $entry .= '->files' if ($line =~ /\[file\]/);
         }
 
         push @list, $entry

VERSION

@@ -1 +1 @@
-0.5.2
+0.7.0

audio/audio.c

@@ -20,10 +20,13 @@
 #include <stdlib.h>
 #include <assert.h>
 
-#include <libavutil/mem.h>
+#include <libavutil/buffer.h>
+#include <libavutil/frame.h>
+#include <libavutil/version.h>
 
 #include "talloc.h"
 #include "common/common.h"
+#include "fmt-conversion.h"
 #include "audio.h"
 
 static void update_redundant_info(struct mp_audio *mpa)
@@ -88,18 +91,12 @@ bool mp_audio_config_valid(const struct mp_audio *mpa)
         && mpa->rate >= 1 && mpa->rate < 10000000;
 }
 
-char *mp_audio_fmt_to_str(int srate, const struct mp_chmap *chmap, int format)
+char *mp_audio_config_to_str_buf(char *buf, size_t buf_sz, struct mp_audio *mpa)
 {
-    char *chstr = mp_chmap_to_str(chmap);
-    char *res = talloc_asprintf(NULL, "%dHz %s %dch %s", srate, chstr,
-                                chmap->num, af_fmt_to_str(format));
-    talloc_free(chstr);
-    return res;
-}
-
-char *mp_audio_config_to_str(struct mp_audio *mpa)
-{
-    return mp_audio_fmt_to_str(mpa->rate, &mpa->channels, mpa->format);
+    snprintf(buf, buf_sz, "%dHz %s %dch %s", mpa->rate,
+             mp_chmap_to_str(&mpa->channels), mpa->channels.num,
+             af_fmt_to_str(mpa->format));
+    return buf;
 }
 
 void mp_audio_force_interleaved_format(struct mp_audio *mpa)
@@ -116,19 +113,27 @@ int mp_audio_psize(struct mp_audio *mpa)
 
 void mp_audio_set_null_data(struct mp_audio *mpa)
 {
-    for (int n = 0; n < MP_NUM_CHANNELS; n++)
+    for (int n = 0; n < MP_NUM_CHANNELS; n++) {
         mpa->planes[n] = NULL;
+        mpa->allocated[n] = NULL;
+    }
     mpa->samples = 0;
 }
 
+static int get_plane_size(const struct mp_audio *mpa, int samples)
+{
+    if (samples < 0 || !mp_audio_config_valid(mpa))
+        return -1;
+    if (samples >= INT_MAX / mpa->sstride)
+        return -1;
+    return MPMAX(samples * mpa->sstride, 1);
+}
+
 static void mp_audio_destructor(void *ptr)
 {
     struct mp_audio *mpa = ptr;
-    for (int n = 0; n < MP_NUM_CHANNELS; n++) {
-        // Note: don't free if not allocated by mp_audio_realloc
-        if (mpa->allocated[n])
-            av_free(mpa->planes[n]);
-    }
+    for (int n = 0; n < MP_NUM_CHANNELS; n++)
+        av_buffer_unref(&mpa->allocated[n]);
 }
 
 /* Reallocate the data stored in mpa->planes[n] so that enough samples are
@@ -147,27 +152,19 @@ static void mp_audio_destructor(void *ptr)
  */
 void mp_audio_realloc(struct mp_audio *mpa, int samples)
 {
-    assert(samples >= 0);
-    if (samples >= INT_MAX / mpa->sstride)
-        abort(); // oom
-    int size = MPMAX(samples * mpa->sstride, 1);
+    int size = get_plane_size(mpa, samples);
+    if (size < 0)
+        abort(); // oom or invalid parameters
     for (int n = 0; n < mpa->num_planes; n++) {
-        if (size != mpa->allocated[n]) {
-            // Note: av_realloc() can't be used (see libavutil doxygen)
-            void *new = av_malloc(size);
-            if (!new)
-                abort();
-            if (mpa->allocated[n])
-                memcpy(new, mpa->planes[n], MPMIN(mpa->allocated[n], size));
-            av_free(mpa->planes[n]);
-            mpa->planes[n] = new;
-            mpa->allocated[n] = size;
+        if (!mpa->allocated[n] || size != mpa->allocated[n]->size) {
+            if (av_buffer_realloc(&mpa->allocated[n], size) < 0)
+                abort(); // OOM
+            mpa->planes[n] = mpa->allocated[n]->data;
         }
     }
     for (int n = mpa->num_planes; n < MP_NUM_CHANNELS; n++) {
-        av_free(mpa->planes[n]);
+        av_buffer_unref(&mpa->allocated[n]);
         mpa->planes[n] = NULL;
-        mpa->allocated[n] = 0;
     }
     talloc_set_destructor(mpa, mp_audio_destructor);
 }
@@ -194,7 +191,7 @@ int mp_audio_get_allocated_size(struct mp_audio *mpa)
 {
     int size = 0;
     for (int n = 0; n < mpa->num_planes; n++) {
-        int s = mpa->allocated[n] / mpa->sstride;
+        int s = mpa->allocated[n] ? mpa->allocated[n]->size / mpa->sstride : 0;
         size = n == 0 ? s : MPMIN(size, s);
     }
     return size;
@@ -240,3 +237,136 @@ void mp_audio_skip_samples(struct mp_audio *data, int samples)
 
     data->samples -= samples;
 }
+
+// Make sure the frame owns the audio data, and if not, copy the data.
+// Return negative value on failure (which means it can't be made writeable).
+// Non-refcounted frames are always considered writeable.
+int mp_audio_make_writeable(struct mp_audio *data)
+{
+    bool ok = true;
+    for (int n = 0; n < MP_NUM_CHANNELS; n++) {
+        if (data->allocated[n])
+            ok &= av_buffer_is_writable(data->allocated[n]);
+    }
+    if (!ok) {
+        struct mp_audio *new = talloc(NULL, struct mp_audio);
+        *new = *data;
+        mp_audio_set_null_data(new); // use format only
+        mp_audio_realloc(new, data->samples);
+        new->samples = data->samples;
+        mp_audio_copy(new, 0, data, 0, data->samples);
+        // "Transfer" the reference.
+        mp_audio_destructor(data);
+        *data = *new;
+        talloc_set_destructor(new, NULL);
+        talloc_free(new);
+    }
+    return 0;
+}
+
+struct mp_audio *mp_audio_from_avframe(struct AVFrame *avframe)
+{
+    AVFrame *tmp = NULL;
+    struct mp_audio *new = talloc_zero(NULL, struct mp_audio);
+    talloc_set_destructor(new, mp_audio_destructor);
+
+    mp_audio_set_format(new, af_from_avformat(avframe->format));
+
+    struct mp_chmap lavc_chmap;
+    mp_chmap_from_lavc(&lavc_chmap, avframe->channel_layout);
+
+#if LIBAVUTIL_VERSION_MICRO >= 100
+    // FFmpeg being special again
+    if (lavc_chmap.num != avframe->channels)
+        mp_chmap_from_channels(&lavc_chmap, avframe->channels);
+#endif
+
+    new->rate = avframe->sample_rate;
+
+    mp_audio_set_channels(new, &lavc_chmap);
+
+    // Force refcounted frame.
+    if (!avframe->buf[0]) {
+        tmp = av_frame_alloc();
+        if (!tmp)
+            goto fail;
+        if (av_frame_ref(tmp, avframe) < 0)
+            goto fail;
+        avframe = tmp;
+    }
+
+    // If we can't handle the format (e.g. too many channels), bail out.
+    if (!mp_audio_config_valid(new) || avframe->nb_extended_buf)
+        goto fail;
+
+    for (int n = 0; n < AV_NUM_DATA_POINTERS; n++) {
+        if (!avframe->buf[n])
+            break;
+        if (n >= MP_NUM_CHANNELS)
+            goto fail;
+        new->allocated[n] = av_buffer_ref(avframe->buf[n]);
+        if (!new->allocated[n])
+            goto fail;
+    }
+
+    for (int n = 0; n < new->num_planes; n++)
+        new->planes[n] = avframe->data[n];
+    new->samples = avframe->nb_samples;
+
+    return new;
+
+fail:
+    talloc_free(new);
+    av_frame_free(&tmp);
+    return NULL;
+}
+
+struct mp_audio_pool {
+    AVBufferPool *avpool;
+    int element_size;
+};
+
+struct mp_audio_pool *mp_audio_pool_create(void *ta_parent)
+{
+    return talloc_zero(ta_parent, struct mp_audio_pool);
+}
+
+static void mp_audio_pool_destructor(void *p)
+{
+    struct mp_audio_pool *pool = p;
+    av_buffer_pool_uninit(&pool->avpool);
+}
+
+// Allocate data using the given format and number of samples.
+// Returns NULL on error.
+struct mp_audio *mp_audio_pool_get(struct mp_audio_pool *pool,
+                                   const struct mp_audio *fmt, int samples)
+{
+    int size = get_plane_size(fmt, samples);
+    if (size < 0)
+        return NULL;
+    if (!pool->avpool || size > pool->element_size) {
+        size_t alloc = ta_calc_prealloc_elems(size);
+        if (alloc >= INT_MAX)
+            return NULL;
+        av_buffer_pool_uninit(&pool->avpool);
+        pool->avpool = av_buffer_pool_init(alloc, NULL);
+        if (!pool->avpool)
+            return NULL;
+        talloc_set_destructor(pool, mp_audio_pool_destructor);
+    }
+    struct mp_audio *new = talloc_ptrtype(NULL, new);
+    talloc_set_destructor(new, mp_audio_destructor);
+    *new = *fmt;
+    mp_audio_set_null_data(new);
+    new->samples = samples;
+    for (int n = 0; n < new->num_planes; n++) {
+        new->allocated[n] = av_buffer_pool_get(pool->avpool);
+        if (!new->allocated[n]) {
+            talloc_free(new);
+            return NULL;
+        }
+        new->planes[n] = new->allocated[n]->data;
+    }
+    return new;
+}

audio/audio.h

@@ -38,7 +38,7 @@ struct mp_audio {
     int bps;            // size of sub-samples (af_fmt2bps(format))
 
     // private
-    int allocated[MP_NUM_CHANNELS]; // use mp_audio_get_allocated_size()
+    struct AVBufferRef *allocated[MP_NUM_CHANNELS];
 };
 
 void mp_audio_set_format(struct mp_audio *mpa, int format);
@@ -49,8 +49,8 @@ void mp_audio_copy_config(struct mp_audio *dst, const struct mp_audio *src);
 bool mp_audio_config_equals(const struct mp_audio *a, const struct mp_audio *b);
 bool mp_audio_config_valid(const struct mp_audio *mpa);
 
-char *mp_audio_fmt_to_str(int srate, const struct mp_chmap *chmap, int format);
-char *mp_audio_config_to_str(struct mp_audio *mpa);
+char *mp_audio_config_to_str_buf(char *buf, size_t buf_sz, struct mp_audio *mpa);
+#define mp_audio_config_to_str(m) mp_audio_config_to_str_buf((char[64]){0}, 64, (m))
 
 void mp_audio_force_interleaved_format(struct mp_audio *mpa);
 
@@ -68,4 +68,13 @@ void mp_audio_copy(struct mp_audio *dst, int dst_offset,
                    struct mp_audio *src, int src_offset, int length);
 void mp_audio_skip_samples(struct mp_audio *data, int samples);
 
+int mp_audio_make_writeable(struct mp_audio *data);
+
+struct AVFrame;
+struct mp_audio *mp_audio_from_avframe(struct AVFrame *avframe);
+
+struct mp_audio_pool *mp_audio_pool_create(void *ta_parent);
+struct mp_audio *mp_audio_pool_get(struct mp_audio_pool *pool,
+                                   const struct mp_audio *fmt, int samples);
+
 #endif

audio/chmap.c

@@ -18,6 +18,7 @@
 #include <stdlib.h>
 #include <assert.h>
 
+#include "common/common.h"
 #include "common/msg.h"
 #include "chmap.h"
 
@@ -86,6 +87,7 @@ static const char *const std_layout_names[][2] = {
     {"7.1(wide-side)",  "fl-fr-fc-lfe-flc-frc-sl-sr"},
     {"octagonal",       "fl-fr-fc-bl-br-bc-sl-sr"},
     {"downmix",         "dl-dr"},
+    {"auto",            ""}, // not in lavc
     {0}
 };
 
@@ -375,34 +377,35 @@ void mp_chmap_get_reorder(int dst[MP_NUM_CHANNELS], const struct mp_chmap *from,
 // Returns something like "fl-fr-fc". If there's a standard layout in lavc
 // order, return that, e.g. "3.0" instead of "fl-fr-fc".
 // Unassigned but valid speakers get names like "sp28".
-char *mp_chmap_to_str(const struct mp_chmap *src)
+char *mp_chmap_to_str_buf(char *buf, size_t buf_size, const struct mp_chmap *src)
 {
-    char *res = talloc_strdup(NULL, "");
+    buf[0] = '\0';
 
-    if (mp_chmap_is_unknown(src))
-        return talloc_asprintf_append_buffer(res, "unknown%d", src->num);
+    if (mp_chmap_is_unknown(src)) {
+        snprintf(buf, buf_size, "unknown%d", src->num);
+        return buf;
+    }
 
     for (int n = 0; n < src->num; n++) {
         int sp = src->speaker[n];
         const char *s = sp < MP_SPEAKER_ID_COUNT ? speaker_names[sp][0] : NULL;
-        char buf[10];
+        char sp_buf[10];
         if (!s) {
-            snprintf(buf, sizeof(buf), "sp%d", sp);
-            s = buf;
+            snprintf(sp_buf, sizeof(sp_buf), "sp%d", sp);
+            s = sp_buf;
         }
-        res = talloc_asprintf_append_buffer(res, "%s%s", n > 0 ? "-" : "", s);
+        mp_snprintf_cat(buf, buf_size, "%s%s", n > 0 ? "-" : "", s);
     }
 
     // To standard layout name
     for (int n = 0; std_layout_names[n][0]; n++) {
-        if (res && strcmp(res, std_layout_names[n][1]) == 0) {
-            talloc_free(res);
-            res = talloc_strdup(NULL, std_layout_names[n][0]);
+        if (strcmp(buf, std_layout_names[n][1]) == 0) {
+            snprintf(buf, buf_size, "%s", std_layout_names[n][0]);
             break;
         }
     }
 
-    return res;
+    return buf;
 }
 
 // If src can be parsed as channel map (as produced by mp_chmap_to_str()),

audio/chmap.h

@@ -20,7 +20,7 @@
 
 #include <inttypes.h>
 #include <stdbool.h>
-#include "bstr/bstr.h"
+#include "misc/bstr.h"
 
 #define MP_NUM_CHANNELS 8
 
@@ -61,7 +61,8 @@ enum mp_speaker_id {
     MP_SPEAKER_ID_UNKNOWN0 = 64,
     MP_SPEAKER_ID_UNKNOWN_LAST = MP_SPEAKER_ID_UNKNOWN0 + MP_NUM_CHANNELS - 1,
 
-    // Including the unassigned IDs in between. This is not a valid ID anymore.
+    // Including the unassigned IDs in between. This is not a valid ID anymore,
+    // but is still within uint8_t.
     MP_SPEAKER_ID_COUNT,
 };
 
@@ -119,7 +120,9 @@ void mp_chmap_reorder_to_lavc(struct mp_chmap *map);
 void mp_chmap_get_reorder(int dst[MP_NUM_CHANNELS], const struct mp_chmap *from,
                           const struct mp_chmap *to);
 
-char *mp_chmap_to_str(const struct mp_chmap *src);
+char *mp_chmap_to_str_buf(char *buf, size_t buf_size, const struct mp_chmap *src);
+#define mp_chmap_to_str(m) mp_chmap_to_str_buf((char[64]){0}, 64, (m))
+
 bool mp_chmap_from_str(struct mp_chmap *dst, bstr src);
 
 struct mp_log;

audio/decode/ad.h

@@ -36,7 +36,7 @@ struct ad_functions {
     int (*init)(struct dec_audio *da, const char *decoder);
     void (*uninit)(struct dec_audio *da);
     int (*control)(struct dec_audio *da, int cmd, void *arg);
-    int (*decode_packet)(struct dec_audio *da);
+    int (*decode_packet)(struct dec_audio *da, struct mp_audio **out);
 };
 
 enum ad_ctrl {

audio/decode/ad_lavc.c

@@ -25,6 +25,7 @@
 #include <libavcodec/avcodec.h>
 #include <libavutil/opt.h>
 #include <libavutil/common.h>
+#include <libavutil/intreadwrite.h>
 
 #include "talloc.h"
 
@@ -37,14 +38,13 @@
 #include "ad.h"
 #include "audio/fmt-conversion.h"
 
-#include "compat/libav.h"
-
 struct priv {
     AVCodecContext *avctx;
     AVFrame *avframe;
     struct mp_audio frame;
     bool force_channel_map;
     struct demux_packet *packet;
+    uint32_t skip_samples;
 };
 
 static void uninit(struct dec_audio *da);
@@ -73,117 +73,6 @@ const struct m_sub_options ad_lavc_conf = {
     },
 };
 
-struct pcm_map
-{
-    int tag;
-    const char *codecs[6]; // {any, 1byte, 2bytes, 3bytes, 4bytes, 8bytes}
-};
-
-// NOTE: these are needed to make rawaudio with demux_mkv work.
-static const struct pcm_map tag_map[] = {
-    // Microsoft PCM
-    {0x0,           {NULL, "pcm_u8", "pcm_s16le", "pcm_s24le", "pcm_s32le"}},
-    {0x1,           {NULL, "pcm_u8", "pcm_s16le", "pcm_s24le", "pcm_s32le"}},
-    // MS PCM, Extended
-    {0xfffe,        {NULL, "pcm_u8", "pcm_s16le", "pcm_s24le", "pcm_s32le"}},
-    // IEEE float
-    {0x3,           {"pcm_f32le", [5] = "pcm_f64le"}},
-    // 'raw '
-    {0x20776172,    {"pcm_s16be", [1] = "pcm_u8"}},
-    // 'twos', used by demux_mkv.c internally
-    {MKTAG('t', 'w', 'o', 's'),
-                    {NULL, "pcm_s8", "pcm_s16be", "pcm_s24be", "pcm_s32be"}},
-    {-1},
-};
-
-// For demux_rawaudio.c; needed because ffmpeg doesn't have these sample
-// formats natively.
-static const struct pcm_map af_map[] = {
-    {AF_FORMAT_U8,              {"pcm_u8"}},
-    {AF_FORMAT_S8,              {"pcm_u8"}},
-    {AF_FORMAT_U16_LE,          {"pcm_u16le"}},
-    {AF_FORMAT_U16_BE,          {"pcm_u16be"}},
-    {AF_FORMAT_S16_LE,          {"pcm_s16le"}},
-    {AF_FORMAT_S16_BE,          {"pcm_s16be"}},
-    {AF_FORMAT_U24_LE,          {"pcm_u24le"}},
-    {AF_FORMAT_U24_BE,          {"pcm_u24be"}},
-    {AF_FORMAT_S24_LE,          {"pcm_s24le"}},
-    {AF_FORMAT_S24_BE,          {"pcm_s24be"}},
-    {AF_FORMAT_U32_LE,          {"pcm_u32le"}},
-    {AF_FORMAT_U32_BE,          {"pcm_u32be"}},
-    {AF_FORMAT_S32_LE,          {"pcm_s32le"}},
-    {AF_FORMAT_S32_BE,          {"pcm_s32be"}},
-    {AF_FORMAT_FLOAT_LE,        {"pcm_f32le"}},
-    {AF_FORMAT_FLOAT_BE,        {"pcm_f32be"}},
-    {AF_FORMAT_DOUBLE_LE,       {"pcm_f64le"}},
-    {AF_FORMAT_DOUBLE_BE,       {"pcm_f64be"}},
-    {-1},
-};
-
-static const char *find_pcm_decoder(const struct pcm_map *map, int format,
-                                    int bits_per_sample)
-{
-    int bytes = (bits_per_sample + 7) / 8;
-    if (bytes == 8)
-        bytes = 5; // 64 bit entry
-    for (int n = 0; map[n].tag != -1; n++) {
-        const struct pcm_map *entry = &map[n];
-        if (entry->tag == format) {
-            const char *dec = NULL;
-            if (bytes >= 1 && bytes <= 5)
-                dec = entry->codecs[bytes];
-            if (!dec)
-                dec = entry->codecs[0];
-            if (dec)
-                return dec;
-        }
-    }
-    return NULL;
-}
-
-static void set_data_from_avframe(struct dec_audio *da)
-{
-    struct priv *priv = da->priv;
-    AVCodecContext *lavc_context = priv->avctx;
-
-    // Note: invalid parameters are rejected by dec_audio.c
-
-    int fmt = lavc_context->sample_fmt;
-    mp_audio_set_format(&da->decoded, af_from_avformat(fmt));
-    if (!da->decoded.format)
-        MP_FATAL(da, "unsupported lavc format %s", av_get_sample_fmt_name(fmt));
-
-    da->decoded.rate = lavc_context->sample_rate;
-
-    struct mp_chmap lavc_chmap;
-    mp_chmap_from_lavc(&lavc_chmap, lavc_context->channel_layout);
-    // No channel layout or layout disagrees with channel count
-    if (lavc_chmap.num != lavc_context->channels)
-        mp_chmap_from_channels(&lavc_chmap, lavc_context->channels);
-    if (priv->force_channel_map) {
-        struct sh_audio *sh_audio = da->header->audio;
-        if (lavc_chmap.num == sh_audio->channels.num)
-            lavc_chmap = sh_audio->channels;
-    }
-    mp_audio_set_channels(&da->decoded, &lavc_chmap);
-
-    da->decoded.samples = priv->avframe->nb_samples;
-    for (int n = 0; n < da->decoded.num_planes; n++)
-        da->decoded.planes[n] = priv->avframe->data[n];
-}
-
-static void set_from_wf(AVCodecContext *avctx, MP_WAVEFORMATEX *wf)
-{
-    avctx->channels = wf->nChannels;
-    avctx->sample_rate = wf->nSamplesPerSec;
-    avctx->bit_rate = wf->nAvgBytesPerSec * 8;
-    avctx->block_align = wf->nBlockAlign;
-    avctx->bits_per_coded_sample = wf->wBitsPerSample;
-
-    if (wf->cbSize > 0)
-        mp_lavc_set_extradata(avctx, wf + 1, wf->cbSize);
-}
-
 static int init(struct dec_audio *da, const char *decoder)
 {
     struct MPOpts *mpopts = da->opts;
@@ -196,13 +85,7 @@ static int init(struct dec_audio *da, const char *decoder)
     struct priv *ctx = talloc_zero(NULL, struct priv);
     da->priv = ctx;
 
-    if (sh_audio->wf && strcmp(decoder, "pcm") == 0) {
-        decoder = find_pcm_decoder(tag_map, sh->format,
-                                   sh_audio->wf->wBitsPerSample);
-    } else if (sh_audio->wf && strcmp(decoder, "mp-pcm") == 0) {
-        decoder = find_pcm_decoder(af_map, sh->format, 0);
-        ctx->force_channel_map = true;
-    }
+    ctx->force_channel_map = sh_audio->force_channels;
 
     lavc_codec = avcodec_find_decoder_by_name(decoder);
     if (!lavc_codec) {
@@ -227,17 +110,23 @@ static int init(struct dec_audio *da, const char *decoder)
     av_opt_set_double(lavc_context, "drc_scale", opts->ac3drc,
                       AV_OPT_SEARCH_CHILDREN);
 
+#if HAVE_AVFRAME_SKIP_SAMPLES
+    // Let decoder add AV_FRAME_DATA_SKIP_SAMPLES.
+    av_opt_set(lavc_context, "flags2", "+skip_manual", AV_OPT_SEARCH_CHILDREN);
+#endif
+
     mp_set_avopts(da->log, lavc_context, opts->avopts);
 
     lavc_context->codec_tag = sh->format;
     lavc_context->sample_rate = sh_audio->samplerate;
     lavc_context->bit_rate = sh_audio->bitrate;
-    lavc_context->channel_layout = mp_chmap_to_lavc(&sh_audio->channels);
+    lavc_context->block_align = sh_audio->block_align;
+    lavc_context->bits_per_coded_sample = sh_audio->bits_per_coded_sample;
+    lavc_context->channels = sh_audio->channels.num;
+    if (!mp_chmap_is_unknown(&sh_audio->channels))
+        lavc_context->channel_layout = mp_chmap_to_lavc(&sh_audio->channels);
 
-    if (sh_audio->wf)
-        set_from_wf(lavc_context, sh_audio->wf);
-
-    // demux_mkv, demux_mpg
+    // demux_mkv
     if (sh_audio->codecdata_len && sh_audio->codecdata &&
             !lavc_context->extradata) {
         mp_lavc_set_extradata(lavc_context, sh_audio->codecdata,
@@ -258,8 +147,6 @@ static int init(struct dec_audio *da, const char *decoder)
 
     if (lavc_context->bit_rate != 0)
         da->bitrate = lavc_context->bit_rate;
-    else if (sh_audio->wf && sh_audio->wf->nAvgBytesPerSec)
-        da->bitrate = sh_audio->wf->nAvgBytesPerSec * 8;
 
     return 1;
 }
@@ -286,21 +173,19 @@ static int control(struct dec_audio *da, int cmd, void *arg)
     switch (cmd) {
     case ADCTRL_RESET:
         avcodec_flush_buffers(ctx->avctx);
-        mp_audio_set_null_data(&da->decoded);
         talloc_free(ctx->packet);
         ctx->packet = NULL;
+        ctx->skip_samples = 0;
         return CONTROL_TRUE;
     }
     return CONTROL_UNKNOWN;
 }
 
-static int decode_packet(struct dec_audio *da)
+static int decode_packet(struct dec_audio *da, struct mp_audio **out)
 {
     struct priv *priv = da->priv;
     AVCodecContext *avctx = priv->avctx;
 
-    mp_audio_set_null_data(&da->decoded);
-
     struct demux_packet *mpkt = priv->packet;
     if (!mpkt) {
         if (demux_read_packet_async(da->header, &mpkt) == 0)
@@ -347,23 +232,55 @@ static int decode_packet(struct dec_audio *da)
     if (!got_frame)
         return mpkt ? AD_OK : AD_EOF;
 
-    set_data_from_avframe(da);
-
     double out_pts = mp_pts_from_av(priv->avframe->pkt_pts, NULL);
     if (out_pts != MP_NOPTS_VALUE) {
         da->pts = out_pts;
         da->pts_offset = 0;
     }
 
-    MP_DBG(da, "Decoded %d -> %d samples\n", in_len, da->decoded.samples);
+    struct mp_audio *mpframe = mp_audio_from_avframe(priv->avframe);
+    if (!mpframe)
+        return AD_ERR;
+
+    struct mp_chmap lavc_chmap = mpframe->channels;
+    if (lavc_chmap.num != avctx->channels)
+        mp_chmap_from_channels(&lavc_chmap, avctx->channels);
+    if (priv->force_channel_map) {
+        struct sh_audio *sh_audio = da->header->audio;
+        if (lavc_chmap.num == sh_audio->channels.num)
+            lavc_chmap = sh_audio->channels;
+    }
+    mp_audio_set_channels(mpframe, &lavc_chmap);
+
+#if HAVE_AVFRAME_SKIP_SAMPLES
+    AVFrameSideData *sd =
+        av_frame_get_side_data(priv->avframe, AV_FRAME_DATA_SKIP_SAMPLES);
+    if (sd && sd->size >= 10) {
+        char *d = sd->data;
+        priv->skip_samples += AV_RL32(d + 0);
+        uint32_t pad = AV_RL32(d + 4);
+        uint32_t skip = MPMIN(priv->skip_samples, mpframe->samples);
+        if (skip) {
+            mp_audio_skip_samples(mpframe, skip);
+            da->pts_offset += skip;
+            priv->skip_samples -= skip;
+        }
+        if (pad <= mpframe->samples)
+            mpframe->samples -= pad;
+    }
+#endif
+
+    *out = mpframe;
+
+    av_frame_unref(priv->avframe);
+
+    MP_DBG(da, "Decoded %d -> %d samples\n", in_len, mpframe->samples);
     return 0;
 }
 
 static void add_decoders(struct mp_decoder_list *list)
 {
     mp_add_lavc_decoders(list, AVMEDIA_TYPE_AUDIO);
-    mp_add_decoder(list, "lavc", "pcm", "pcm", "Raw PCM");
-    mp_add_decoder(list, "lavc", "mp-pcm", "mp-pcm", "Raw PCM");
 }
 
 const struct ad_functions ad_lavc = {

audio/decode/ad_mpg123.c

@@ -43,6 +43,7 @@ struct ad_mpg123_context {
     short delay;
     /* If the stream is actually VBR. */
     char vbr;
+    struct mp_audio frame;
 };
 
 static void uninit(struct dec_audio *da)
@@ -197,27 +198,25 @@ static int set_format(struct dec_audio *da)
     int encoding;
     ret = mpg123_getformat(con->handle, &rate, &channels, &encoding);
     if (ret == MPG123_OK) {
-        mp_audio_set_num_channels(&da->decoded, channels);
-        da->decoded.rate = rate;
+        mp_audio_set_num_channels(&con->frame, channels);
+        con->frame.rate = rate;
         int af = mpg123_format_to_af(encoding);
         if (!af) {
             /* This means we got a funny custom build of libmpg123 that only supports an unknown format. */
             MP_ERR(da, "Bad encoding from mpg123: %i.\n", encoding);
             return MPG123_ERR;
         }
-        mp_audio_set_format(&da->decoded, af);
+        mp_audio_set_format(&con->frame, af);
         con->sample_size = channels * af_fmt2bps(af);
     }
     return ret;
 }
 
-static int decode_packet(struct dec_audio *da)
+static int decode_packet(struct dec_audio *da, struct mp_audio **out)
 {
     struct ad_mpg123_context *con = da->priv;
     int ret;
 
-    mp_audio_set_null_data(&da->decoded);
-
     struct demux_packet *pkt;
     if (demux_read_packet_async(da->header, &pkt) == 0)
         return AD_WAIT;
@@ -257,8 +256,11 @@ static int decode_packet(struct dec_audio *da)
     }
 
     int got_samples = bytes / con->sample_size;
-    da->decoded.planes[0] = audio;
-    da->decoded.samples = got_samples;
+    *out = mp_audio_pool_get(da->pool, &con->frame, got_samples);
+    if (!*out)
+        return AD_ERR;
+
+    memcpy((*out)->planes[0], audio, bytes);
 
     update_info(da);
     return 0;
@@ -274,7 +276,6 @@ static int control(struct dec_audio *da, int cmd, void *arg)
 
     switch (cmd) {
     case ADCTRL_RESET:
-        mp_audio_set_null_data(&da->decoded);
         mpg123_close(con->handle);
 
         if (mpg123_open_feed(con->handle) != MPG123_OK) {

audio/decode/ad_spdif.c

@@ -40,6 +40,7 @@ struct spdifContext {
     int              out_buffer_len;
     uint8_t          out_buffer[OUTBUF_SIZE];
     bool             need_close;
+    struct mp_audio  fmt;
 };
 
 static int write_packet(void *p, uint8_t *buf, int buf_size)
@@ -117,13 +118,13 @@ static int init(struct dec_audio *da, const char *decoder)
     switch (stream->codec->codec_id) {
     case AV_CODEC_ID_AAC:
         spdif_ctx->iec61937_packet_size = 16384;
-        sample_format                   = AF_FORMAT_IEC61937_LE;
+        sample_format                   = AF_FORMAT_S_AAC;
         samplerate                      = 48000;
         num_channels                    = 2;
         break;
     case AV_CODEC_ID_AC3:
         spdif_ctx->iec61937_packet_size = 6144;
-        sample_format                   = AF_FORMAT_AC3_LE;
+        sample_format                   = AF_FORMAT_S_AC3;
         samplerate                      = 48000;
         num_channels                    = 2;
         break;
@@ -131,40 +132,40 @@ static int init(struct dec_audio *da, const char *decoder)
         if (da->opts->dtshd) {
             av_dict_set(&format_opts, "dtshd_rate", "768000", 0); // 4*192000
             spdif_ctx->iec61937_packet_size = 32768;
-            sample_format                   = AF_FORMAT_IEC61937_LE;
+            sample_format                   = AF_FORMAT_S_DTSHD;
             samplerate                      = 192000;
             num_channels                    = 2*4;
         } else {
             spdif_ctx->iec61937_packet_size = 32768;
-            sample_format                   = AF_FORMAT_AC3_LE;
+            sample_format                   = AF_FORMAT_S_DTS;
             samplerate                      = 48000;
             num_channels                    = 2;
         }
         break;
     case AV_CODEC_ID_EAC3:
         spdif_ctx->iec61937_packet_size = 24576;
-        sample_format                   = AF_FORMAT_IEC61937_LE;
+        sample_format                   = AF_FORMAT_S_EAC3;
         samplerate                      = 192000;
         num_channels                    = 2;
         break;
     case AV_CODEC_ID_MP3:
         spdif_ctx->iec61937_packet_size = 4608;
-        sample_format                   = AF_FORMAT_MPEG2;
+        sample_format                   = AF_FORMAT_S_MP3;
         samplerate                      = 48000;
         num_channels                    = 2;
         break;
     case AV_CODEC_ID_TRUEHD:
         spdif_ctx->iec61937_packet_size = 61440;
-        sample_format                   = AF_FORMAT_IEC61937_LE;
+        sample_format                   = AF_FORMAT_S_TRUEHD;
         samplerate                      = 192000;
         num_channels                    = 8;
         break;
     default:
         abort();
     }
-    mp_audio_set_num_channels(&da->decoded, num_channels);
-    mp_audio_set_format(&da->decoded, sample_format);
-    da->decoded.rate = samplerate;
+    mp_audio_set_num_channels(&spdif_ctx->fmt, num_channels);
+    mp_audio_set_format(&spdif_ctx->fmt, sample_format);
+    spdif_ctx->fmt.rate = samplerate;
 
     if (avformat_write_header(lavf_ctx, &format_opts) < 0) {
         MP_FATAL(da, "libavformat spdif initialization failed.\n");
@@ -182,13 +183,11 @@ fail:
     return 0;
 }
 
-static int decode_packet(struct dec_audio *da)
+static int decode_packet(struct dec_audio *da, struct mp_audio **out)
 {
     struct spdifContext *spdif_ctx = da->priv;
     AVFormatContext     *lavf_ctx  = spdif_ctx->lavf_ctx;
 
-    mp_audio_set_null_data(&da->decoded);
-
     spdif_ctx->out_buffer_len  = 0;
 
     struct demux_packet *mpkt;
@@ -212,8 +211,12 @@ static int decode_packet(struct dec_audio *da)
     if (ret < 0)
         return AD_ERR;
 
-    da->decoded.planes[0] = spdif_ctx->out_buffer;
-    da->decoded.samples = spdif_ctx->out_buffer_len / da->decoded.sstride;
+    int samples = spdif_ctx->out_buffer_len / spdif_ctx->fmt.sstride;
+    *out = mp_audio_pool_get(da->pool, &spdif_ctx->fmt, samples);
+    if (!*out)
+        return AD_ERR;
+
+    memcpy((*out)->planes[0], spdif_ctx->out_buffer, spdif_ctx->out_buffer_len);
 
     return 0;
 }

audio/decode/dec_audio.c

@@ -28,7 +28,7 @@
 #include "config.h"
 #include "common/codecs.h"
 #include "common/msg.h"
-#include "bstr/bstr.h"
+#include "misc/bstr.h"
 
 #include "stream/stream.h"
 #include "demux/demux.h"
@@ -48,27 +48,14 @@ extern const struct ad_functions ad_lavc;
 extern const struct ad_functions ad_spdif;
 
 static const struct ad_functions * const ad_drivers[] = {
+    &ad_lavc,
 #if HAVE_MPG123
     &ad_mpg123,
 #endif
-    &ad_lavc,
     &ad_spdif,
     NULL
 };
 
-// Drop audio buffer and reinit it (after format change)
-// Returns whether the format was valid at all.
-static bool reinit_audio_buffer(struct dec_audio *da)
-{
-    if (!mp_audio_config_valid(&da->decoded)) {
-        MP_ERR(da, "Audio decoder did not specify audio "
-               "format, or requested an unsupported configuration!\n");
-        return false;
-    }
-    mp_audio_buffer_reinit(da->decode_buffer, &da->decoded);
-    return true;
-}
-
 static void uninit_decoder(struct dec_audio *d_audio)
 {
     if (d_audio->ad_driver) {
@@ -89,7 +76,6 @@ static int init_audio_codec(struct dec_audio *d_audio, const char *decoder)
         return 0;
     }
 
-    d_audio->decode_buffer = mp_audio_buffer_create(NULL);
     return 1;
 }
 
@@ -164,136 +150,26 @@ void audio_uninit(struct dec_audio *d_audio)
 {
     if (!d_audio)
         return;
-    if (d_audio->afilter) {
-        MP_VERBOSE(d_audio, "Uninit audio filters...\n");
-        af_destroy(d_audio->afilter);
-        d_audio->afilter = NULL;
-    }
+    MP_VERBOSE(d_audio, "Uninit audio filters...\n");
+    af_destroy(d_audio->afilter);
     uninit_decoder(d_audio);
-    talloc_free(d_audio->decode_buffer);
     talloc_free(d_audio);
 }
 
-
-int audio_init_filters(struct dec_audio *d_audio, int in_samplerate,
-                       int *out_samplerate, struct mp_chmap *out_channels,
-                       int *out_format)
-{
-    if (!d_audio->afilter)
-        d_audio->afilter = af_new(d_audio->global);
-    struct af_stream *afs = d_audio->afilter;
-
-    // input format: same as codec's output format:
-    mp_audio_buffer_get_format(d_audio->decode_buffer, &afs->input);
-    // Sample rate can be different when adjusting playback speed
-    afs->input.rate = in_samplerate;
-
-    // output format: same as ao driver's input format (if missing, fallback to input)
-    afs->output.rate = *out_samplerate;
-    mp_audio_set_channels(&afs->output, out_channels);
-    mp_audio_set_format(&afs->output, *out_format);
-
-    afs->replaygain_data = d_audio->replaygain_data;
-
-    char *s_from = mp_audio_config_to_str(&afs->input);
-    char *s_to = mp_audio_config_to_str(&afs->output);
-    MP_VERBOSE(d_audio, "Building audio filter chain for %s -> %s...\n", s_from, s_to);
-    talloc_free(s_from);
-    talloc_free(s_to);
-
-    // let's autoprobe it!
-    if (af_init(afs) != 0) {
-        af_destroy(afs);
-        d_audio->afilter = NULL;
-        return 0;   // failed :(
-    }
-
-    *out_samplerate = afs->output.rate;
-    *out_channels = afs->output.channels;
-    *out_format = afs->output.format;
-
-    return 1;
-}
-
 /* Decode packets until we know the audio format. Then reinit the buffer.
  * Returns AD_OK on success, negative AD_* code otherwise.
  * Also returns AD_OK if already initialized (and does nothing).
  */
 int initial_audio_decode(struct dec_audio *da)
 {
-    while (!mp_audio_config_valid(&da->decoded)) {
-        if (da->decoded.samples > 0)
-            return AD_ERR; // invalid format, rather than uninitialized
-        int ret = da->ad_driver->decode_packet(da);
+    while (!da->waiting) {
+        int ret = da->ad_driver->decode_packet(da, &da->waiting);
         if (ret < 0)
             return ret;
     }
-    if (mp_audio_buffer_samples(da->decode_buffer) > 0) // avoid accidental flush
-        return AD_OK;
-    return reinit_audio_buffer(da) ? AD_OK : AD_ERR;
-}
-
-// Filter len bytes of input, put result into outbuf.
-static int filter_n_bytes(struct dec_audio *da, struct mp_audio_buffer *outbuf,
-                          int len)
-{
-    bool format_change = false;
-    int error = 0;
-
-    assert(len > 0); // would break EOF logic below
-
-    while (mp_audio_buffer_samples(da->decode_buffer) < len) {
-        // Check for a format change
-        struct mp_audio config;
-        mp_audio_buffer_get_format(da->decode_buffer, &config);
-        format_change = !mp_audio_config_equals(&da->decoded, &config);
-        if (format_change) {
-            error = AD_EOF; // drain remaining data left in the current buffer
-            break;
-        }
-        if (da->decoded.samples > 0) {
-            int copy = MPMIN(da->decoded.samples, len);
-            struct mp_audio append = da->decoded;
-            append.samples = copy;
-            mp_audio_buffer_append(da->decode_buffer, &append);
-            mp_audio_skip_samples(&da->decoded, copy);
-            da->pts_offset += copy;
-            continue;
-        }
-        error = da->ad_driver->decode_packet(da);
-        if (error < 0)
-            break;
-    }
-
-    if (error == AD_WAIT)
-        return error;
-
-    // Filter
-    struct mp_audio filter_data;
-    mp_audio_buffer_peek(da->decode_buffer, &filter_data);
-    filter_data.rate = da->afilter->input.rate; // due to playback speed change
-    len = MPMIN(filter_data.samples, len);
-    filter_data.samples = len;
-    bool eof = error == AD_EOF && filter_data.samples == 0;
-
-    if (af_filter(da->afilter, &filter_data, eof ? AF_FILTER_FLAG_EOF : 0) < 0)
-        return AD_ERR;
-
-    mp_audio_buffer_append(outbuf, &filter_data);
-    if (error == AD_EOF && filter_data.samples > 0)
-        error = 0; // don't end playback yet
-
-    // remove processed data from decoder buffer:
-    mp_audio_buffer_skip(da->decode_buffer, len);
-
-    // if format was changed, and all data was drained, execute the format change
-    if (format_change && eof) {
-        error = AD_NEW_FMT;
-        if (!reinit_audio_buffer(da))
-            error = AD_ERR; // switch to invalid format
-    }
-
-    return error;
+    talloc_steal(da, da->waiting);
+    da->decode_format = *da->waiting;
+    return mp_audio_config_valid(da->waiting) ? AD_OK : AD_ERR;
 }
 
 /* Try to get at least minsamples decoded+filtered samples in outbuf
@@ -301,52 +177,55 @@ static int filter_n_bytes(struct dec_audio *da, struct mp_audio_buffer *outbuf,
  * Return 0 on success, or negative AD_* error code.
  * In the former case outbuf has at least minsamples buffered on return.
  * In case of EOF/error it might or might not be. */
-int audio_decode(struct dec_audio *d_audio, struct mp_audio_buffer *outbuf,
+int audio_decode(struct dec_audio *da, struct mp_audio_buffer *outbuf,
                  int minsamples)
 {
-    if (!d_audio->afilter)
+    struct af_stream *afs = da->afilter;
+    if (afs->initialized < 1)
         return AD_ERR;
 
-    // Indicates that a filter seems to be buffering large amounts of data
-    int huge_filter_buffer = 0;
+    MP_STATS(da, "start audio");
 
-    /* Filter output size will be about filter_multiplier times input size.
-     * If some filter buffers audio in big blocks this might only hold
-     * as average over time. */
-    double filter_multiplier = af_calc_filter_multiplier(d_audio->afilter);
-
-    int prev_buffered = -1;
     int res = 0;
-    MP_STATS(d_audio, "start audio");
     while (res >= 0 && minsamples >= 0) {
         int buffered = mp_audio_buffer_samples(outbuf);
-        if (minsamples < buffered || buffered == prev_buffered)
+        if (minsamples < buffered)
             break;
-        prev_buffered = buffered;
 
-        int decsamples = (minsamples - buffered) / filter_multiplier;
-        // + some extra for possible filter buffering, and avoid 0
-        decsamples += 512;
+        res = 0;
 
-        if (huge_filter_buffer) {
-            /* Some filter must be doing significant buffering if the estimated
-             * input length didn't produce enough output from filters.
-             * Feed the filters 250 samples at a time until we have enough
-             * output. Very small amounts could make filtering inefficient while
-             * large amounts can make mpv demux the file unnecessarily far ahead
-             * to get audio data and buffer video frames in memory while doing
-             * so. However the performance impact of either is probably not too
-             * significant as long as the value is not completely insane. */
-            decsamples = 250;
+        struct mp_audio *mpa = da->waiting;
+        if (!mpa)
+            res = da->ad_driver->decode_packet(da, &mpa);
+
+        if (res != AD_EOF) {
+            if (res < 0)
+                break;
+            if (!mpa )
+                continue;
         }
 
-        /* if this iteration does not fill buffer, we must have lots
-         * of buffering in filters */
-        huge_filter_buffer = 1;
+        if (mpa) {
+            da->pts_offset += mpa->samples;
+            da->decode_format = *mpa;
+            mp_audio_set_null_data(&da->decode_format);
+            // On format change, make sure to drain the filter chain.
+            if (!mp_audio_config_equals(&afs->input, mpa)) {
+                res = AD_NEW_FMT;
+                da->waiting = talloc_steal(da, mpa);
+                mpa = NULL;
+            }
+        }
 
-        res = filter_n_bytes(d_audio, outbuf, decsamples);
+        if (mpa)
+            da->waiting = NULL;
+
+        if (af_filter(afs, mpa, outbuf) < 0)
+            return AD_ERR;
     }
-    MP_STATS(d_audio, "end audio");
+
+    MP_STATS(da, "end audio");
+
     return res;
 }
 
@@ -354,10 +233,11 @@ void audio_reset_decoding(struct dec_audio *d_audio)
 {
     if (d_audio->ad_driver)
         d_audio->ad_driver->control(d_audio, ADCTRL_RESET, NULL);
-    if (d_audio->afilter)
-        af_control_all(d_audio->afilter, AF_CONTROL_RESET, NULL);
+    af_control_all(d_audio->afilter, AF_CONTROL_RESET, NULL);
     d_audio->pts = MP_NOPTS_VALUE;
     d_audio->pts_offset = 0;
-    if (d_audio->decode_buffer)
-        mp_audio_buffer_clear(d_audio->decode_buffer);
+    if (d_audio->waiting) {
+        talloc_free(d_audio->waiting);
+        d_audio->waiting = NULL;
+    }
 }

audio/decode/dec_audio.h

@@ -33,13 +33,13 @@ struct dec_audio {
     struct mpv_global *global;
     const struct ad_functions *ad_driver;
     struct sh_stream *header;
-    struct mp_audio_buffer *decode_buffer;
     struct af_stream *afilter;
     char *decoder_desc;
-    struct replaygain_data *replaygain_data;
     int init_retries;
+    struct mp_audio_pool *pool;
+    struct mp_audio decode_format;
+    struct mp_audio *waiting;   // used on format-change
     // set by decoder
-    struct mp_audio decoded;    // decoded audio set by last decode_packet() call
     int bitrate;                // input bitrate, can change with VBR sources
     // last known pts value in output from decoder
     double pts;
@@ -65,8 +65,4 @@ int initial_audio_decode(struct dec_audio *d_audio);
 void audio_reset_decoding(struct dec_audio *d_audio);
 void audio_uninit(struct dec_audio *d_audio);
 
-int audio_init_filters(struct dec_audio *d_audio, int in_samplerate,
-                       int *out_samplerate, struct mp_chmap *out_channels,
-                       int *out_format);
-
 #endif /* MPLAYER_DEC_AUDIO_H */

audio/filter/af.c

@@ -28,6 +28,7 @@
 #include "options/m_option.h"
 #include "options/m_config.h"
 
+#include "audio/audio_buffer.h"
 #include "af.h"
 
 // Static list of filters
@@ -53,6 +54,7 @@ extern const struct af_info af_info_center;
 extern const struct af_info af_info_sinesuppress;
 extern const struct af_info af_info_karaoke;
 extern const struct af_info af_info_scaletempo;
+extern const struct af_info af_info_forcespeed;
 extern const struct af_info af_info_bs2b;
 extern const struct af_info af_info_lavfi;
 extern const struct af_info af_info_convert24;
@@ -83,6 +85,7 @@ static const struct af_info *const filter_list[] = {
     &af_info_center,
     &af_info_sinesuppress,
     &af_info_karaoke,
+    &af_info_forcespeed,
     &af_info_scaletempo,
 #if HAVE_LIBBS2B
     &af_info_bs2b,
@@ -121,13 +124,6 @@ const struct m_obj_list af_obj_list = {
     },
 };
 
-static bool af_config_equals(struct mp_audio *a, struct mp_audio *b)
-{
-    return a->format == b->format
-        && mp_chmap_equals(&a->channels, &b->channels)
-        && a->rate   == b->rate;
-}
-
 static void af_copy_unset_fields(struct mp_audio *dst, struct mp_audio *src)
 {
     if (dst->format == AF_FORMAT_UNKNOWN)
@@ -162,7 +158,7 @@ static int output_control(struct af_instance* af, int cmd, void* arg)
         *filter_output = *output;
         af_copy_unset_fields(filter_output, in);
         *in = *filter_output;
-        return af_config_equals(in, &orig_in) ? AF_OK : AF_FALSE;
+        return mp_audio_config_equals(in, &orig_in) ? AF_OK : AF_FALSE;
     }
     }
     return AF_UNKNOWN;
@@ -312,11 +308,8 @@ static void af_print_filter_chain(struct af_stream *s, struct af_instance *at,
     while (af) {
         char b[128] = {0};
         mp_snprintf_cat(b, sizeof(b), "  [%s] ", af->info->name);
-        if (af->data) {
-            char *info = mp_audio_config_to_str(af->data);
-            mp_snprintf_cat(b, sizeof(b), "%s", info);
-            talloc_free(info);
-        }
+        if (af->data)
+            mp_snprintf_cat(b, sizeof(b), "%s", mp_audio_config_to_str(af->data));
         if (af == at)
             mp_snprintf_cat(b, sizeof(b), " <-");
         MP_MSG(s, msg_level, "%s\n", b);
@@ -324,9 +317,7 @@ static void af_print_filter_chain(struct af_stream *s, struct af_instance *at,
         af = af->next;
     }
 
-    char *info = mp_audio_config_to_str(&s->output);
-    MP_MSG(s, msg_level, "  [ao] %s\n", info);
-    talloc_free(info);
+    MP_MSG(s, msg_level, "  [ao] %s\n", mp_audio_config_to_str(&s->output));
 }
 
 static int af_count_filters(struct af_stream *s)
@@ -523,11 +514,16 @@ static int af_reinit(struct af_stream *s)
         // Reset just in case...
         mp_audio_set_null_data(&in);
 
+        if (!mp_audio_config_valid(&in))
+            goto error;
+
         int rv = af->control(af, AF_CONTROL_REINIT, &in);
         if (rv == AF_OK && !mp_audio_config_equals(&in, af->prev->data))
             rv = AF_FALSE; // conversion filter needed
         switch (rv) {
         case AF_OK:
+            if (!mp_audio_config_valid(af->data))
+                goto error;
             af = af->next;
             break;
         case AF_FALSE: { // Configuration filter is needed
@@ -558,22 +554,27 @@ static int af_reinit(struct af_stream *s)
             MP_ERR(s, "Reinitialization did not work, "
                    "audio filter '%s' returned error code %i\n",
                    af->info->name, rv);
-            af_print_filter_chain(s, af, MSGL_ERR);
-            return AF_ERROR;
+            goto error;
         }
     }
 
-    af_print_filter_chain(s, NULL, MSGL_V);
-
     /* Set previously unset fields in s->output to those of the filter chain
      * output. This is used to make the output format fixed, and even if you
      * insert new filters or change the input format, the output format won't
      * change. (Audio outputs generally can't change format at runtime.) */
     af_copy_unset_fields(&s->output, &s->filter_output);
-    return af_config_equals(&s->output, &s->filter_output) ? AF_OK : AF_ERROR;
+    if (mp_audio_config_equals(&s->output, &s->filter_output)) {
+        s->initialized = 1;
+        af_print_filter_chain(s, NULL, MSGL_V);
+        return AF_OK;
+    }
+
+    goto error;
 
 negotiate_error:
     MP_ERR(s, "Unable to convert audio input format to output format.\n");
+error:
+    s->initialized = -1;
     af_print_filter_chain(s, af, MSGL_ERR);
     return AF_ERROR;
 }
@@ -583,6 +584,7 @@ void af_uninit(struct af_stream *s)
 {
     while (s->first->next && s->first->next != s->last)
         af_remove(s, s->first->next);
+    s->initialized = 0;
 }
 
 struct af_stream *af_new(struct mpv_global *global)
@@ -632,10 +634,6 @@ void af_destroy(struct af_stream *s)
    The return value is 0 if success and -1 if failure */
 int af_init(struct af_stream *s)
 {
-    // Sanity check
-    if (!s)
-        return -1;
-
     // Precaution in case caller is misbehaving
     mp_audio_set_null_data(&s->input);
     mp_audio_set_null_data(&s->output);
@@ -645,17 +643,20 @@ int af_init(struct af_stream *s)
         // Add all filters in the list (if there are any)
         struct m_obj_settings *list = s->opts->af_settings;
         for (int i = 0; list && list[i].name; i++) {
-            if (!af_prepend(s, s->last, list[i].name, list[i].attribs)) {
+            struct af_instance *af =
+                af_prepend(s, s->last, list[i].name, list[i].attribs);
+            if (!af) {
                 af_uninit(s);
+                s->initialized = -1;
                 return -1;
             }
+            af->label = talloc_strdup(af, list[i].label);
         }
     }
 
     if (af_reinit(s) != AF_OK) {
         // Something is stuffed audio out will not work
         MP_ERR(s, "Could not create audio filter chain.\n");
-        af_uninit(s);
         return -1;
     }
     return 0;
@@ -668,9 +669,6 @@ int af_init(struct af_stream *s)
 struct af_instance *af_add(struct af_stream *s, char *name, char **args)
 {
     struct af_instance *new;
-    // Sanity check
-    if (!s || !s->first || !name)
-        return NULL;
     // Insert the filter somewhere nice
     if (af_is_conversion_filter(s->first->next))
         new = af_append(s, s->first->next, name, args);
@@ -681,30 +679,85 @@ struct af_instance *af_add(struct af_stream *s, char *name, char **args)
 
     // Reinitalize the filter list
     if (af_reinit(s) != AF_OK) {
-        af_uninit(s);
-        af_init(s);
+        af_remove(s, new);
+        if (af_reinit(s) != AF_OK) {
+            af_uninit(s);
+            af_init(s);
+        }
         return NULL;
     }
     return new;
 }
 
-/* Filter data chunk through the filters in the list.
- * On success, *data is set to the filtered data/format.
- * Warning: input audio data will be overwritten.
+struct af_instance *af_find_by_label(struct af_stream *s, char *label)
+{
+    for (struct af_instance *af = s->first; af; af = af->next) {
+        if (af->label && strcmp(af->label, label) == 0)
+            return af;
+    }
+    return NULL;
+}
+
+/* Remove the first filter that matches this name. Return number of filters
+ * removed (0, 1), or a negative error code if reinit after removing failed.
  */
-int af_filter(struct af_stream *s, struct mp_audio *data, int flags)
+int af_remove_by_label(struct af_stream *s, char *label)
+{
+    struct af_instance *af = af_find_by_label(s, label);
+    if (!af)
+        return 0;
+    af_remove(s, af);
+    if (af_reinit(s) != AF_OK) {
+        af_uninit(s);
+        af_init(s);
+        return -1;
+    }
+    return 1;
+}
+
+/* Feed "data" to the chain, and write results to output. "data" needs to be
+ * a refcounted frame, although refcounting is not used yet.
+ * data==NULL means EOF.
+ */
+int af_filter(struct af_stream *s, struct mp_audio *data,
+              struct mp_audio_buffer *output)
 {
     struct af_instance *af = s->first;
-    assert(mp_audio_config_equals(af->data, data));
+    assert(s->initialized > 0);
+    int flags = 0;
+    int r = 0;
+    struct mp_audio tmp;
+    char dummy[MP_NUM_CHANNELS];
+    if (data) {
+        assert(mp_audio_config_equals(af->data, data));
+        r = mp_audio_make_writeable(data);
+    } else {
+        data = &tmp;
+        *data = *(af->data);
+        mp_audio_set_null_data(data);
+        flags = AF_FILTER_FLAG_EOF;
+        for (int n = 0; n < MP_NUM_CHANNELS; n++)
+            data->planes[n] = &dummy[n];
+    }
+    if (r < 0)
+        goto done;
+    struct mp_audio frame = *data;
+    for (int n = 0; n < MP_NUM_CHANNELS; n++)
+        frame.allocated[n] = NULL;
     // Iterate through all filters
     while (af) {
-        int r = af->filter(af, data, flags);
+        r = af->filter(af, &frame, flags);
         if (r < 0)
-            return r;
-        assert(mp_audio_config_equals(af->data, data));
+            goto done;
+        assert(mp_audio_config_equals(af->data, &frame));
         af = af->next;
     }
-    return 0;
+    mp_audio_buffer_append(output, &frame);
+
+done:
+    if (data != &tmp)
+        talloc_free(data);
+    return r;
 }
 
 // Calculate average ratio of filter output samples to input samples.

audio/filter/af.h

@@ -76,10 +76,13 @@ struct af_instance {
                  * the number of samples passed though. (Ratio of input
                  * and output, e.g. mul=4 => 1 sample becomes 4 samples) .*/
     bool auto_inserted; // inserted by af.c, such as conversion filters
+    char *label;
 };
 
 // Current audio stream
 struct af_stream {
+    int initialized; // 0: no, 1: yes, -1: attempted to, but failed
+
     // The first and last filter in the list
     struct af_instance *first;
     struct af_instance *last;
@@ -117,6 +120,7 @@ enum af_control {
     AF_CONTROL_SET_PAN_BALANCE,
     AF_CONTROL_GET_PAN_BALANCE,
     AF_CONTROL_SET_PLAYBACK_SPEED,
+    AF_CONTROL_SET_PLAYBACK_SPEED_RESAMPLE,
 };
 
 // Argument for AF_CONTROL_SET_PAN_LEVEL
@@ -130,7 +134,11 @@ void af_destroy(struct af_stream *s);
 int af_init(struct af_stream *s);
 void af_uninit(struct af_stream *s);
 struct af_instance *af_add(struct af_stream *s, char *name, char **args);
-int af_filter(struct af_stream *s, struct mp_audio *data, int flags);
+int af_remove_by_label(struct af_stream *s, char *label);
+struct af_instance *af_find_by_label(struct af_stream *s, char *label);
+struct mp_audio_buffer;
+int af_filter(struct af_stream *s, struct mp_audio *data,
+              struct mp_audio_buffer *output);
 struct af_instance *af_control_any_rev(struct af_stream *s, int cmd, void *arg);
 void af_control_all(struct af_stream *s, int cmd, void *arg);
 

audio/filter/af_bs2b.c

@@ -48,20 +48,12 @@ static int filter_##name(struct af_instance *af, struct mp_audio *data, int f) \
 }
 
 FILTER(f, float)
-FILTER(fbe, float)
-FILTER(fle, float)
-FILTER(s32be, int32_t)
-FILTER(u32be, uint32_t)
-FILTER(s32le, int32_t)
-FILTER(u32le, uint32_t)
-FILTER(s24be, bs2b_int24_t)
-FILTER(u24be, bs2b_uint24_t)
-FILTER(s24le, bs2b_int24_t)
-FILTER(u24le, bs2b_uint24_t)
-FILTER(s16be, int16_t)
-FILTER(u16be, uint16_t)
-FILTER(s16le, int16_t)
-FILTER(u16le, uint16_t)
+FILTER(s32, int32_t)
+FILTER(u32, uint32_t)
+FILTER(s24, bs2b_int24_t)
+FILTER(u24, bs2b_uint24_t)
+FILTER(s16, int16_t)
+FILTER(u16, uint16_t)
 FILTER(s8, int8_t)
 FILTER(u8, uint8_t)
 
@@ -85,47 +77,26 @@ static int control(struct af_instance *af, int cmd, void *arg)
         /* check for formats supported by libbs2b
            and assign corresponding handlers */
         switch (format) {
-            case AF_FORMAT_FLOAT_BE:
-                af->filter = filter_fbe;
+            case AF_FORMAT_FLOAT:
+                af->filter = filter_f;
                 break;
-            case AF_FORMAT_FLOAT_LE:
-                af->filter = filter_fle;
+            case AF_FORMAT_S32:
+                af->filter = filter_s32;
                 break;
-            case AF_FORMAT_S32_BE:
-                af->filter = filter_s32be;
+            case AF_FORMAT_U32:
+                af->filter = filter_u32;
                 break;
-            case AF_FORMAT_U32_BE:
-                af->filter = filter_u32be;
+            case AF_FORMAT_S24:
+                af->filter = filter_s24;
                 break;
-            case AF_FORMAT_S32_LE:
-                af->filter = filter_s32le;
+            case AF_FORMAT_U24:
+                af->filter = filter_u24;
                 break;
-            case AF_FORMAT_U32_LE:
-                af->filter = filter_u32le;
+            case AF_FORMAT_S16:
+                af->filter = filter_s16;
                 break;
-            case AF_FORMAT_S24_BE:
-                af->filter = filter_s24be;
-                break;
-            case AF_FORMAT_U24_BE:
-                af->filter = filter_u24be;
-                break;
-            case AF_FORMAT_S24_LE:
-                af->filter = filter_s24le;
-                break;
-            case AF_FORMAT_U24_LE:
-                af->filter = filter_u24le;
-                break;
-            case AF_FORMAT_S16_BE:
-                af->filter = filter_s16be;
-                break;
-            case AF_FORMAT_U16_BE:
-                af->filter = filter_u16be;
-                break;
-            case AF_FORMAT_S16_LE:
-                af->filter = filter_s16le;
-                break;
-            case AF_FORMAT_U16_LE:
-                af->filter = filter_u16le;
+            case AF_FORMAT_U16:
+                af->filter = filter_u16;
                 break;
             case AF_FORMAT_S8:
                 af->filter = filter_s8;

audio/filter/af_convert24.c

@@ -20,6 +20,7 @@
 
 #include "audio/format.h"
 #include "af.h"
+#include "osdep/endian.h"
 
 static bool test_conversion(int src_format, int dst_format)
 {

audio/filter/af_convertsignendian.c

@@ -20,21 +20,16 @@
 
 #include "af.h"
 #include "audio/format.h"
-#include "compat/mpbswap.h"
+#include "osdep/endian.h"
 
 static bool test_conversion(int src_format, int dst_format)
 {
     if ((src_format & AF_FORMAT_PLANAR) ||
         (dst_format & AF_FORMAT_PLANAR))
         return false;
-    int src_noend = src_format & ~AF_FORMAT_END_MASK;
-    int dst_noend = dst_format & ~AF_FORMAT_END_MASK;
-    // We can swap endian for all formats, but sign only for integer formats.
-    if (src_noend == dst_noend)
-        return true;
-    if (((src_noend & ~AF_FORMAT_SIGN_MASK) ==
-         (dst_noend & ~AF_FORMAT_SIGN_MASK)) &&
-        ((src_noend & AF_FORMAT_POINT_MASK) == AF_FORMAT_I))
+    if (((src_format & ~AF_FORMAT_SIGN_MASK) ==
+         (dst_format & ~AF_FORMAT_SIGN_MASK)) &&
+        ((src_format & AF_FORMAT_TYPE_MASK) == AF_FORMAT_I))
         return true;
     return false;
 }
@@ -63,34 +58,11 @@ static int control(struct af_instance *af, int cmd, void *arg)
     return AF_UNKNOWN;
 }
 
-static void endian(void *data, int len, int bps)
-{
-    switch (bps) {
-    case 2:
-        for (int i = 0; i < len; i++) {
-            ((uint16_t*)data)[i] = bswap_16(((uint16_t *)data)[i]);
-        }
-        break;
-    case 3:
-        for(int i = 0; i < len; i++) {
-            uint8_t s = ((uint8_t *)data)[3 * i];
-            ((uint8_t *)data)[3 * i] = ((uint8_t *)data)[3 * i + 2];
-            ((uint8_t *)data)[3 * i + 2] = s;
-        }
-        break;
-    case 4:
-        for(int i = 0; i < len; i++) {
-            ((uint32_t*)data)[i] = bswap_32(((uint32_t *)data)[i]);
-        }
-        break;
-    }
-}
-
-static void si2us(void *data, int len, int bps, bool le)
+static void si2us(void *data, int len, int bps)
 {
     ptrdiff_t i = -(len * bps);
     uint8_t *p = &((uint8_t *)data)[len * bps];
-    if (le && bps > 1)
+    if (BYTE_ORDER == LITTLE_ENDIAN && bps > 1)
         p += bps - 1;
     if (len <= 0)
         return;
@@ -105,12 +77,8 @@ static int filter(struct af_instance *af, struct mp_audio *data, int flags)
     int outfmt = af->data->format;
     size_t len = data->samples * data->nch;
 
-    if ((infmt & AF_FORMAT_END_MASK) != (outfmt & AF_FORMAT_END_MASK))
-        endian(data->planes[0], len, data->bps);
-
     if ((infmt & AF_FORMAT_SIGN_MASK) != (outfmt & AF_FORMAT_SIGN_MASK))
-        si2us(data->planes[0], len, data->bps,
-              (outfmt & AF_FORMAT_END_MASK) == AF_FORMAT_LE);
+        si2us(data->planes[0], len, data->bps);
 
     mp_audio_set_format(data, outfmt);
     return 0;
@@ -124,8 +92,8 @@ static int af_open(struct af_instance *af)
 }
 
 const struct af_info af_info_convertsignendian = {
-    .info = "Convert between sample format sign/endian",
-    .name = "convertsignendian",
+    .info = "Convert between sample format sign",
+    .name = "convertsign",
     .open = af_open,
     .test_conversion = test_conversion,
 };

audio/filter/af_export.c

@@ -78,8 +78,7 @@ static int control(struct af_instance* af, int cmd, void* arg)
     int mapsize;
 
     // Free previous buffers
-    if (s->buf)
-      free(s->buf[0]);
+    free(s->buf[0]);
 
     // unmap previous area
     if(s->mmap_area)
@@ -94,8 +93,10 @@ static int control(struct af_instance* af, int cmd, void* arg)
 
     // Allocate new buffers (as one continuous block)
     s->buf[0] = calloc(s->sz*af->data->nch, af->data->bps);
-    if(NULL == s->buf[0])
+    if(NULL == s->buf[0]) {
       MP_FATAL(af, "Out of memory\n");
+      return AF_ERROR;
+    }
     for(i = 1; i < af->data->nch; i++)
       s->buf[i] = (uint8_t *)s->buf[0] + i*s->sz*af->data->bps;
 
@@ -147,8 +148,8 @@ static int control(struct af_instance* af, int cmd, void* arg)
 static void uninit( struct af_instance* af )
 {
     af_export_t* s = af->priv;
-    if (s->buf)
-      free(s->buf[0]);
+
+    free(s->buf[0]);
 
     // Free mmaped area
     if(s->mmap_area)

audio/filter/af_forcespeed.c

@@ -0,0 +1,69 @@
+/*
+ * This file is part of mpv.
+ *
+ * mpv is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * mpv is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with mpv.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "af.h"
+
+struct priv {
+    double speed;
+};
+
+static int control(struct af_instance *af, int cmd, void *arg)
+{
+    struct priv *priv = af->priv;
+
+    switch (cmd) {
+    case AF_CONTROL_REINIT: {
+        struct mp_audio *in = arg;
+        struct mp_audio orig_in = *in;
+        struct mp_audio *out = af->data;
+
+        mp_audio_copy_config(out, in);
+        out->rate = in->rate * priv->speed;
+
+        return mp_audio_config_equals(in, &orig_in) ? AF_OK : AF_FALSE;
+    }
+    case AF_CONTROL_SET_PLAYBACK_SPEED_RESAMPLE: {
+        priv->speed = *(double *)arg;
+        return AF_OK;
+    }
+    }
+    return AF_UNKNOWN;
+}
+
+static int filter(struct af_instance *af, struct mp_audio *data, int flags)
+{
+    mp_audio_copy_config(data, af->data);
+    return 0;
+}
+
+static int af_open(struct af_instance *af)
+{
+    struct priv *priv = af->priv;
+    af->control = control;
+    af->filter = filter;
+    priv->speed = 1.0;
+    return AF_OK;
+}
+
+#define OPT_BASE_STRUCT struct priv
+
+const struct af_info af_info_forcespeed = {
+    .info = "Force audio speed",
+    .name = "forcespeed",
+    .open = af_open,
+    .priv_size = sizeof(struct priv),
+};

audio/filter/af_format.c

@@ -80,12 +80,12 @@ static int control(struct af_instance *af, int cmd, void *arg)
         force_out_params(af, out);
 
         if (in->nch != out->nch || in->bps != out->bps) {
-            MP_ERR(af, "[af_format] Forced input/output formats are incompatible.\n");
+            MP_ERR(af, "Forced input/output formats are incompatible.\n");
             return AF_ERROR;
         }
 
         if (priv->fail) {
-            MP_ERR(af, "[af_format] Failing on purpose.\n");
+            MP_ERR(af, "Failing on purpose.\n");
             return AF_ERROR;
         }
 

audio/filter/af_hrtf.c

@@ -284,6 +284,27 @@ static inline void update_ch(af_hrtf_t *s, short *in, const int k)
     s->ba_r[k] = in[4] + in[1] + in[3];
 }
 
+static void clear_coeff(af_hrtf_t *s, float *c)
+{
+    memset(c, 0, s->dlbuflen * sizeof(float));
+}
+
+static void reset(af_hrtf_t *s)
+{
+    clear_coeff(s, s->lf);
+    clear_coeff(s, s->rf);
+    clear_coeff(s, s->lr);
+    clear_coeff(s, s->rr);
+    clear_coeff(s, s->cf);
+    clear_coeff(s, s->cr);
+    clear_coeff(s, s->ba_l);
+    clear_coeff(s, s->ba_r);
+    clear_coeff(s, s->fwrbuf_l);
+    clear_coeff(s, s->fwrbuf_r);
+    clear_coeff(s, s->fwrbuf_lr);
+    clear_coeff(s, s->fwrbuf_rr);
+}
+
 /* Initialization and runtime control */
 static int control(struct af_instance *af, int cmd, void* arg)
 {
@@ -292,30 +313,27 @@ static int control(struct af_instance *af, int cmd, void* arg)
 
     switch(cmd) {
     case AF_CONTROL_REINIT:
-        af->data->rate   = ((struct mp_audio*)arg)->rate;
-        if(af->data->rate != 48000) {
-            // automatic samplerate adjustment in the filter chain
-            // is not yet supported.
-            MP_ERR(af, "ERROR: Sampling rate is not 48000 Hz (%d)!\n",
-                   af->data->rate);
-            return AF_ERROR;
-        }
+        reset(s);
+        af->data->rate = 48000;
         mp_audio_set_channels_old(af->data, ((struct mp_audio*)arg)->nch);
-            if(af->data->nch == 2) {
-               /* 2 channel input */
-               if(s->decode_mode != HRTF_MIX_MATRIX2CH) {
-                  /* Default behavior is stereo mixing. */
-                  s->decode_mode = HRTF_MIX_STEREO;
-               }
+        if(af->data->nch == 2) {
+            /* 2 channel input */
+            if(s->decode_mode != HRTF_MIX_MATRIX2CH) {
+                /* Default behavior is stereo mixing. */
+                s->decode_mode = HRTF_MIX_STEREO;
             }
-            else if (af->data->nch < 5)
-              mp_audio_set_channels_old(af->data, 5);
+        } else if (af->data->nch < 5) {
+            mp_audio_set_channels_old(af->data, 5);
+        }
         mp_audio_set_format(af->data, AF_FORMAT_S16);
         test_output_res = af_test_output(af, (struct mp_audio*)arg);
         // after testing input set the real output format
         mp_audio_set_num_channels(af->data, 2);
         s->print_flag = 1;
         return test_output_res;
+    case AF_CONTROL_RESET:
+        reset(s);
+        return AF_OK;
     }
 
     return AF_UNKNOWN;

audio/filter/af_lavcac3enc.c

@@ -30,6 +30,7 @@
 #include <libavutil/audioconvert.h>
 #include <libavutil/intreadwrite.h>
 #include <libavutil/common.h>
+#include <libavutil/bswap.h>
 #include <libavutil/mem.h>
 
 #include "common/common.h"
@@ -87,7 +88,7 @@ static int control(struct af_instance *af, int cmd, void *arg)
         if (in->nch > AC3_MAX_CHANNELS)
             mp_audio_set_num_channels(in, AC3_MAX_CHANNELS);
 
-        mp_audio_set_format(af->data, AF_FORMAT_AC3_BE);
+        mp_audio_set_format(af->data, AF_FORMAT_S_AC3);
         mp_audio_set_num_channels(af->data, 2);
 
         if (!mp_audio_config_equals(in, &orig_in))
@@ -150,6 +151,12 @@ static void uninit(struct af_instance* af)
     }
 }
 
+static void swap_16(uint16_t *ptr, size_t size)
+{
+    for (size_t n = 0; n < size; n++)
+        ptr[n] = av_bswap16(ptr[n]);
+}
+
 // Filter data through filter
 static int filter(struct af_instance* af, struct mp_audio* audio, int flags)
 {
@@ -223,11 +230,11 @@ static int filter(struct af_instance* af, struct mp_audio* audio, int flags)
             frame_size = AC3_FRAME_SIZE * 2 * 2;
             header_len = 8;
 
-            AV_WB16(hdr,     0xF872);   // iec 61937 syncword 1
-            AV_WB16(hdr + 2, 0x4E1F);   // iec 61937 syncword 2
-            hdr[4] = bsmod;             // bsmod
-            hdr[5] = 0x01;              // data-type ac3
-            AV_WB16(hdr + 6, len << 3); // number of bits in payload
+            AV_WL16(hdr,     0xF872);   // iec 61937 syncword 1
+            AV_WL16(hdr + 2, 0x4E1F);   // iec 61937 syncword 2
+            hdr[5] = bsmod;             // bsmod
+            hdr[4] = 0x01;              // data-type ac3
+            AV_WL16(hdr + 6, len << 3); // number of bits in payload
         }
 
         size_t max_size = (max_out_samples - out->samples) * out->sstride;
@@ -239,6 +246,7 @@ static int filter(struct af_instance* af, struct mp_audio* audio, int flags)
         memcpy(buf + header_len, s->pkt.data, s->pkt.size);
         memset(buf + header_len + s->pkt.size, 0,
                frame_size - (header_len + s->pkt.size));
+        swap_16((uint16_t *)(buf + header_len), s->pkt.size / 2);
         out->samples += frame_size / out->sstride;
     }
 
@@ -312,12 +320,13 @@ const struct af_info af_info_lavcac3enc = {
     .priv_size = sizeof(struct af_ac3enc_s),
     .priv_defaults = &(const struct af_ac3enc_s){
         .cfg_add_iec61937_header = 1,
-        .cfg_min_channel_num = 5,
+        .cfg_bit_rate = 640,
+        .cfg_min_channel_num = 3,
     },
     .options = (const struct m_option[]) {
         OPT_FLAG("tospdif", cfg_add_iec61937_header, 0),
         OPT_CHOICE_OR_INT("bitrate", cfg_bit_rate, 0, 32, 640,
-                          ({"default", 0})),
+                          ({"auto", 0}, {"default", 0})),
         OPT_INTRANGE("minch", cfg_min_channel_num, 0, 2, 6),
         {0}
     },

audio/filter/af_lavrresample.c

@@ -214,8 +214,7 @@ static int configure_lavrr(struct af_instance *af, struct mp_audio *in,
     if (avresample_open(s->avrctx) < 0 ||
         avresample_open(s->avrctx_out) < 0)
     {
-        MP_ERR(af, "Cannot open "
-                "Libavresample Context. \n");
+        MP_ERR(af, "Cannot open Libavresample Context. \n");
         return AF_ERROR;
     }
     return AF_OK;
@@ -339,7 +338,7 @@ static int filter(struct af_instance *af, struct mp_audio *data, int flags)
     if (needs_reorder(s->reorder_out, out->nch)) {
         if (af_fmt_is_planar(out->format)) {
             reorder_planes(data, s->reorder_out);
-        } else {
+        } else if (out->samples) {
             int out_size = out->samples * out->sstride;
             if (talloc_get_size(s->reorder_buffer) < out_size)
                 s->reorder_buffer = talloc_realloc_size(s, s->reorder_buffer, out_size);
@@ -347,7 +346,8 @@ static int filter(struct af_instance *af, struct mp_audio *data, int flags)
             int out_samples = avresample_convert(s->avrctx_out,
                     (uint8_t **) data->planes, out_size, out->samples,
                     (uint8_t **) out->planes, out_size, out->samples);
-            assert(out_samples == data->samples);
+            if (out_samples < 0)
+                MP_ERR(af, "Reordering failed.\n");
         }
     }
 
@@ -371,8 +371,7 @@ static int af_open(struct af_instance *af)
     if (s->avrctx && s->avrctx_out) {
         return AF_OK;
     } else {
-        MP_ERR(af, "Cannot initialize "
-               "Libavresample Context. \n");
+        MP_ERR(af, "Cannot initialize Libavresample Context. \n");
         uninit(af);
         return AF_ERROR;
     }

audio/filter/af_scaletempo.c

@@ -271,7 +271,7 @@ static int control(struct af_instance *af, int cmd, void *arg)
     switch (cmd) {
     case AF_CONTROL_REINIT: {
         struct mp_audio *data = (struct mp_audio *)arg;
-        float srate = data->rate / 1000;
+        float srate  = data->rate / 1000.0;
         int nch = data->nch;
         int use_int = 0;
 

audio/format.c

@@ -30,9 +30,6 @@
 
 int af_fmt2bps(int format)
 {
-    if (AF_FORMAT_IS_AC3(format)) return 2;
-    if (format == AF_FORMAT_UNKNOWN)
-        return 0;
     switch (format & AF_FORMAT_BITS_MASK) {
     case AF_FORMAT_8BIT:  return 1;
     case AF_FORMAT_16BIT: return 2;
@@ -62,7 +59,7 @@ static int bits_to_mask(int bits)
 
 int af_fmt_change_bits(int format, int bits)
 {
-    if (!af_fmt_is_valid(format) || (format & AF_FORMAT_SPECIAL_MASK))
+    if (!af_fmt_is_valid(format))
         return 0;
     int mask = bits_to_mask(bits);
     format = (format & ~AF_FORMAT_BITS_MASK) | mask;
@@ -109,35 +106,31 @@ bool af_fmt_is_planar(int format)
     return !!(format & AF_FORMAT_PLANAR);
 }
 
-#define FMT(string, id)                                 \
-    {string,            id},
-
-#define FMT_ENDIAN(string, id)                          \
-    {string,            id},                            \
-    {string "le",       MP_CONCAT(id, _LE)},            \
-    {string "be",       MP_CONCAT(id, _BE)},            \
-
 const struct af_fmt_entry af_fmtstr_table[] = {
-    FMT("mpeg2",                AF_FORMAT_MPEG2)
-    FMT_ENDIAN("ac3",           AF_FORMAT_AC3)
-    FMT_ENDIAN("iec61937",      AF_FORMAT_IEC61937)
+    {"u8",          AF_FORMAT_U8},
+    {"s8",          AF_FORMAT_S8},
+    {"u16",         AF_FORMAT_U16},
+    {"s16",         AF_FORMAT_S16},
+    {"u24",         AF_FORMAT_U24},
+    {"s24",         AF_FORMAT_S24},
+    {"u32",         AF_FORMAT_U32},
+    {"s32",         AF_FORMAT_S32},
+    {"float",       AF_FORMAT_FLOAT},
+    {"double",      AF_FORMAT_DOUBLE},
 
-    FMT("u8",                   AF_FORMAT_U8)
-    FMT("s8",                   AF_FORMAT_S8)
-    FMT_ENDIAN("u16",           AF_FORMAT_U16)
-    FMT_ENDIAN("s16",           AF_FORMAT_S16)
-    FMT_ENDIAN("u24",           AF_FORMAT_U24)
-    FMT_ENDIAN("s24",           AF_FORMAT_S24)
-    FMT_ENDIAN("u32",           AF_FORMAT_U32)
-    FMT_ENDIAN("s32",           AF_FORMAT_S32)
-    FMT_ENDIAN("float",         AF_FORMAT_FLOAT)
-    FMT_ENDIAN("double",        AF_FORMAT_DOUBLE)
+    {"u8p",         AF_FORMAT_U8P},
+    {"s16p",        AF_FORMAT_S16P},
+    {"s32p",        AF_FORMAT_S32P},
+    {"floatp",      AF_FORMAT_FLOATP},
+    {"doublep",     AF_FORMAT_DOUBLEP},
 
-    FMT("u8p",                  AF_FORMAT_U8P)
-    FMT("s16p",                 AF_FORMAT_S16P)
-    FMT("s32p",                 AF_FORMAT_S32P)
-    FMT("floatp",               AF_FORMAT_FLOATP)
-    FMT("doublep",              AF_FORMAT_DOUBLEP)
+    {"spdif-aac",   AF_FORMAT_S_AAC},
+    {"spdif-ac3",   AF_FORMAT_S_AC3},
+    {"spdif-dts",   AF_FORMAT_S_DTS},
+    {"spdif-dtshd", AF_FORMAT_S_DTSHD},
+    {"spdif-eac3",  AF_FORMAT_S_EAC3},
+    {"spdif-mp3",   AF_FORMAT_S_MP3},
+    {"spdif-truehd",AF_FORMAT_S_TRUEHD},
 
     {0}
 };
@@ -199,22 +192,17 @@ int af_format_conversion_score(int dst_format, int src_format)
         return INT_MIN;
     if (dst_format == src_format)
         return 1024;
-    // Just endian swapping (separate because it works for special formats)
-    if ((dst_format & ~AF_FORMAT_END_MASK) == (src_format & ~AF_FORMAT_END_MASK))
-        return 1024 - 2;
     // Can't be normally converted
     if (AF_FORMAT_IS_SPECIAL(dst_format) || AF_FORMAT_IS_SPECIAL(src_format))
         return INT_MIN;
     int score = 1024;
     if (FMT_DIFF(AF_FORMAT_INTERLEAVING_MASK, dst_format, src_format))
         score -= 1;     // has to (de-)planarize
-    if (FMT_DIFF(AF_FORMAT_END_MASK, dst_format, src_format))
-        score -= 2;     // has to swap endian
     if (FMT_DIFF(AF_FORMAT_SIGN_MASK, dst_format, src_format))
         score -= 4;     // has to swap sign
-    if (FMT_DIFF(AF_FORMAT_POINT_MASK, dst_format, src_format)) {
+    if (FMT_DIFF(AF_FORMAT_TYPE_MASK, dst_format, src_format)) {
         int dst_bits = dst_format & AF_FORMAT_BITS_MASK;
-        if ((dst_format & AF_FORMAT_POINT_MASK) == AF_FORMAT_F) {
+        if ((dst_format & AF_FORMAT_TYPE_MASK) == AF_FORMAT_F) {
             // For int->float, always prefer 32 bit float.
             score -= dst_bits == AF_FORMAT_32BIT ? 8 : 0;
         } else {
@@ -230,7 +218,7 @@ int af_format_conversion_score(int dst_format, int src_format)
         }
     }
     // Consider this the worst case.
-    if (FMT_DIFF(AF_FORMAT_POINT_MASK, dst_format, src_format))
+    if (FMT_DIFF(AF_FORMAT_TYPE_MASK, dst_format, src_format))
         score -= 2048;  // has to convert float<->int
     return score;
 }

audio/format.h

@@ -25,121 +25,77 @@
 
 #include <stdbool.h>
 
-#include "osdep/endian.h"
-#include "bstr/bstr.h"
-
-#if BYTE_ORDER == BIG_ENDIAN
-#define AF_SELECT_LE_BE(LE, BE) BE
-#else
-#define AF_SELECT_LE_BE(LE, BE) LE
-#endif
-
-// Endianness
-#define AF_FORMAT_BE            (0<<0) // Big Endian
-#define AF_FORMAT_LE            (1<<0) // Little Endian
-#define AF_FORMAT_END_MASK      (1<<0)
-
-#define AF_FORMAT_NE AF_SELECT_LE_BE(AF_FORMAT_LE, AF_FORMAT_BE)
+#include "misc/bstr.h"
 
 // Signed/unsigned
-#define AF_FORMAT_SI            (0<<1) // Signed
-#define AF_FORMAT_US            (1<<1) // Unsigned
-#define AF_FORMAT_SIGN_MASK     (1<<1)
+#define AF_FORMAT_SI            (0<<0) // Signed
+#define AF_FORMAT_US            (1<<0) // Unsigned
+#define AF_FORMAT_SIGN_MASK     (1<<0)
 
 // Bits used
 // Some code assumes they're sorted by size.
-#define AF_FORMAT_8BIT          (0<<3)
-#define AF_FORMAT_16BIT         (1<<3)
-#define AF_FORMAT_24BIT         (2<<3)
-#define AF_FORMAT_32BIT         (3<<3)
-#define AF_FORMAT_64BIT         (4<<3)
-#define AF_FORMAT_BITS_MASK     (7<<3)
+#define AF_FORMAT_8BIT          (0<<1)
+#define AF_FORMAT_16BIT         (1<<1)
+#define AF_FORMAT_24BIT         (2<<1)
+#define AF_FORMAT_32BIT         (3<<1)
+#define AF_FORMAT_64BIT         (4<<1)
+#define AF_FORMAT_BITS_MASK     (7<<1)
 
-// Special flags refering to non pcm data (note: 1<<6, 2<<6, 5<<6 unused)
-#define AF_FORMAT_S_MPEG2       (3<<6) // MPEG(2) audio
-#define AF_FORMAT_S_AC3         (4<<6) // Dolby Digital AC3
-#define AF_FORMAT_S_IEC61937    (6<<6)
-#define AF_FORMAT_SPECIAL_MASK  (7<<6)
-
-// Fixed or floating point
-#define AF_FORMAT_I             (1<<9) // Int
-#define AF_FORMAT_F             (2<<9) // Foating point
-#define AF_FORMAT_POINT_MASK    (3<<9)
+// Fixed/floating point/special
+#define AF_FORMAT_I             (1<<4) // Int
+#define AF_FORMAT_F             (2<<4) // Foating point
+#define AF_FORMAT_S             (4<<4) // special (IEC61937)
+#define AF_FORMAT_TYPE_MASK     (7<<4)
 
 // Interleaving (planar formats have data for each channel in separate planes)
-#define AF_FORMAT_INTERLEAVED        (0<<11) // must be 0
-#define AF_FORMAT_PLANAR             (1<<11)
-#define AF_FORMAT_INTERLEAVING_MASK  (1<<11)
+#define AF_FORMAT_INTERLEAVED        (0<<7) // must be 0
+#define AF_FORMAT_PLANAR             (1<<7)
+#define AF_FORMAT_INTERLEAVING_MASK  (1<<7)
+
+#define AF_FORMAT_S_CODEC(n)    ((n)<<8)
+#define AF_FORMAT_S_CODEC_MASK  (15 <<8) // 16 codecs max.
 
 #define AF_FORMAT_MASK          ((1<<12)-1)
 
 #define AF_INTP (AF_FORMAT_I|AF_FORMAT_PLANAR)
 #define AF_FLTP (AF_FORMAT_F|AF_FORMAT_PLANAR)
+#define AF_FORMAT_S_(n) (AF_FORMAT_S_CODEC(n)|AF_FORMAT_S|AF_FORMAT_16BIT)
 
 // actual sample formats
 enum af_format {
     AF_FORMAT_UNKNOWN   = 0,
 
-    AF_FORMAT_U8        = (AF_FORMAT_I|AF_FORMAT_US|AF_FORMAT_8BIT|AF_FORMAT_NE),
-    AF_FORMAT_S8        = (AF_FORMAT_I|AF_FORMAT_SI|AF_FORMAT_8BIT|AF_FORMAT_NE),
-    AF_FORMAT_U16_LE    = (AF_FORMAT_I|AF_FORMAT_US|AF_FORMAT_16BIT|AF_FORMAT_LE),
-    AF_FORMAT_U16_BE    = (AF_FORMAT_I|AF_FORMAT_US|AF_FORMAT_16BIT|AF_FORMAT_BE),
-    AF_FORMAT_S16_LE    = (AF_FORMAT_I|AF_FORMAT_SI|AF_FORMAT_16BIT|AF_FORMAT_LE),
-    AF_FORMAT_S16_BE    = (AF_FORMAT_I|AF_FORMAT_SI|AF_FORMAT_16BIT|AF_FORMAT_BE),
-    AF_FORMAT_U24_LE    = (AF_FORMAT_I|AF_FORMAT_US|AF_FORMAT_24BIT|AF_FORMAT_LE),
-    AF_FORMAT_U24_BE    = (AF_FORMAT_I|AF_FORMAT_US|AF_FORMAT_24BIT|AF_FORMAT_BE),
-    AF_FORMAT_S24_LE    = (AF_FORMAT_I|AF_FORMAT_SI|AF_FORMAT_24BIT|AF_FORMAT_LE),
-    AF_FORMAT_S24_BE    = (AF_FORMAT_I|AF_FORMAT_SI|AF_FORMAT_24BIT|AF_FORMAT_BE),
-    AF_FORMAT_U32_LE    = (AF_FORMAT_I|AF_FORMAT_US|AF_FORMAT_32BIT|AF_FORMAT_LE),
-    AF_FORMAT_U32_BE    = (AF_FORMAT_I|AF_FORMAT_US|AF_FORMAT_32BIT|AF_FORMAT_BE),
-    AF_FORMAT_S32_LE    = (AF_FORMAT_I|AF_FORMAT_SI|AF_FORMAT_32BIT|AF_FORMAT_LE),
-    AF_FORMAT_S32_BE    = (AF_FORMAT_I|AF_FORMAT_SI|AF_FORMAT_32BIT|AF_FORMAT_BE),
+    AF_FORMAT_U8        = (AF_FORMAT_I|AF_FORMAT_US|AF_FORMAT_8BIT),
+    AF_FORMAT_S8        = (AF_FORMAT_I|AF_FORMAT_SI|AF_FORMAT_8BIT),
+    AF_FORMAT_U16       = (AF_FORMAT_I|AF_FORMAT_US|AF_FORMAT_16BIT),
+    AF_FORMAT_S16       = (AF_FORMAT_I|AF_FORMAT_SI|AF_FORMAT_16BIT),
+    AF_FORMAT_U24       = (AF_FORMAT_I|AF_FORMAT_US|AF_FORMAT_24BIT),
+    AF_FORMAT_S24       = (AF_FORMAT_I|AF_FORMAT_SI|AF_FORMAT_24BIT),
+    AF_FORMAT_U32       = (AF_FORMAT_I|AF_FORMAT_US|AF_FORMAT_32BIT),
+    AF_FORMAT_S32       = (AF_FORMAT_I|AF_FORMAT_SI|AF_FORMAT_32BIT),
 
-    AF_FORMAT_FLOAT_LE  = (AF_FORMAT_F|AF_FORMAT_32BIT|AF_FORMAT_LE),
-    AF_FORMAT_FLOAT_BE  = (AF_FORMAT_F|AF_FORMAT_32BIT|AF_FORMAT_BE),
-
-    AF_FORMAT_DOUBLE_LE = (AF_FORMAT_F|AF_FORMAT_64BIT|AF_FORMAT_LE),
-    AF_FORMAT_DOUBLE_BE = (AF_FORMAT_F|AF_FORMAT_64BIT|AF_FORMAT_BE),
-
-    AF_FORMAT_AC3_LE    = (AF_FORMAT_S_AC3|AF_FORMAT_16BIT|AF_FORMAT_LE),
-    AF_FORMAT_AC3_BE    = (AF_FORMAT_S_AC3|AF_FORMAT_16BIT|AF_FORMAT_BE),
-
-    AF_FORMAT_IEC61937_LE = (AF_FORMAT_S_IEC61937|AF_FORMAT_16BIT|AF_FORMAT_LE),
-    AF_FORMAT_IEC61937_BE = (AF_FORMAT_S_IEC61937|AF_FORMAT_16BIT|AF_FORMAT_BE),
-
-    AF_FORMAT_MPEG2     = (AF_FORMAT_S_MPEG2),
+    AF_FORMAT_FLOAT     = (AF_FORMAT_F|AF_FORMAT_32BIT),
+    AF_FORMAT_DOUBLE    = (AF_FORMAT_F|AF_FORMAT_64BIT),
 
     // Planar variants
-    AF_FORMAT_U8P       = (AF_INTP|AF_FORMAT_US|AF_FORMAT_8BIT|AF_FORMAT_NE),
-    AF_FORMAT_S16P      = (AF_INTP|AF_FORMAT_SI|AF_FORMAT_16BIT|AF_FORMAT_NE),
-    AF_FORMAT_S32P      = (AF_INTP|AF_FORMAT_SI|AF_FORMAT_32BIT|AF_FORMAT_NE),
-    AF_FORMAT_FLOATP    = (AF_FLTP|AF_FORMAT_32BIT|AF_FORMAT_NE),
-    AF_FORMAT_DOUBLEP   = (AF_FLTP|AF_FORMAT_64BIT|AF_FORMAT_NE),
+    AF_FORMAT_U8P       = (AF_INTP|AF_FORMAT_US|AF_FORMAT_8BIT),
+    AF_FORMAT_S16P      = (AF_INTP|AF_FORMAT_SI|AF_FORMAT_16BIT),
+    AF_FORMAT_S32P      = (AF_INTP|AF_FORMAT_SI|AF_FORMAT_32BIT),
+    AF_FORMAT_FLOATP    = (AF_FLTP|AF_FORMAT_32BIT),
+    AF_FORMAT_DOUBLEP   = (AF_FLTP|AF_FORMAT_64BIT),
 
-    // Native endian variants
-    AF_FORMAT_U16       = AF_SELECT_LE_BE(AF_FORMAT_U16_LE, AF_FORMAT_U16_BE),
-    AF_FORMAT_S16       = AF_SELECT_LE_BE(AF_FORMAT_S16_LE, AF_FORMAT_S16_BE),
-    AF_FORMAT_U24       = AF_SELECT_LE_BE(AF_FORMAT_U24_LE, AF_FORMAT_U24_BE),
-    AF_FORMAT_S24       = AF_SELECT_LE_BE(AF_FORMAT_S24_LE, AF_FORMAT_S24_BE),
-    AF_FORMAT_U32       = AF_SELECT_LE_BE(AF_FORMAT_U32_LE, AF_FORMAT_U32_BE),
-    AF_FORMAT_S32       = AF_SELECT_LE_BE(AF_FORMAT_S32_LE, AF_FORMAT_S32_BE),
-
-    AF_FORMAT_FLOAT     = AF_SELECT_LE_BE(AF_FORMAT_FLOAT_LE, AF_FORMAT_FLOAT_BE),
-    AF_FORMAT_DOUBLE    = AF_SELECT_LE_BE(AF_FORMAT_DOUBLE_LE, AF_FORMAT_DOUBLE_BE),
-
-    AF_FORMAT_AC3       = AF_SELECT_LE_BE(AF_FORMAT_AC3_LE, AF_FORMAT_AC3_BE),
-    AF_FORMAT_IEC61937  = AF_SELECT_LE_BE(AF_FORMAT_IEC61937_LE, AF_FORMAT_IEC61937_BE),
+    // All of these use IEC61937 framing, and otherwise pretend to be like PCM.
+    AF_FORMAT_S_AAC     = AF_FORMAT_S_(0),
+    AF_FORMAT_S_AC3     = AF_FORMAT_S_(1),
+    AF_FORMAT_S_DTS     = AF_FORMAT_S_(2),
+    AF_FORMAT_S_DTSHD   = AF_FORMAT_S_(3),
+    AF_FORMAT_S_EAC3    = AF_FORMAT_S_(4),
+    AF_FORMAT_S_MP3     = AF_FORMAT_S_(5),
+    AF_FORMAT_S_TRUEHD  = AF_FORMAT_S_(6),
 };
 
-#define AF_FORMAT_IS_AC3(fmt) \
-    (((fmt) & AF_FORMAT_SPECIAL_MASK) == AF_FORMAT_S_AC3)
-
-#define AF_FORMAT_IS_IEC61937(fmt) \
-    (AF_FORMAT_IS_AC3(fmt) ||      \
-     ((fmt) & AF_FORMAT_SPECIAL_MASK) == AF_FORMAT_S_IEC61937)
-
-#define AF_FORMAT_IS_SPECIAL(fmt) \
-    ((fmt & AF_FORMAT_SPECIAL_MASK) != 0)
+#define AF_FORMAT_IS_IEC61937(f) (((f) & AF_FORMAT_TYPE_MASK) == AF_FORMAT_S)
+#define AF_FORMAT_IS_SPECIAL(f) AF_FORMAT_IS_IEC61937(f)
 
 struct af_fmt_entry {
     const char *name;

audio/mixer.c

@@ -37,6 +37,7 @@ struct mixer {
     struct af_stream *af;
     // Static, dependent on ao/softvol settings
     bool softvol;                       // use AO (true) or af_volume (false)
+    bool ao_softvol;                    // AO has private or per-app volume
     bool emulate_mute;                  // if true, emulate mute with volume=0
     // Last known values (possibly out of sync with reality)
     float vol_l, vol_r;
@@ -245,11 +246,13 @@ char *mixer_get_volume_restore_data(struct mixer *mixer)
 
 static void probe_softvol(struct mixer *mixer)
 {
+    mixer->ao_softvol =
+        ao_control(mixer->ao, AOCONTROL_HAS_SOFT_VOLUME, 0) == 1 ||
+        ao_control(mixer->ao, AOCONTROL_HAS_PER_APP_VOLUME, 0) == 1;
+
     if (mixer->opts->softvol == SOFTVOL_AUTO) {
         // No system-wide volume => fine with AO volume control.
-        mixer->softvol =
-            ao_control(mixer->ao, AOCONTROL_HAS_TEMP_VOLUME, 0) != 1 ||
-            ao_control(mixer->ao, AOCONTROL_HAS_PER_APP_VOLUME, 0) != 1;
+        mixer->softvol = !mixer->ao_softvol;
     } else {
         mixer->softvol = mixer->opts->softvol == SOFTVOL_YES;
     }
@@ -282,8 +285,7 @@ static void restore_volume(struct mixer *mixer)
     const char *prev_driver = mixer->driver;
     mixer->driver = mixer->softvol ? "softvol" : ao_get_name(ao);
 
-    bool restore =
-        mixer->softvol || ao_control(ao, AOCONTROL_HAS_TEMP_VOLUME, 0) == 1;
+    bool restore = mixer->softvol || mixer->ao_softvol;
 
     // Restore old parameters if volume won't survive reinitialization.
     // But not if volume scale is possibly different.
@@ -361,7 +363,7 @@ void mixer_uninit_audio(struct mixer *mixer)
         return;
 
     checkvolume(mixer);
-    if (mixer->muted_by_us) {
+    if (mixer->muted_by_us && !mixer->softvol && !mixer->ao_softvol) {
         /* Current audio output API combines playing the remaining buffered
          * audio and uninitializing the AO into one operation, even though
          * ideally unmute would happen between those two steps. We can't do

audio/out/ao.c

@@ -29,6 +29,7 @@
 #include "audio/format.h"
 #include "audio/audio.h"
 
+#include "input/input.h"
 #include "options/options.h"
 #include "options/m_config.h"
 #include "common/msg.h"
@@ -60,18 +61,15 @@ static const struct ao_driver * const audio_out_drivers[] = {
 #if HAVE_PULSE
     &audio_out_pulse,
 #endif
-#if HAVE_SNDIO
-    &audio_out_sndio,
-#endif
 #if HAVE_ALSA
     &audio_out_alsa,
 #endif
-#if HAVE_DSOUND
-    &audio_out_dsound,
-#endif
 #if HAVE_WASAPI
     &audio_out_wasapi,
 #endif
+#if HAVE_DSOUND
+    &audio_out_dsound,
+#endif
 #if HAVE_OSS_AUDIO
     &audio_out_oss,
 #endif
@@ -87,6 +85,12 @@ static const struct ao_driver * const audio_out_drivers[] = {
 #endif
 #if HAVE_SDL1 || HAVE_SDL2
     &audio_out_sdl,
+#endif
+#if HAVE_SNDIO
+    &audio_out_sndio,
+#endif
+#if HAVE_COREAUDIO
+    &audio_out_coreaudio_exclusive,
 #endif
     &audio_out_null,
     // should not be auto-selected:
@@ -94,9 +98,6 @@ static const struct ao_driver * const audio_out_drivers[] = {
 #if HAVE_ENCODING
     &audio_out_lavc,
 #endif
-#if HAVE_COREAUDIO
-    &audio_out_coreaudio_exclusive,
-#endif
 #if HAVE_RSOUND
     &audio_out_rsound,
 #endif
@@ -128,12 +129,34 @@ const struct m_obj_list ao_obj_list = {
     .allow_trailer = true,
 };
 
-static struct ao *ao_create(bool probing, struct mpv_global *global,
-                            struct input_ctx *input_ctx,
-                            struct encode_lavc_context *encode_lavc_ctx,
-                            int samplerate, int format, struct mp_chmap channels,
-                            char *name, char **args)
+// Return true if the ao with the given name matches the opt (--audio-device
+// contents). If opt is empty, matching always succeeds. If out_device is not
+// NULL, it will be set to the implied audio device (or NULL).
+static bool match_ao_driver(const char *ao_name, char *opt, char **out_device)
 {
+    bstr ao_nameb = bstr0(ao_name);
+    bstr optb = bstr0(opt);
+
+    char *dummy;
+    if (!out_device)
+        out_device = &dummy;
+    *out_device = NULL;
+
+    if (!optb.len || bstr_equals0(optb, "auto"))
+        return true;
+    if (!bstr_startswith(optb, ao_nameb))
+        return false;
+    if (optb.len > ao_nameb.len && optb.start[ao_nameb.len] != '/')
+        return false;
+    char *split = strchr(opt, '/');
+    *out_device = split ? split + 1 : NULL;
+    return true;
+}
+
+static struct ao *ao_alloc(bool probing, struct mpv_global *global,
+                           struct input_ctx *input_ctx, char *name, char **args)
+{
+    struct MPOpts *opts = global->opts;
     struct mp_log *log = mp_log_new(NULL, global->log, "ao");
     struct m_obj_desc desc;
     if (!m_obj_list_find(&desc, &ao_obj_list, bstr0(name))) {
@@ -146,33 +169,62 @@ static struct ao *ao_create(bool probing, struct mpv_global *global,
     *ao = (struct ao) {
         .driver = desc.p,
         .probing = probing,
-        .encode_lavc_ctx = encode_lavc_ctx,
         .input_ctx = input_ctx,
-        .samplerate = samplerate,
-        .channels = channels,
-        .format = format,
         .log = mp_log_new(ao, log, name),
+        .def_buffer = opts->audio_buffer,
     };
-    if (ao->driver->encode != !!ao->encode_lavc_ctx)
-        goto error;
     struct m_config *config = m_config_from_obj_desc(ao, ao->log, &desc);
-    if (m_config_apply_defaults(config, name, global->opts->ao_defs) < 0)
+    if (m_config_apply_defaults(config, name, opts->ao_defs) < 0)
         goto error;
     if (m_config_set_obj_params(config, args) < 0)
         goto error;
     ao->priv = config->optstruct;
+    return ao;
+error:
+    talloc_free(ao);
+    return NULL;
+}
 
-    char *chmap = mp_chmap_to_str(&ao->channels);
+static struct ao *ao_alloc_pb(bool probing, struct mpv_global *global,
+                              struct input_ctx *input_ctx,
+                              struct encode_lavc_context *encode_lavc_ctx,
+                              int samplerate, int format, struct mp_chmap channels,
+                              char *name, char **args)
+{
+    struct MPOpts *opts = global->opts;
+    struct ao *ao = ao_alloc(probing, global, input_ctx, name, args);
+    if (!ao)
+        return NULL;
+    ao->samplerate = samplerate;
+    ao->channels = channels;
+    ao->format = format;
+    ao->encode_lavc_ctx = encode_lavc_ctx;
+    if (ao->driver->encode != !!ao->encode_lavc_ctx) {
+        talloc_free(ao);
+        return NULL;
+    }
+
+    match_ao_driver(ao->driver->name, opts->audio_device, &ao->device);
+    ao->device = talloc_strdup(ao, ao->device);
+
+    ao->client_name = talloc_strdup(ao, opts->audio_client_name);
+
+    return ao;
+}
+
+static int ao_init(struct ao *ao)
+{
     MP_VERBOSE(ao, "requested format: %d Hz, %s channels, %s\n",
-               ao->samplerate, chmap, af_fmt_to_str(ao->format));
-    talloc_free(chmap);
+               ao->samplerate, mp_chmap_to_str(&ao->channels),
+               af_fmt_to_str(ao->format));
 
     ao->api = ao->driver->play ? &ao_api_push : &ao_api_pull;
     ao->api_priv = talloc_zero_size(ao, ao->api->priv_size);
     assert(!ao->api->priv_defaults && !ao->api->options);
 
-    if (ao->driver->init(ao) < 0)
-        goto error;
+    int r = ao->driver->init(ao);
+    if (r < 0)
+        return r;
 
     ao->sstride = af_fmt2bps(ao->format);
     ao->num_planes = 1;
@@ -183,18 +235,40 @@ static struct ao *ao_create(bool probing, struct mpv_global *global,
     }
     ao->bps = ao->samplerate * ao->sstride;
 
-    if (!ao->device_buffer && ao->driver->get_space) {
+    if (!ao->device_buffer && ao->driver->get_space)
         ao->device_buffer = ao->driver->get_space(ao);
+    if (ao->device_buffer)
         MP_VERBOSE(ao, "device buffer: %d samples.\n", ao->device_buffer);
-    }
-    ao->buffer = MPMAX(ao->device_buffer, MIN_BUFFER * ao->samplerate);
+    ao->buffer = MPMAX(ao->device_buffer, ao->def_buffer * ao->samplerate);
     MP_VERBOSE(ao, "using soft-buffer of %d samples.\n", ao->buffer);
 
-    if (ao->api->init(ao) < 0)
-        goto error;
+    return ao->api->init(ao);
+}
 
-    return ao;
-error:
+static struct ao *ao_create(bool probing, struct mpv_global *global,
+                            struct input_ctx *input_ctx,
+                            struct encode_lavc_context *encode_lavc_ctx,
+                            int samplerate, int format, struct mp_chmap channels,
+                            char *name, char **args)
+{
+    struct ao *ao = ao_alloc_pb(probing, global, input_ctx, encode_lavc_ctx,
+                                samplerate, format, channels, name, args);
+    if (ao && ao_init(ao) >= 0)
+        return ao;
+    // Silly exception for coreaudio spdif redirection
+    if (ao && ao->redirect) {
+        char redirect[80], device[80];
+        snprintf(redirect, sizeof(redirect), "%s", ao->redirect);
+        snprintf(device, sizeof(device), "%s", ao->device ? ao->device : "");
+        talloc_free(ao);
+        ao = ao_alloc_pb(probing, global, input_ctx, encode_lavc_ctx,
+                         samplerate, format, channels, redirect, args);
+        if (ao) {
+            ao->device = talloc_strdup(ao, device);
+            if (ao_init(ao) >= 0)
+                return ao;
+        }
+    }
     talloc_free(ao);
     return NULL;
 }
@@ -204,9 +278,10 @@ struct ao *ao_init_best(struct mpv_global *global,
                         struct encode_lavc_context *encode_lavc_ctx,
                         int samplerate, int format, struct mp_chmap channels)
 {
+    struct MPOpts *opts = global->opts;
     struct mp_log *log = mp_log_new(NULL, global->log, "ao");
     struct ao *ao = NULL;
-    struct m_obj_settings *ao_list = global->opts->audio_driver_list;
+    struct m_obj_settings *ao_list = opts->audio_driver_list;
     if (ao_list && ao_list[0].name) {
         for (int n = 0; ao_list[n].name; n++) {
             if (strlen(ao_list[n].name) == 0)
@@ -223,15 +298,21 @@ struct ao *ao_init_best(struct mpv_global *global,
         }
         goto done;
     }
-autoprobe:
+autoprobe: ;
     // now try the rest...
+    bool matched_dev = false;
     for (int i = 0; audio_out_drivers[i]; i++) {
+        char *name = (char *)audio_out_drivers[i]->name;
+        if (!match_ao_driver(name, opts->audio_device, NULL))
+            continue;
+        matched_dev = true;
         ao = ao_create(true, global, input_ctx, encode_lavc_ctx,
-                       samplerate, format, channels,
-                       (char *)audio_out_drivers[i]->name, NULL);
+                       samplerate, format, channels, name, NULL);
         if (ao)
             goto done;
     }
+    if (!matched_dev)
+        mp_err(log, "--audio-device option refers to missing output driver.\n");
 done:
     talloc_free(log);
     return ao;
@@ -258,16 +339,7 @@ int ao_play(struct ao *ao, void **data, int samples, int flags)
 
 int ao_control(struct ao *ao, enum aocontrol cmd, void *arg)
 {
-    switch (cmd) {
-    case AOCONTROL_HAS_TEMP_VOLUME:
-        return !ao->no_persistent_volume;
-    case AOCONTROL_HAS_PER_APP_VOLUME:
-        return !!ao->per_application_mixer;
-    default:
-        if (ao->api->control)
-            return ao->api->control(ao, cmd, arg);
-    }
-    return CONTROL_UNKNOWN;
+    return ao->api->control ? ao->api->control(ao, cmd, arg) : CONTROL_UNKNOWN;
 }
 
 // Return size of the buffered data in seconds. Can include the device latency.
@@ -324,6 +396,22 @@ bool ao_eof_reached(struct ao *ao)
     return ao->api->get_eof ? ao->api->get_eof(ao) : true;
 }
 
+// Query the AO_EVENT_*s as requested by the events parameter, and return them.
+int ao_query_and_reset_events(struct ao *ao, int events)
+{
+    int actual_events = 0;
+    if (atomic_load(&ao->request_reload)) // don't need to reset it
+        actual_events |= AO_EVENT_RELOAD;
+    return actual_events & events;
+}
+
+// Request that the player core destroys and recreates the AO.
+void ao_request_reload(struct ao *ao)
+{
+    atomic_store(&ao->request_reload, true);
+    mp_input_wakeup(ao->input_ctx);
+}
+
 bool ao_chmap_sel_adjust(struct ao *ao, const struct mp_chmap_sel *s,
                          struct mp_chmap *map)
 {
@@ -360,3 +448,48 @@ bool ao_untimed(struct ao *ao)
 {
     return ao->untimed;
 }
+
+struct ao_device_list *ao_get_device_list(struct mpv_global *global)
+{
+    struct ao_device_list *list = talloc_zero(NULL, struct ao_device_list);
+    MP_TARRAY_APPEND(list, list->devices, list->num_devices,
+        (struct ao_device_desc){"auto", "Autoselect device"});
+    for (int n = 0; audio_out_drivers[n]; n++) {
+        const struct ao_driver *d = audio_out_drivers[n];
+        if (d == &audio_out_null)
+            break; // don't add unsafe/special entries
+        struct ao *ao = ao_alloc(true, global, NULL, (char *)d->name, NULL);
+        if (!ao)
+            continue;
+        int num = list->num_devices;
+        if (d->list_devs)
+            d->list_devs(ao, list);
+        // Add at least a default entry
+        if (list->num_devices == num)
+            ao_device_list_add(list, ao, &(struct ao_device_desc){"", "Default"});
+        talloc_free(ao);
+    }
+    return list;
+}
+
+void ao_device_list_add(struct ao_device_list *list, struct ao *ao,
+                        struct ao_device_desc *e)
+{
+    struct ao_device_desc c = *e;
+    const char *dname = ao->driver->name;
+    c.name = c.name[0] ? talloc_asprintf(list, "%s/%s", dname, c.name)
+                       : talloc_strdup(list, dname);
+    c.desc = talloc_strdup(list, c.desc);
+    MP_TARRAY_APPEND(list, list->devices, list->num_devices, c);
+}
+
+void ao_print_devices(struct mpv_global *global, struct mp_log *log)
+{
+    struct ao_device_list *list = ao_get_device_list(global);
+    mp_info(log, "List of detected audio devices:\n");
+    for (int n = 0; n < list->num_devices; n++) {
+        struct ao_device_desc *desc = &list->devices[n];
+        mp_info(log, "  '%s' (%s)\n", desc->name, desc->desc);
+    }
+    talloc_free(list);
+}

audio/out/ao.h

@@ -21,7 +21,7 @@
 
 #include <stdbool.h>
 
-#include "bstr/bstr.h"
+#include "misc/bstr.h"
 #include "common/common.h"
 #include "audio/chmap.h"
 #include "audio/chmap_sel.h"
@@ -36,7 +36,9 @@ enum aocontrol {
     AOCONTROL_SET_MUTE,
     // Has char* as argument, which contains the desired stream title.
     AOCONTROL_UPDATE_STREAM_TITLE,
-    AOCONTROL_HAS_TEMP_VOLUME,
+    // the AO does the equivalent of af_volume (return CONTROL_TRUE if yes)
+    AOCONTROL_HAS_SOFT_VOLUME,
+    // like above, but volume persists (per app), mpv won't restore volume
     AOCONTROL_HAS_PER_APP_VOLUME,
 };
 
@@ -44,11 +46,25 @@ enum aocontrol {
 // data might be written again, instead of closing the AO.
 #define AOPLAY_FINAL_CHUNK 1
 
+enum {
+    AO_EVENT_RELOAD = 1,
+};
+
 typedef struct ao_control_vol {
     float left;
     float right;
 } ao_control_vol_t;
 
+struct ao_device_desc {
+    const char *name;   // symbolic name; will be set on ao->device
+    const char *desc;   // verbose human readable name
+};
+
+struct ao_device_list {
+    struct ao_device_desc *devices;
+    int num_devices;
+};
+
 struct ao;
 struct mpv_global;
 struct input_ctx;
@@ -73,5 +89,10 @@ void ao_pause(struct ao *ao);
 void ao_resume(struct ao *ao);
 void ao_drain(struct ao *ao);
 bool ao_eof_reached(struct ao *ao);
+int ao_query_and_reset_events(struct ao *ao, int events);
+void ao_request_reload(struct ao *ao);
+
+struct ao_device_list *ao_get_device_list(struct mpv_global *global);
+void ao_print_devices(struct mpv_global *global, struct mp_log *log);
 
 #endif /* MPLAYER_AUDIO_OUT_H */

audio/out/ao_alsa.c

@@ -38,12 +38,13 @@
 #include "options/options.h"
 #include "options/m_option.h"
 #include "common/msg.h"
-
-#define ALSA_PCM_NEW_HW_PARAMS_API
-#define ALSA_PCM_NEW_SW_PARAMS_API
+#include "osdep/endian.h"
 
 #include <alsa/asoundlib.h>
 
+#define HAVE_CHMAP_API \
+    (defined(SND_CHMAP_API_VERSION) && SND_CHMAP_API_VERSION >= (1 << 16))
+
 #include "ao.h"
 #include "internal.h"
 #include "audio/format.h"
@@ -53,7 +54,7 @@ struct priv {
     snd_pcm_format_t alsa_fmt;
     int can_pause;
     snd_pcm_sframes_t prepause_frames;
-    float delay_before_pause;
+    double delay_before_pause;
     int buffersize; // in frames
     int outburst; // in frames
 
@@ -77,10 +78,12 @@ struct priv {
         } \
     } while (0)
 
-static float get_delay(struct ao *ao);
-static void uninit(struct ao *ao);
+#define CHECK_ALSA_WARN(message) \
+    do { \
+        if (err < 0) \
+            MP_WARN(ao, "%s: %s\n", (message), snd_strerror(err)); \
+    } while (0)
 
-/* to set/get/query special features/parameters */
 static int control(struct ao *ao, enum aocontrol cmd, void *arg)
 {
     struct priv *p = ao->priv;
@@ -99,13 +102,11 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
         long get_vol, set_vol;
         float f_multi;
 
-        if (AF_FORMAT_IS_IEC61937(ao->format))
+        if (AF_FORMAT_IS_SPECIAL(ao->format))
             return CONTROL_FALSE;
 
-        //allocate simple id
         snd_mixer_selem_id_alloca(&sid);
 
-        //sets simple-mixer index and name
         snd_mixer_selem_id_set_index(sid, p->cfg_mixer_index);
         snd_mixer_selem_id_set_name(sid, p->cfg_mixer_name);
 
@@ -137,7 +138,6 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
             ao_control_vol_t *vol = arg;
             set_vol = vol->left / f_multi + pmin + 0.5;
 
-            //setting channels
             err = snd_mixer_selem_set_playback_volume
                     (elem, SND_MIXER_SCHN_FRONT_LEFT, set_vol);
             CHECK_ALSA_ERROR("Error setting left channel");
@@ -149,8 +149,7 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
                     (elem, SND_MIXER_SCHN_FRONT_RIGHT, set_vol);
             CHECK_ALSA_ERROR("Error setting right channel");
             MP_DBG(ao, "right=%li, pmin=%li, pmax=%li, mult=%f\n",
-                   set_vol, pmin, pmax,
-                   f_multi);
+                   set_vol, pmin, pmax, f_multi);
             break;
         }
         case AOCONTROL_GET_VOLUME: {
@@ -208,25 +207,15 @@ alsa_error:
 static const int mp_to_alsa_format[][2] = {
     {AF_FORMAT_S8,          SND_PCM_FORMAT_S8},
     {AF_FORMAT_U8,          SND_PCM_FORMAT_U8},
-    {AF_FORMAT_U16_LE,      SND_PCM_FORMAT_U16_LE},
-    {AF_FORMAT_U16_BE,      SND_PCM_FORMAT_U16_BE},
-    {AF_FORMAT_S16_LE,      SND_PCM_FORMAT_S16_LE},
-    {AF_FORMAT_S16_BE,      SND_PCM_FORMAT_S16_BE},
-    {AF_FORMAT_U32_LE,      SND_PCM_FORMAT_U32_LE},
-    {AF_FORMAT_U32_BE,      SND_PCM_FORMAT_U32_BE},
-    {AF_FORMAT_S32_LE,      SND_PCM_FORMAT_S32_LE},
-    {AF_FORMAT_S32_BE,      SND_PCM_FORMAT_S32_BE},
-    {AF_FORMAT_U24_LE,      SND_PCM_FORMAT_U24_3LE},
-    {AF_FORMAT_U24_BE,      SND_PCM_FORMAT_U24_3BE},
-    {AF_FORMAT_S24_LE,      SND_PCM_FORMAT_S24_3LE},
-    {AF_FORMAT_S24_BE,      SND_PCM_FORMAT_S24_3BE},
-    {AF_FORMAT_FLOAT_LE,    SND_PCM_FORMAT_FLOAT_LE},
-    {AF_FORMAT_FLOAT_BE,    SND_PCM_FORMAT_FLOAT_BE},
-    {AF_FORMAT_AC3_LE,      SND_PCM_FORMAT_S16_LE},
-    {AF_FORMAT_AC3_BE,      SND_PCM_FORMAT_S16_BE},
-    {AF_FORMAT_IEC61937_LE, SND_PCM_FORMAT_S16_LE},
-    {AF_FORMAT_IEC61937_BE, SND_PCM_FORMAT_S16_BE},
-    {AF_FORMAT_MPEG2,       SND_PCM_FORMAT_MPEG},
+    {AF_FORMAT_U16,         SND_PCM_FORMAT_U16},
+    {AF_FORMAT_S16,         SND_PCM_FORMAT_S16},
+    {AF_FORMAT_U32,         SND_PCM_FORMAT_U32},
+    {AF_FORMAT_S32,         SND_PCM_FORMAT_S32},
+    {AF_FORMAT_U24,
+            MP_SELECT_LE_BE(SND_PCM_FORMAT_U24_3LE, SND_PCM_FORMAT_U24_3BE)},
+    {AF_FORMAT_S24,
+            MP_SELECT_LE_BE(SND_PCM_FORMAT_S24_3LE, SND_PCM_FORMAT_S24_3BE)},
+    {AF_FORMAT_FLOAT,       SND_PCM_FORMAT_FLOAT},
     {AF_FORMAT_UNKNOWN,     SND_PCM_FORMAT_UNKNOWN},
 };
 
@@ -240,6 +229,98 @@ static int find_alsa_format(int af_format)
     return SND_PCM_FORMAT_UNKNOWN;
 }
 
+#if HAVE_CHMAP_API
+
+static const int alsa_to_mp_channels[][2] = {
+    {SND_CHMAP_FL,      MP_SP(FL)},
+    {SND_CHMAP_FR,      MP_SP(FR)},
+    {SND_CHMAP_RL,      MP_SP(BL)},
+    {SND_CHMAP_RR,      MP_SP(BR)},
+    {SND_CHMAP_FC,      MP_SP(FC)},
+    {SND_CHMAP_LFE,     MP_SP(LFE)},
+    {SND_CHMAP_SL,      MP_SP(SL)},
+    {SND_CHMAP_SR,      MP_SP(SR)},
+    {SND_CHMAP_RC,      MP_SP(BC)},
+    {SND_CHMAP_FLC,     MP_SP(FLC)},
+    {SND_CHMAP_FRC,     MP_SP(FRC)},
+    {SND_CHMAP_FLW,     MP_SP(WL)},
+    {SND_CHMAP_FRW,     MP_SP(WR)},
+    {SND_CHMAP_TC,      MP_SP(TC)},
+    {SND_CHMAP_TFL,     MP_SP(TFL)},
+    {SND_CHMAP_TFR,     MP_SP(TFR)},
+    {SND_CHMAP_TFC,     MP_SP(TFC)},
+    {SND_CHMAP_TRL,     MP_SP(TBL)},
+    {SND_CHMAP_TRR,     MP_SP(TBR)},
+    {SND_CHMAP_TRC,     MP_SP(TBC)},
+    {SND_CHMAP_MONO,    MP_SP(FC)},
+    {SND_CHMAP_LAST,    MP_SPEAKER_ID_COUNT}
+};
+
+static int find_mp_channel(int alsa_channel)
+{
+    for (int i = 0; alsa_to_mp_channels[i][1] != MP_SPEAKER_ID_COUNT; i++) {
+        if (alsa_to_mp_channels[i][0] == alsa_channel)
+            return alsa_to_mp_channels[i][1];
+    }
+
+    return MP_SPEAKER_ID_COUNT;
+}
+
+static int find_alsa_channel(int mp_channel)
+{
+    for (int i = 0; alsa_to_mp_channels[i][1] != MP_SPEAKER_ID_COUNT; i++) {
+        if (alsa_to_mp_channels[i][1] == mp_channel)
+            return alsa_to_mp_channels[i][0];
+    }
+
+    return SND_CHMAP_UNKNOWN;
+}
+
+static bool query_chmaps(struct ao *ao, struct mp_chmap *chmap)
+{
+    struct priv *p = ao->priv;
+    struct mp_chmap_sel chmap_sel = {0};
+
+    snd_pcm_chmap_query_t **maps = snd_pcm_query_chmaps(p->alsa);
+    if (!maps)
+        return false;
+
+    for (int i = 0; maps[i] != NULL; i++) {
+        if (maps[i]->map.channels > MP_NUM_CHANNELS) {
+            MP_VERBOSE(ao, "skipping ALSA channel map with too many channels.\n");
+            continue;
+        }
+
+        struct mp_chmap entry = {.num = maps[i]->map.channels};
+        for (int c = 0; c < entry.num; c++)
+            entry.speaker[c] = find_mp_channel(maps[i]->map.pos[c]);
+
+        if (mp_chmap_is_valid(&entry)) {
+            MP_VERBOSE(ao, "Got supported channel map: %s (type %s)\n",
+                       mp_chmap_to_str(&entry),
+                       snd_pcm_chmap_type_name(maps[i]->type));
+            mp_chmap_sel_add_map(&chmap_sel, &entry);
+        } else {
+            char tmp[128];
+            if (snd_pcm_chmap_print(&maps[i]->map, sizeof(tmp), tmp) > 0)
+                MP_VERBOSE(ao, "skipping unknown ALSA channel map: %s\n", tmp);
+        }
+    }
+
+    snd_pcm_free_chmaps(maps);
+
+    return ao_chmap_sel_adjust(ao, &chmap_sel, chmap);
+}
+
+#else /* HAVE_CHMAP_API */
+
+static bool query_chmaps(struct ao *ao, struct mp_chmap *chmap)
+{
+    return false;
+}
+
+#endif /* else HAVE_CHMAP_API */
+
 // Lists device names and their implied channel map.
 // The second item must be resolvable with mp_chmap_from_str().
 // Source: http://www.alsa-project.org/main/index.php/DeviceNames
@@ -257,11 +338,9 @@ static const char *const device_channel_layouts[][2] = {
     {"surround71",      "fl-fr-bl-br-fc-lfe-sl-sr"},
 };
 
-#define ARRAY_LEN(x) (sizeof(x) / sizeof((x)[0]))
+#define NUM_ALSA_CHMAPS MP_ARRAY_SIZE(device_channel_layouts)
 
-#define NUM_ALSA_CHMAPS ARRAY_LEN(device_channel_layouts)
-
-static const char *select_chmap(struct ao *ao)
+static const char *select_chmap(struct ao *ao, struct mp_chmap *chmap)
 {
     struct mp_chmap_sel sel = {0};
     struct mp_chmap maps[NUM_ALSA_CHMAPS];
@@ -270,18 +349,16 @@ static const char *select_chmap(struct ao *ao)
         mp_chmap_sel_add_map(&sel, &maps[n]);
     };
 
-    if (!ao_chmap_sel_adjust(ao, &sel, &ao->channels))
-        return NULL;
+    if (!ao_chmap_sel_adjust(ao, &sel, chmap))
+        return "default";
 
     for (int n = 0; n < NUM_ALSA_CHMAPS; n++) {
-        if (mp_chmap_equals(&ao->channels, &maps[n]))
+        if (mp_chmap_equals(chmap, &maps[n]))
             return device_channel_layouts[n][0];
     }
 
-    char *name = mp_chmap_to_str(&ao->channels);
     MP_ERR(ao, "channel layout %s (%d ch) not supported.\n",
-           name, ao->channels.num);
-    talloc_free(name);
+           mp_chmap_to_str(chmap), chmap->num);
     return "default";
 }
 
@@ -302,34 +379,44 @@ static int map_iec958_srate(int srate)
     }
 }
 
+// ALSA device strings can have parameters. They are usually appended to the
+// device name. Since there can be various forms, and we (sometimes) want to
+// append them to unknown device strings, which possibly already include params.
+static char *append_params(void *ta_parent, const char *device, const char *p)
+{
+    if (!p || !p[0])
+        return talloc_strdup(ta_parent, device);
+
+    int len = strlen(device);
+    char *end = strchr(device, ':');
+    if (!end) {
+        /* no existing parameters: add it behind device name */
+        return talloc_asprintf(ta_parent, "%s:%s", device, p);
+    } else if (end[1] == '\0') {
+        /* ":" but no parameters */
+        return talloc_asprintf(ta_parent, "%s%s", device, p);
+    } else if (end[1] == '{' && device[len - 1] == '}') {
+        /* parameters in config syntax: add it inside the { } block */
+        return talloc_asprintf(ta_parent, "%.*s %s}", len - 1, device, p);
+    } else {
+        /* a simple list of parameters: add it at the end of the list */
+        return talloc_asprintf(ta_parent, "%s,%s", device, p);
+    }
+    abort();
+}
+
 static int try_open_device(struct ao *ao, const char *device, int open_mode)
 {
     struct priv *p = ao->priv;
 
     if (AF_FORMAT_IS_IEC61937(ao->format)) {
         void *tmp = talloc_new(NULL);
-        /* to set the non-audio bit, use AES0=6 */
         char *params = talloc_asprintf(tmp,
                         "AES0=%d,AES1=%d,AES2=0,AES3=%d",
                         IEC958_AES0_NONAUDIO | IEC958_AES0_PRO_EMPHASIS_NONE,
                         IEC958_AES1_CON_ORIGINAL | IEC958_AES1_CON_PCM_CODER,
                         map_iec958_srate(ao->samplerate));
-        const char *ac3_device = device;
-        int len = strlen(device);
-        char *end = strchr(device, ':');
-        if (!end) {
-            /* no existing parameters: add it behind device name */
-            ac3_device = talloc_asprintf(tmp, "%s:%s", device, params);
-        } else if (end[1] == '\0') {
-            /* ":" but no parameters */
-            ac3_device = talloc_asprintf(tmp, "%s%s", device, params);
-        } else if (end[1] == '{' && device[len - 1] == '}') {
-            /* parameters in config syntax: add it inside the { } block */
-            ac3_device = talloc_asprintf(tmp, "%.*s %s}", len - 1, device, params);
-        } else {
-            /* a simple list of parameters: add it at the end of the list */
-            ac3_device = talloc_asprintf(tmp, "%s,%s", device, params);
-        }
+        const char *ac3_device = append_params(tmp, device, params);
         int err = snd_pcm_open
                     (&p->alsa, ac3_device, SND_PCM_STREAM_PLAYBACK, open_mode);
         talloc_free(tmp);
@@ -340,49 +427,49 @@ static int try_open_device(struct ao *ao, const char *device, int open_mode)
     return snd_pcm_open(&p->alsa, device, SND_PCM_STREAM_PLAYBACK, open_mode);
 }
 
-/*
-    open & setup audio device
-    return: 0=success -1=fail
- */
+static void uninit(struct ao *ao)
+{
+    struct priv *p = ao->priv;
+
+    if (p->alsa) {
+        int err;
+
+        err = snd_pcm_close(p->alsa);
+        CHECK_ALSA_ERROR("pcm close error");
+    }
+
+alsa_error: ;
+}
+
 static int init(struct ao *ao)
 {
-    int err;
-    snd_pcm_uframes_t chunk_size;
-    snd_pcm_uframes_t bufsize;
-    snd_pcm_uframes_t boundary;
-
     struct priv *p = ao->priv;
+    int err;
 
     if (!p->cfg_ni)
         ao->format = af_fmt_from_planar(ao->format);
 
-    /* switch for spdif
-     * sets opening sequence for SPDIF
-     * sets also the playback and other switches 'on the fly'
-     * while opening the abstract alias for the spdif subdevice
-     * 'iec958'
-     */
+    struct mp_chmap implied_chmap = ao->channels;
     const char *device;
     if (AF_FORMAT_IS_IEC61937(ao->format)) {
         device = "iec958";
         MP_VERBOSE(ao, "playing AC3/iec61937/iec958, %i channels\n",
                    ao->channels.num);
     } else {
-        device = select_chmap(ao);
-        if (strcmp(device, "default") != 0 && (ao->format & AF_FORMAT_F)) {
-            // hack - use the converter plugin (why the heck?)
+        device = select_chmap(ao, &implied_chmap);
+        // Not-default likely means a hw device - enable software conversions.
+        if (strcmp(device, "default") != 0)
             device = talloc_asprintf(ao, "plug:%s", device);
-        }
     }
+    if (ao->device)
+        device = ao->device;
     if (p->cfg_device && p->cfg_device[0])
         device = p->cfg_device;
 
     MP_VERBOSE(ao, "using device: %s\n", device);
     MP_VERBOSE(ao, "using ALSA version: %s\n", snd_asoundlib_version());
 
-    int open_mode = p->cfg_block ? 0 : SND_PCM_NONBLOCK;
-    //modes = 0, SND_PCM_NONBLOCK, SND_PCM_ASYNC
-    err = try_open_device(ao, device, open_mode);
+    err = try_open_device(ao, device, p->cfg_block ? 0 : SND_PCM_NONBLOCK);
     if (err < 0) {
         if (err != -EBUSY && !p->cfg_block) {
             MP_WARN(ao, "Open in nonblock-mode "
@@ -405,11 +492,18 @@ static int init(struct ao *ao)
     snd_pcm_hw_params_alloca(&alsa_hwparams);
     snd_pcm_sw_params_alloca(&alsa_swparams);
 
-    // setting hw-parameters
     err = snd_pcm_hw_params_any(p->alsa, alsa_hwparams);
     CHECK_ALSA_ERROR("Unable to get initial parameters");
 
-    p->alsa_fmt = find_alsa_format(ao->format);
+    if (AF_FORMAT_IS_IEC61937(ao->format)) {
+        if (ao->format == AF_FORMAT_S_MP3) {
+            p->alsa_fmt = SND_PCM_FORMAT_MPEG;
+        } else {
+            p->alsa_fmt = SND_PCM_FORMAT_S16;
+        }
+    } else {
+        p->alsa_fmt = find_alsa_format(ao->format);
+    }
     if (p->alsa_fmt == SND_PCM_FORMAT_UNKNOWN) {
         p->alsa_fmt = SND_PCM_FORMAT_S16;
         ao->format = AF_FORMAT_S16;
@@ -417,15 +511,12 @@ static int init(struct ao *ao)
 
     err = snd_pcm_hw_params_test_format(p->alsa, alsa_hwparams, p->alsa_fmt);
     if (err < 0) {
+        if (AF_FORMAT_IS_IEC61937(ao->format))
+            CHECK_ALSA_ERROR("Unable to set IEC61937 format");
         MP_INFO(ao, "Format %s is not supported by hardware, trying default.\n",
                 af_fmt_to_str(ao->format));
-        p->alsa_fmt = SND_PCM_FORMAT_S16_LE;
-        if (AF_FORMAT_IS_AC3(ao->format))
-            ao->format = AF_FORMAT_AC3_LE;
-        else if (AF_FORMAT_IS_IEC61937(ao->format))
-            ao->format = AF_FORMAT_IEC61937_LE;
-        else
-            ao->format = AF_FORMAT_S16_LE;
+        p->alsa_fmt = SND_PCM_FORMAT_S16;
+        ao->format = AF_FORMAT_S16;
     }
 
     err = snd_pcm_hw_params_set_format(p->alsa, alsa_hwparams, p->alsa_fmt);
@@ -442,11 +533,24 @@ static int init(struct ao *ao)
     }
     CHECK_ALSA_ERROR("Unable to set access type");
 
+    struct mp_chmap dev_chmap = ao->channels;
+    if (query_chmaps(ao, &dev_chmap)) {
+        ao->channels = dev_chmap;
+    } else {
+        dev_chmap.num = 0;
+        ao->channels = implied_chmap;
+    }
+
     int num_channels = ao->channels.num;
     err = snd_pcm_hw_params_set_channels_near
             (p->alsa, alsa_hwparams, &num_channels);
     CHECK_ALSA_ERROR("Unable to set channels");
 
+    if (num_channels > MP_NUM_CHANNELS) {
+        MP_FATAL(ao, "Too many audio channels (%d).\n", num_channels);
+        goto alsa_error;
+    }
+
     if (num_channels != ao->channels.num) {
         MP_ERR(ao, "Couldn't get requested number of channels.\n");
         mp_chmap_from_channels_alsa(&ao->channels, num_channels);
@@ -465,25 +569,79 @@ static int init(struct ao *ao)
 
     err = snd_pcm_hw_params_set_buffer_time_near
             (p->alsa, alsa_hwparams, &(unsigned int){BUFFER_TIME}, NULL);
-    CHECK_ALSA_ERROR("Unable to set buffer time near");
+    CHECK_ALSA_WARN("Unable to set buffer time near");
 
     err = snd_pcm_hw_params_set_periods_near
             (p->alsa, alsa_hwparams, &(unsigned int){FRAGCOUNT}, NULL);
-    CHECK_ALSA_ERROR("Unable to set periods");
+    CHECK_ALSA_WARN("Unable to set periods");
 
     /* finally install hardware parameters */
     err = snd_pcm_hw_params(p->alsa, alsa_hwparams);
     CHECK_ALSA_ERROR("Unable to set hw-parameters");
 
-    // end setting hw-params
+    /* end setting hw-params */
 
-    // gets buffersize for control
+#if HAVE_CHMAP_API
+    if (mp_chmap_is_valid(&dev_chmap)) {
+        snd_pcm_chmap_t *alsa_chmap =
+            calloc(1, sizeof(*alsa_chmap) +
+                      sizeof(alsa_chmap->pos[0]) * dev_chmap.num);
+        if (!alsa_chmap)
+            goto alsa_error;
+
+        alsa_chmap->channels = dev_chmap.num;
+        for (int c = 0; c < dev_chmap.num; c++)
+            alsa_chmap->pos[c] = find_alsa_channel(dev_chmap.speaker[c]);
+
+        char tmp[128];
+        if (snd_pcm_chmap_print(alsa_chmap, sizeof(tmp), tmp) > 0)
+            MP_VERBOSE(ao, "trying to set ALSA channel map: %s\n", tmp);
+
+        err = snd_pcm_set_chmap(p->alsa, alsa_chmap);
+        if (err == -ENXIO) {
+            MP_WARN(ao, "Device does not support requested channel map\n");
+        } else {
+            CHECK_ALSA_WARN("Channel map setup failed");
+        }
+    }
+
+    snd_pcm_chmap_t *alsa_chmap = snd_pcm_get_chmap(p->alsa);
+    if (alsa_chmap) {
+        char tmp[128];
+        if (snd_pcm_chmap_print(alsa_chmap, sizeof(tmp), tmp) > 0)
+            MP_VERBOSE(ao, "channel map reported by ALSA: %s\n", tmp);
+
+        struct mp_chmap chmap = {.num = alsa_chmap->channels};
+        for (int c = 0; c < chmap.num; c++)
+            chmap.speaker[c] = find_mp_channel(alsa_chmap->pos[c]);
+
+        MP_VERBOSE(ao, "which we understand as: %s\n", mp_chmap_to_str(&chmap));
+
+        if (mp_chmap_is_valid(&chmap)) {
+            if (mp_chmap_equals(&chmap, &ao->channels)) {
+                MP_VERBOSE(ao, "which is what we requested.\n");
+            } else if (chmap.num == ao->channels.num) {
+                MP_VERBOSE(ao, "using the ALSA channel map.\n");
+                ao->channels = chmap;
+            } else {
+                MP_WARN(ao, "ALSA channel map conflicts with channel count!\n");
+            }
+        } else {
+            MP_WARN(ao, "Got unknown channel map from ALSA.\n");
+        }
+
+        free(alsa_chmap);
+    }
+#endif
+
+    snd_pcm_uframes_t bufsize;
     err = snd_pcm_hw_params_get_buffer_size(alsa_hwparams, &bufsize);
     CHECK_ALSA_ERROR("Unable to get buffersize");
 
     p->buffersize = bufsize;
     MP_VERBOSE(ao, "got buffersize=%i samples\n", p->buffersize);
 
+    snd_pcm_uframes_t chunk_size;
     err = snd_pcm_hw_params_get_period_size(alsa_hwparams, &chunk_size, NULL);
     CHECK_ALSA_ERROR("Unable to get period size");
 
@@ -494,6 +652,7 @@ static int init(struct ao *ao)
     err = snd_pcm_sw_params_current(p->alsa, alsa_swparams);
     CHECK_ALSA_ERROR("Unable to get sw-parameters");
 
+    snd_pcm_uframes_t boundary;
     err = snd_pcm_sw_params_get_boundary(alsa_swparams, &boundary);
     CHECK_ALSA_ERROR("Unable to get boundary");
 
@@ -513,40 +672,17 @@ static int init(struct ao *ao)
     CHECK_ALSA_ERROR("Unable to set silence size");
 
     err = snd_pcm_sw_params(p->alsa, alsa_swparams);
-    CHECK_ALSA_ERROR("Unable to get sw-parameters");
+    CHECK_ALSA_ERROR("Unable to set sw-parameters");
 
     /* end setting sw-params */
 
     p->can_pause = snd_pcm_hw_params_can_pause(alsa_hwparams);
 
-    MP_VERBOSE(ao, "opened: %d Hz/%d channels/%d bps/%d samples buffer/%s\n",
-               ao->samplerate, ao->channels.num, af_fmt2bits(ao->format),
-               p->buffersize, snd_pcm_format_description(p->alsa_fmt));
-
     return 0;
 
 alsa_error:
     uninit(ao);
     return -1;
-} // end init
-
-
-/* close audio device */
-static void uninit(struct ao *ao)
-{
-    struct priv *p = ao->priv;
-
-    if (p->alsa) {
-        int err;
-
-        err = snd_pcm_close(p->alsa);
-        CHECK_ALSA_ERROR("pcm close error");
-
-        MP_VERBOSE(ao, "uninit: pcm closed\n");
-    }
-
-alsa_error:
-    p->alsa = NULL;
 }
 
 static void drain(struct ao *ao)
@@ -555,6 +691,46 @@ static void drain(struct ao *ao)
     snd_pcm_drain(p->alsa);
 }
 
+static int get_space(struct ao *ao)
+{
+    struct priv *p = ao->priv;
+    snd_pcm_status_t *status;
+    int err;
+
+    snd_pcm_status_alloca(&status);
+
+    err = snd_pcm_status(p->alsa, status);
+    CHECK_ALSA_ERROR("cannot get pcm status");
+
+    unsigned space = snd_pcm_status_get_avail(status);
+    if (space > p->buffersize) // Buffer underrun?
+        space = p->buffersize;
+    return space / p->outburst * p->outburst;
+
+alsa_error:
+    return 0;
+}
+
+/* delay in seconds between first and last sample in buffer */
+static double get_delay(struct ao *ao)
+{
+    struct priv *p = ao->priv;
+    snd_pcm_sframes_t delay;
+
+    if (snd_pcm_state(p->alsa) == SND_PCM_STATE_PAUSED)
+        return p->delay_before_pause;
+
+    if (snd_pcm_delay(p->alsa, &delay) < 0)
+        return 0;
+
+    if (delay < 0) {
+        /* underrun - move the application pointer forward to catch up */
+        snd_pcm_forward(p->alsa, -delay);
+        delay = 0;
+    }
+    return delay / (double)ao->samplerate;
+}
+
 static void audio_pause(struct ao *ao)
 {
     struct priv *p = ao->priv;
@@ -571,7 +747,7 @@ static void audio_pause(struct ao *ao)
         if (snd_pcm_delay(p->alsa, &p->prepause_frames) < 0
             || p->prepause_frames < 0)
             p->prepause_frames = 0;
-        p->delay_before_pause = p->prepause_frames / (float)ao->samplerate;
+        p->delay_before_pause = p->prepause_frames / (double)ao->samplerate;
 
         err = snd_pcm_drop(p->alsa);
         CHECK_ALSA_ERROR("pcm drop error");
@@ -608,7 +784,6 @@ static void audio_resume(struct ao *ao)
 alsa_error: ;
 }
 
-/* stop playing and empty buffers (for seeking/pause) */
 static void reset(struct ao *ao)
 {
     struct priv *p = ao->priv;
@@ -641,8 +816,7 @@ static int play(struct ao *ao, void **data, int samples, int flags)
             res = snd_pcm_writei(p->alsa, data[0], samples);
         }
 
-        if (res == -EINTR) {
-            /* nothing to do */
+        if (res == -EINTR || res == -EAGAIN) { /* retry */
             res = 0;
         } else if (res == -ESTRPIPE) {  /* suspend */
             audio_resume(ao);
@@ -661,46 +835,6 @@ alsa_error:
     return -1;
 }
 
-static int get_space(struct ao *ao)
-{
-    struct priv *p = ao->priv;
-    snd_pcm_status_t *status;
-    int err;
-
-    snd_pcm_status_alloca(&status);
-
-    err = snd_pcm_status(p->alsa, status);
-    CHECK_ALSA_ERROR("cannot get pcm status");
-
-    unsigned space = snd_pcm_status_get_avail(status);
-    if (space > p->buffersize) // Buffer underrun?
-        space = p->buffersize;
-    return space;
-
-alsa_error:
-    return 0;
-}
-
-/* delay in seconds between first and last sample in buffer */
-static float get_delay(struct ao *ao)
-{
-    struct priv *p = ao->priv;
-    snd_pcm_sframes_t delay;
-
-    if (snd_pcm_state(p->alsa) == SND_PCM_STATE_PAUSED)
-        return p->delay_before_pause;
-
-    if (snd_pcm_delay(p->alsa, &delay) < 0)
-        return 0;
-
-    if (delay < 0) {
-        /* underrun - move the application pointer forward to catch up */
-        snd_pcm_forward(p->alsa, -delay);
-        delay = 0;
-    }
-    return (float)delay / (float)ao->samplerate;
-}
-
 #define MAX_POLL_FDS 20
 static int audio_wait(struct ao *ao, pthread_mutex_t *lock)
 {
@@ -735,10 +869,34 @@ alsa_error:
     return -1;
 }
 
+static void list_devs(struct ao *ao, struct ao_device_list *list)
+{
+    void **hints;
+    if (snd_device_name_hint(-1, "pcm", &hints) < 0)
+        return;
+
+    for (int n = 0; hints[n]; n++) {
+        char *name = snd_device_name_get_hint(hints[n], "NAME");
+        char *desc = snd_device_name_get_hint(hints[n], "DESC");
+        char *io = snd_device_name_get_hint(hints[n], "IOID");
+        if (io && strcmp(io, "Output") != 0)
+            continue;
+        char desc2[1024];
+        snprintf(desc2, sizeof(desc2), "%s", desc ? desc : "");
+        for (int i = 0; desc2[i]; i++) {
+            if (desc2[i] == '\n')
+                desc2[i] = '/';
+        }
+        ao_device_list_add(list, ao, &(struct ao_device_desc){name, desc2});
+    }
+
+    snd_device_name_free_hint(hints);
+}
+
 #define OPT_BASE_STRUCT struct priv
 
 const struct ao_driver audio_out_alsa = {
-    .description = "ALSA-0.9.x-1.x audio output",
+    .description = "ALSA audio output",
     .name      = "alsa",
     .init      = init,
     .uninit    = uninit,
@@ -752,6 +910,7 @@ const struct ao_driver audio_out_alsa = {
     .drain     = drain,
     .wait      = audio_wait,
     .wakeup    = ao_wakeup_poll,
+    .list_devs = list_devs,
     .priv_size = sizeof(struct priv),
     .priv_defaults = &(const struct priv) {
         .cfg_block = 1,

audio/out/ao_coreaudio.c

@@ -33,10 +33,6 @@ struct priv {
     AudioUnit audio_unit;
 
     uint64_t hw_latency_us;
-
-    // options
-    int opt_device_id;
-    int opt_list;
 };
 
 bool ca_layout_to_mp_chmap(struct ao *ao, AudioChannelLayout *layout,
@@ -136,6 +132,10 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
         return get_volume(ao, arg);
     case AOCONTROL_SET_VOLUME:
         return set_volume(ao, arg);
+    case AOCONTROL_HAS_SOFT_VOLUME:
+        return CONTROL_TRUE;
+    case AOCONTROL_HAS_PER_APP_VOLUME:
+        return CONTROL_TRUE;
     }
     return CONTROL_UNKNOWN;
 }
@@ -145,14 +145,15 @@ static bool init_audiounit(struct ao *ao, AudioStreamBasicDescription asbd);
 
 static int init(struct ao *ao)
 {
+    if (AF_FORMAT_IS_IEC61937(ao->format)) {
+        MP_WARN(ao, "detected IEC61937, redirecting to coreaudio_exclusive\n");
+        ao->redirect = "coreaudio_exclusive";
+        return CONTROL_ERROR;
+    }
+
     struct priv *p = ao->priv;
 
-    if (p->opt_list) ca_print_device_list(ao);
-
-    ao->per_application_mixer = true;
-    ao->no_persistent_volume  = true;
-
-    OSStatus err = ca_select_device(ao, p->opt_device_id, &p->device);
+    OSStatus err = ca_select_device(ao, ao->device, &p->device);
     CHECK_CA_ERROR("failed to select device");
 
     if (!init_chmap(ao))
@@ -220,9 +221,9 @@ static bool init_audiounit(struct ao *ao, AudioStreamBasicDescription asbd)
 
     AudioComponentDescription desc = (AudioComponentDescription) {
         .componentType         = kAudioUnitType_Output,
-        .componentSubType      = (p->opt_device_id < 0) ?
-                                    kAudioUnitSubType_DefaultOutput :
-                                    kAudioUnitSubType_HALOutput,
+        .componentSubType      = (ao->device) ?
+                                    kAudioUnitSubType_HALOutput :
+                                    kAudioUnitSubType_DefaultOutput,
         .componentManufacturer = kAudioUnitManufacturer_Apple,
         .componentFlags        = 0,
         .componentFlagsMask    = 0,
@@ -439,10 +440,6 @@ const struct ao_driver audio_out_coreaudio = {
     .control   = control,
     .pause     = stop,
     .resume    = start,
+    .list_devs = ca_get_device_list,
     .priv_size = sizeof(struct priv),
-    .options = (const struct m_option[]) {
-        OPT_INT("device_id", opt_device_id, 0, OPTDEF_INT(-1)),
-        OPT_FLAG("list", opt_list, 0),
-        {0}
-    },
 };

audio/out/ao_coreaudio_exclusive.c

@@ -333,10 +333,6 @@ struct priv {
     bool changed_mixing;
     int stream_asbd_changed;
     bool muted;
-
-    // options
-    int opt_device_id;
-    int opt_list;
 };
 
 static int get_ring_size(struct ao *ao)
@@ -402,16 +398,15 @@ static int init(struct ao *ao)
 {
     struct priv *p = ao->priv;
 
-    if (p->opt_list) ca_print_device_list(ao);
-
-    OSStatus err = ca_select_device(ao, p->opt_device_id, &p->device);
+    OSStatus err = ca_select_device(ao, ao->device, &p->device);
     CHECK_CA_ERROR("failed to select device");
 
     ao->format = af_fmt_from_planar(ao->format);
 
     bool supports_digital = false;
-    /* Probe whether device support S/PDIF stream output if input is AC3 stream. */
-    if (AF_FORMAT_IS_AC3(ao->format)) {
+    /* Probe whether device support S/PDIF stream output if input is AC3 stream,
+     * or anything else IEC61937-framed. */
+    if (AF_FORMAT_IS_IEC61937(ao->format)) {
         if (ca_device_supports_digital(ao, p->device))
             supports_digital = true;
     }
@@ -521,14 +516,7 @@ static int init_digital(struct ao *ao, AudioStreamBasicDescription asbd)
     err = ca_enable_device_listener(p->device, changed);
     CHECK_CA_ERROR("cannot install format change listener during init");
 
-#if BYTE_ORDER == BIG_ENDIAN
-    if (!(p->stream_asdb.mFormatFlags & kAudioFormatFlagIsBigEndian))
-#else
-    /* tell mplayer that we need a byteswap on AC3 streams, */
-    if (p->stream_asbd.mFormatID & kAudioFormat60958AC3)
-        ao->format = AF_FORMAT_AC3_LE;
-    else if (p->stream_asbd.mFormatFlags & kAudioFormatFlagIsBigEndian)
-#endif
+    if (p->stream_asbd.mFormatFlags & kAudioFormatFlagIsBigEndian)
         MP_WARN(ao, "stream has non-native byte order, output may fail\n");
 
     ao->samplerate = p->stream_asbd.mSampleRate;
@@ -593,12 +581,12 @@ static int get_space(struct ao *ao)
     return mp_ring_available(p->buffer) / ao->sstride;
 }
 
-static float get_delay(struct ao *ao)
+static double get_delay(struct ao *ao)
 {
     // FIXME: should also report the delay of coreaudio itself (hardware +
     // internal buffers)
     struct priv *p = ao->priv;
-    return mp_ring_buffered(p->buffer) / (float)ao->bps;
+    return mp_ring_buffered(p->buffer) / (double)ao->bps;
 }
 
 static void uninit(struct ao *ao)
@@ -666,6 +654,7 @@ const struct ao_driver audio_out_coreaudio_exclusive = {
     .reset     = reset,
     .pause     = audio_pause,
     .resume    = audio_resume,
+    .list_devs = ca_get_device_list,
     .priv_size = sizeof(struct priv),
     .priv_defaults = &(const struct priv){
         .muted = false,
@@ -675,9 +664,4 @@ const struct ao_driver audio_out_coreaudio_exclusive = {
         .stream_idx = -1,
         .changed_mixing = false,
     },
-    .options = (const struct m_option[]) {
-        OPT_INT("device_id", opt_device_id, 0, OPTDEF_INT(-1)),
-        OPT_FLAG("list", opt_list, 0),
-        {0}
-    },
 };

audio/out/ao_coreaudio_utils.c

@@ -25,44 +25,51 @@
 #include "audio/out/ao_coreaudio_utils.h"
 #include "audio/out/ao_coreaudio_properties.h"
 #include "osdep/timer.h"
+#include "osdep/endian.h"
 #include "audio/format.h"
 
-void ca_print_device_list(struct ao *ao)
+static bool ca_is_output_device(struct ao *ao, AudioDeviceID dev)
 {
-    char *help = talloc_strdup(NULL, "Available output devices:\n");
+    size_t n_buffers;
+    AudioBufferList *buffers;
+    const ca_scope scope = kAudioDevicePropertyStreamConfiguration;
+    CA_GET_ARY_O(dev, scope, &buffers, &n_buffers);
+    talloc_free(buffers);
+    return n_buffers > 0;
+}
 
+void ca_get_device_list(struct ao *ao, struct ao_device_list *list)
+{
     AudioDeviceID *devs;
     size_t n_devs;
-
     OSStatus err =
         CA_GET_ARY(kAudioObjectSystemObject, kAudioHardwarePropertyDevices,
                    &devs, &n_devs);
-
     CHECK_CA_ERROR("Failed to get list of output devices.");
-
     for (int i = 0; i < n_devs; i++) {
-        char *name;
-        err = CA_GET_STR(devs[i], kAudioObjectPropertyName, &name);
-
-        if (err == noErr)
-            talloc_steal(devs, name);
-        else
-            name = "Unknown";
-
-        help = talloc_asprintf_append(
-                help, "  * %s (id: %" PRIu32 ")\n", name, devs[i]);
+        if (!ca_is_output_device(ao, devs[i]))
+            continue;
+        char name[32];
+        char *desc;
+        sprintf(name, "%d", devs[i]);
+        err = CA_GET_STR(devs[i], kAudioObjectPropertyName, &desc);
+        if (err != noErr)
+            desc = "Unknown";
+        ao_device_list_add(list, ao, &(struct ao_device_desc){name, desc});
     }
-
     talloc_free(devs);
-
 coreaudio_error:
-    MP_INFO(ao, "%s", help);
-    talloc_free(help);
+    return;
 }
 
-OSStatus ca_select_device(struct ao *ao, int selection, AudioDeviceID *device)
+OSStatus ca_select_device(struct ao *ao, char* name, AudioDeviceID *device)
 {
     OSStatus err = noErr;
+    int selection = name ? strtol(name, (char **)NULL, 10) : -1;
+    if (errno == EINVAL || errno == ERANGE) {
+        selection = -1;
+        MP_WARN(ao, "device identifier '%s' is invalid\n", name);
+    }
     *device = 0;
     if (selection < 0) {
         // device not set by user, get the default one
@@ -75,14 +82,14 @@ OSStatus ca_select_device(struct ao *ao, int selection, AudioDeviceID *device)
     }
 
     if (mp_msg_test(ao->log, MSGL_V)) {
-        char *name;
-        err = CA_GET_STR(*device, kAudioObjectPropertyName, &name);
+        char *desc;
+        err = CA_GET_STR(*device, kAudioObjectPropertyName, &desc);
         CHECK_CA_ERROR("could not get selected audio device name");
 
         MP_VERBOSE(ao, "selected audio output device: %s (%" PRIu32 ")\n",
-                       name, *device);
+                       desc, *device);
 
-        talloc_free(name);
+        talloc_free(desc);
     }
 
 coreaudio_error:
@@ -129,20 +136,21 @@ bool check_ca_st(struct ao *ao, int level, OSStatus code, const char *message)
 void ca_fill_asbd(struct ao *ao, AudioStreamBasicDescription *asbd)
 {
     asbd->mSampleRate       = ao->samplerate;
-    asbd->mFormatID         = AF_FORMAT_IS_AC3(ao->format) ?
+    // Set "AC3" for other spdif formats too - unknown if that works.
+    asbd->mFormatID         = AF_FORMAT_IS_IEC61937(ao->format) ?
                               kAudioFormat60958AC3 :
                               kAudioFormatLinearPCM;
     asbd->mChannelsPerFrame = ao->channels.num;
     asbd->mBitsPerChannel   = af_fmt2bits(ao->format);
     asbd->mFormatFlags      = kAudioFormatFlagIsPacked;
 
-    if ((ao->format & AF_FORMAT_POINT_MASK) == AF_FORMAT_F)
+    if ((ao->format & AF_FORMAT_TYPE_MASK) == AF_FORMAT_F)
         asbd->mFormatFlags |= kAudioFormatFlagIsFloat;
 
     if ((ao->format & AF_FORMAT_SIGN_MASK) == AF_FORMAT_SI)
         asbd->mFormatFlags |= kAudioFormatFlagIsSignedInteger;
 
-    if ((ao->format & AF_FORMAT_END_MASK) == AF_FORMAT_BE)
+    if (BYTE_ORDER == BIG_ENDIAN)
         asbd->mFormatFlags |= kAudioFormatFlagIsBigEndian;
 
     asbd->mFramesPerPacket = 1;

audio/out/ao_coreaudio_utils.h

@@ -46,8 +46,8 @@ bool check_ca_st(struct ao *ao, int level, OSStatus code, const char *message);
         if (err != noErr) goto label; \
     } while (0)
 
-void ca_print_device_list(struct ao *ao);
-OSStatus ca_select_device(struct ao *ao, int selection, AudioDeviceID *device);
+void ca_get_device_list(struct ao *ao, struct ao_device_list *list);
+OSStatus ca_select_device(struct ao *ao, char* name, AudioDeviceID *device);
 
 void ca_fill_asbd(struct ao *ao, AudioStreamBasicDescription *asbd);
 void ca_print_asbd(struct ao *ao, const char *description,

audio/out/ao_dsound.c

@@ -41,6 +41,7 @@
 #include "internal.h"
 #include "common/msg.h"
 #include "osdep/timer.h"
+#include "osdep/io.h"
 #include "options/m_option.h"
 
 /**
@@ -98,9 +99,10 @@ struct priv {
     int outburst;                   ///play in multiple of chunks of this size
 
     int cfg_device;
-};
+    int cfg_buffersize;
 
-static float get_delay(struct ao *ao);
+    struct ao_device_list *listing; ///temporary during list_devs()
+};
 
 /***************************************************************************************/
 
@@ -171,11 +173,84 @@ static BOOL CALLBACK DirectSoundEnum(LPGUID guid, LPCSTR desc, LPCSTR module,
         if (guid)
             memcpy(&p->device, guid, sizeof(GUID));
     }
+    char *guidstr = talloc_strdup(NULL, "");
+    if (guid) {
+        wchar_t guidwstr[80] = {0};
+        StringFromGUID2(guid, guidwstr, MP_ARRAY_SIZE(guidwstr));
+        char *nstr = mp_to_utf8(NULL, guidwstr);
+        if (nstr) {
+            talloc_free(guidstr);
+            guidstr = nstr;
+        }
+    }
+    if (p->device_num < 0 && ao->device) {
+        if (strcmp(ao->device, guidstr) == 0) {
+            MP_VERBOSE(ao, "<--");
+            p->device_num = p->device_index;
+            if (guid)
+                memcpy(&p->device, guid, sizeof(GUID));
+        }
+    }
+    if (p->listing) {
+        struct ao_device_desc e = {guidstr, desc};
+        ao_device_list_add(p->listing, ao, &e);
+    }
+    talloc_free(guidstr);
+
     MP_VERBOSE(ao, "\n");
     p->device_index++;
     return TRUE;
 }
 
+static void EnumDevs(struct ao *ao)
+{
+    struct priv *p = ao->priv;
+
+    p->device_index = 0;
+    p->device_num = p->cfg_device;
+
+    HRESULT (WINAPI *OurDirectSoundEnumerate)(LPDSENUMCALLBACKA, LPVOID);
+    OurDirectSoundEnumerate = (void *)GetProcAddress(p->hdsound_dll,
+                                                     "DirectSoundEnumerateA");
+
+    if (OurDirectSoundEnumerate == NULL) {
+        MP_ERR(ao, "GetProcAddress FAILED\n");
+        return;
+    }
+
+    // Enumerate all directsound p->devices
+    MP_VERBOSE(ao, "Output Devices:\n");
+    OurDirectSoundEnumerate(DirectSoundEnum, ao);
+}
+
+static int LoadDirectSound(struct ao *ao)
+{
+    struct priv *p = ao->priv;
+
+    // initialize directsound
+    p->hdsound_dll = LoadLibrary("DSOUND.DLL");
+    if (p->hdsound_dll == NULL) {
+        MP_ERR(ao, "cannot load DSOUND.DLL\n");
+        return 0;
+    }
+
+    return 1;
+}
+
+static void list_devs(struct ao *ao, struct ao_device_list *list)
+{
+    struct priv *p = ao->priv;
+    bool need_init = !p->hdsound_dll;
+    if (need_init && !LoadDirectSound(ao))
+        return;
+
+    p->listing = list;
+    EnumDevs(ao);
+    p->listing = NULL;
+
+    if (need_init)
+        UninitDirectSound(ao);
+}
 
 /**
 \brief initilize direct sound
@@ -187,34 +262,23 @@ static int InitDirectSound(struct ao *ao)
 
     DSCAPS dscaps;
 
-    // initialize directsound
-    HRESULT (WINAPI *OurDirectSoundCreate)(LPGUID, LPDIRECTSOUND *, LPUNKNOWN);
-    HRESULT (WINAPI *OurDirectSoundEnumerate)(LPDSENUMCALLBACKA, LPVOID);
-    p->device_index = 0;
-    p->device_num = p->cfg_device;
-
-    p->hdsound_dll = LoadLibrary("DSOUND.DLL");
-    if (p->hdsound_dll == NULL) {
-        MP_ERR(ao, "cannot load DSOUND.DLL\n");
+    if (!LoadDirectSound(ao))
         return 0;
-    }
-    OurDirectSoundCreate = (void *)GetProcAddress(p->hdsound_dll,
-                                                  "DirectSoundCreate");
-    OurDirectSoundEnumerate = (void *)GetProcAddress(p->hdsound_dll,
-                                                     "DirectSoundEnumerateA");
 
-    if (OurDirectSoundCreate == NULL || OurDirectSoundEnumerate == NULL) {
+    HRESULT (WINAPI *OurDirectSoundCreate)(LPGUID, LPDIRECTSOUND *, LPUNKNOWN);
+    OurDirectSoundCreate =
+        (void *)GetProcAddress(p->hdsound_dll, "DirectSoundCreate");
+
+    if (OurDirectSoundCreate == NULL) {
         MP_ERR(ao, "GetProcAddress FAILED\n");
         FreeLibrary(p->hdsound_dll);
         return 0;
     }
 
-    // Enumerate all directsound p->devices
-    MP_VERBOSE(ao, "Output Devices:\n");
-    OurDirectSoundEnumerate(DirectSoundEnum, ao);
+    EnumDevs(ao);
 
     // Create the direct sound object
-    if (FAILED(OurDirectSoundCreate((p->device_num) ? &p->device : NULL,
+    if (FAILED(OurDirectSoundCreate((p->device_num > 0) ? &p->device : NULL,
                                     &p->hds, NULL)))
     {
         MP_ERR(ao, "cannot create a DirectSound device\n");
@@ -350,6 +414,8 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
         IDirectSoundBuffer_SetVolume(p->hdsbuf, volume);
         return CONTROL_OK;
     }
+    case AOCONTROL_HAS_SOFT_VOLUME:
+        return CONTROL_TRUE;
     }
     return CONTROL_UNKNOWN;
 }
@@ -370,7 +436,6 @@ static int init(struct ao *ao)
     if (!InitDirectSound(ao))
         return -1;
 
-    ao->no_persistent_volume = true;
     p->audio_volume = 100;
 
     // ok, now create the buffers
@@ -380,34 +445,33 @@ static int init(struct ao *ao)
     int format = af_fmt_from_planar(ao->format);
     int rate = ao->samplerate;
 
-    if (AF_FORMAT_IS_AC3(format))
-        format = AF_FORMAT_AC3;
-    else {
+    if (!AF_FORMAT_IS_IEC61937(format)) {
         struct mp_chmap_sel sel = {0};
         mp_chmap_sel_add_waveext(&sel);
         if (!ao_chmap_sel_adjust(ao, &sel, &ao->channels))
             return -1;
     }
     switch (format) {
-    case AF_FORMAT_AC3:
-    case AF_FORMAT_S24_LE:
-    case AF_FORMAT_S16_LE:
+    case AF_FORMAT_S24:
+    case AF_FORMAT_S16:
     case AF_FORMAT_U8:
         break;
     default:
+        if (AF_FORMAT_IS_IEC61937(format))
+            break;
         MP_VERBOSE(ao, "format %s not supported defaulting to Signed 16-bit Little-Endian\n",
                    af_fmt_to_str(format));
-        format = AF_FORMAT_S16_LE;
+        format = AF_FORMAT_S16;
     }
     //set our audio parameters
     ao->samplerate = rate;
     ao->format = format;
     ao->bps = ao->channels.num * rate * af_fmt2bps(format);
-    int buffersize = ao->bps * 0.100; // space for 100ms
+    int buffersize = ao->bps * p->cfg_buffersize / 1000;
     MP_VERBOSE(ao, "Samplerate:%iHz Channels:%i Format:%s\n", rate,
                ao->channels.num, af_fmt_to_str(format));
-    MP_VERBOSE(ao, "Buffersize:%d bytes (%d msec)\n",
-               buffersize, buffersize / ao->bps * 1000);
+    MP_VERBOSE(ao, "Buffersize:%d bytes (%f msec)\n",
+               buffersize, buffersize * 1000.0 / ao->bps);
 
     //fill waveformatex
     ZeroMemory(&wformat, sizeof(WAVEFORMATEXTENSIBLE));
@@ -415,7 +479,8 @@ static int init(struct ao *ao)
                     ? sizeof(WAVEFORMATEXTENSIBLE) - sizeof(WAVEFORMATEX) : 0;
     wformat.Format.nChannels = ao->channels.num;
     wformat.Format.nSamplesPerSec = rate;
-    if (AF_FORMAT_IS_AC3(format)) {
+    if (AF_FORMAT_IS_IEC61937(format)) {
+        // Whether it also works with e.g. DTS is unknown, but probably does.
         wformat.Format.wFormatTag = WAVE_FORMAT_DOLBY_AC3_SPDIF;
         wformat.Format.wBitsPerSample = 16;
         wformat.Format.nBlockAlign = 4;
@@ -583,7 +648,7 @@ static int get_space(struct ao *ao)
     int space = check_free_buffer_size(ao);
     if (space < p->min_free_space)
         return 0;
-    return (space - p->min_free_space) / ao->sstride;
+    return (space - p->min_free_space) / p->outburst * p->outburst / ao->sstride;
 }
 
 /**
@@ -611,12 +676,12 @@ static int play(struct ao *ao, void **data, int samples, int flags)
 \brief get the delay between the first and last sample in the buffer
 \return delay in seconds
 */
-static float get_delay(struct ao *ao)
+static double get_delay(struct ao *ao)
 {
     struct priv *p = ao->priv;
 
     int space = check_free_buffer_size(ao);
-    return (float)(p->buffer_size - space) / (float)ao->bps;
+    return (p->buffer_size - space) / (double)ao->bps;
 }
 
 #define OPT_BASE_STRUCT struct priv
@@ -633,9 +698,11 @@ const struct ao_driver audio_out_dsound = {
     .pause     = audio_pause,
     .resume    = audio_resume,
     .reset     = reset,
+    .list_devs = list_devs,
     .priv_size = sizeof(struct priv),
     .options = (const struct m_option[]) {
-        OPT_INT("device", cfg_device, 0),
+        OPT_INT("device", cfg_device, 0, OPTDEF_INT(-1)),
+        OPT_INTRANGE("buffersize", cfg_buffersize, 0, 1, 10000, OPTDEF_INT(200)),
         {0}
     },
 };

audio/out/ao_lavc.c

@@ -28,7 +28,6 @@
 #include <libavutil/common.h>
 #include <libavutil/audioconvert.h>
 
-#include "compat/libav.h"
 #include "config.h"
 #include "options/options.h"
 #include "common/common.h"
@@ -57,6 +56,8 @@ struct priv {
 
     AVRational worst_time_base;
     int worst_time_base_is_stream;
+
+    bool shutdown;
 };
 
 static void select_format(struct ao *ao, AVCodec *codec)
@@ -175,6 +176,7 @@ static int init(struct ao *ao)
 
 fail:
     pthread_mutex_unlock(&ao->encode_lavc_ctx->lock);
+    ac->shutdown = true;
     return -1;
 }
 
@@ -185,6 +187,9 @@ static void uninit(struct ao *ao)
     struct priv *ac = ao->priv;
     struct encode_lavc_context *ectx = ao->encode_lavc_ctx;
 
+    if (!ac || ac->shutdown)
+        return;
+
     pthread_mutex_lock(&ectx->lock);
 
     if (!encode_lavc_start(ectx)) {
@@ -202,6 +207,8 @@ static void uninit(struct ao *ao)
     }
 
     pthread_mutex_unlock(&ectx->lock);
+
+    ac->shutdown = true;
 }
 
 // return: how many bytes can be played without blocking
@@ -235,8 +242,9 @@ static int encode(struct ao *ao, double apts, void **data)
         frame->format = af_to_avformat(ao->format);
         frame->nb_samples = ac->aframesize;
 
-        assert(ao->channels.num <= AV_NUM_DATA_POINTERS);
-        for (int n = 0; n < ao->channels.num; n++)
+        size_t num_planes = af_fmt_is_planar(ao->format) ? ao->channels.num : 1;
+        assert(num_planes <= AV_NUM_DATA_POINTERS);
+        for (int n = 0; n < num_planes; n++)
             frame->extended_data[n] = data[n];
 
         frame->linesize[0] = frame->nb_samples * ao->sstride;
@@ -346,12 +354,12 @@ static int play(struct ao *ao, void **data, int samples, int flags)
     size_t num_planes = af_fmt_is_planar(ao->format) ? ao->channels.num : 1;
 
     void *tempdata = NULL;
+    void *padded[MP_NUM_CHANNELS];
 
     if ((flags & AOPLAY_FINAL_CHUNK) && (samples % ac->aframesize)) {
        tempdata = talloc_new(NULL);
        size_t bytelen = samples * ao->sstride;
        size_t extralen = (ac->aframesize - 1) * ao->sstride;
-       void *padded[MP_NUM_CHANNELS];
        for (int n = 0; n < num_planes; n++) {
            padded[n] = talloc_size(tempdata, bytelen + extralen);
            memcpy(padded[n], data[n], bytelen);
@@ -431,7 +439,7 @@ static int play(struct ao *ao, void **data, int samples, int flags)
     outpts += encode_lavc_getoffset(ectx, ac->stream);
 
     while (samples - bufpos >= ac->aframesize) {
-        void *start[MP_NUM_CHANNELS];
+        void *start[MP_NUM_CHANNELS] = {0};
         for (int n = 0; n < num_planes; n++)
             start[n] = (char *)data[n] + bufpos * ao->sstride;
         encode(ao, outpts + bufpos / (double) ao->samplerate, start);

audio/out/ao_null.c

@@ -148,7 +148,8 @@ static int get_space(struct ao *ao)
     struct priv *priv = ao->priv;
 
     drain(ao);
-    return priv->buffersize - priv->latency - priv->buffered;
+    int samples = priv->buffersize - priv->latency - priv->buffered;
+    return samples / priv->outburst * priv->outburst;
 }
 
 static int play(struct ao *ao, void **data, int samples, int flags)
@@ -175,7 +176,7 @@ static int play(struct ao *ao, void **data, int samples, int flags)
     return accepted;
 }
 
-static float get_delay(struct ao *ao)
+static double get_delay(struct ao *ao)
 {
     struct priv *priv = ao->priv;
 

audio/out/ao_openal.c

@@ -76,6 +76,8 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
         vol->left = vol->right = volume * 100;
         return CONTROL_TRUE;
     }
+    case AOCONTROL_HAS_SOFT_VOLUME:
+        return CONTROL_TRUE;
     }
     return CONTROL_UNKNOWN;
 }
@@ -99,6 +101,17 @@ static int validate_device_opt(struct mp_log *log, const m_option_t *opt,
     return 0;
 }
 
+static void list_devs(struct ao *ao, struct ao_device_list *list)
+{
+    if (alcIsExtensionPresent(NULL, "ALC_ENUMERATE_ALL_EXT") != AL_TRUE)
+        return;
+    const char *devs = alcGetString(NULL, ALC_ALL_DEVICES_SPECIFIER);
+    while (devs && *devs) {
+        ao_device_list_add(list, ao, &(struct ao_device_desc){devs, devs});
+        devs = devs + strlen(devs) + 1;
+    }
+}
+
 struct speaker {
     int id;
     float pos[3];
@@ -132,7 +145,6 @@ static int init(struct ao *ao)
         return -1;
     }
     ao_data = ao;
-    ao->no_persistent_volume = true;
     struct mp_chmap_sel sel = {0};
     for (i = 0; speaker_pos[i].id != -1; i++)
         mp_chmap_sel_add_speaker(&sel, speaker_pos[i].id);
@@ -150,7 +162,10 @@ static int init(struct ao *ao)
             goto err_out;
         }
     }
-    dev = alcOpenDevice(p->cfg_device && p->cfg_device[0] ? p->cfg_device : NULL);
+    char *dev_name = p->cfg_device;
+    if (!dev_name || !dev_name[0])
+        dev_name = ao->device;
+    dev = alcOpenDevice(dev_name && dev_name[0] ? dev_name : NULL);
     if (!dev) {
         MP_FATAL(ao, "could not open device\n");
         goto err_out;
@@ -278,12 +293,12 @@ static int play(struct ao *ao, void **data, int samples, int flags)
     return num * CHUNK_SAMPLES;
 }
 
-static float get_delay(struct ao *ao)
+static double get_delay(struct ao *ao)
 {
     ALint queued;
     unqueue_buffers();
     alGetSourcei(sources[0], AL_BUFFERS_QUEUED, &queued);
-    return queued * CHUNK_SAMPLES / (float)ao->samplerate;
+    return queued * CHUNK_SAMPLES / (double)ao->samplerate;
 }
 
 #define OPT_BASE_STRUCT struct priv
@@ -301,6 +316,7 @@ const struct ao_driver audio_out_openal = {
     .resume    = audio_resume,
     .reset     = reset,
     .drain     = drain,
+    .list_devs = list_devs,
     .priv_size = sizeof(struct priv),
     .options = (const struct m_option[]) {
         OPT_STRING_VALIDATE("device", cfg_device, 0, validate_device_opt),

audio/out/ao_oss.c

@@ -31,6 +31,7 @@
 #include <sys/types.h>
 #include <sys/stat.h>
 #include <fcntl.h>
+#include <poll.h>
 #include <errno.h>
 #include <string.h>
 #include <strings.h>
@@ -38,6 +39,8 @@
 #include "config.h"
 #include "options/options.h"
 #include "common/msg.h"
+#include "osdep/timer.h"
+#include "osdep/endian.h"
 
 #if HAVE_SYS_SOUNDCARD_H
 #include <sys/soundcard.h>
@@ -52,14 +55,20 @@
 #include "ao.h"
 #include "internal.h"
 
+// Define to 0 if the device must be reopened to reset it (stop all playback,
+// clear the buffer), and the device should be closed when unused.
+// Define to 1 if SNDCTL_DSP_RESET should be used to reset without close.
+#define KEEP_DEVICE (defined(SNDCTL_DSP_RESET) && !defined(__NetBSD__))
+
 struct priv {
     int audio_fd;
-    int prepause_space;
+    int prepause_samples;
     int oss_mixer_channel;
-    audio_buf_info zz;
     int audio_delay_method;
     int buffersize;
     int outburst;
+    bool device_failed;
+    double audio_end;
 
     char *dsp;
     char *oss_mixer_device;
@@ -81,53 +90,60 @@ static const struct mp_chmap oss_layouts[MP_NUM_CHANNELS + 1] = {
     MP_CHMAP8(FL, FR, BL, BR, FC, LFE, SL, SR), // 7.1
 };
 
+#if !defined(AFMT_S16_NE) && defined(AFMT_S16_LE) && defined(AFMT_S16_BE)
+#define AFMT_S16_NE MP_SELECT_LE_BE(AFMT_S16_LE, AFMT_S16_BE)
+#endif
+
+#if !defined(AFMT_U16_NE) && defined(AFMT_U16_LE) && defined(AFMT_U16_BE)
+#define AFMT_U16_NE MP_SELECT_LE_BE(AFMT_U16_LE, AFMT_U16_BE)
+#endif
+
+#if !defined(AFMT_U24_NE) && defined(AFMT_U24_LE) && defined(AFMT_U24_BE)
+#define AFMT_U24_NE MP_SELECT_LE_BE(AFMT_U24_LE, AFMT_U24_BE)
+#endif
+
+#if !defined(AFMT_S24_NE) && defined(AFMT_S24_LE) && defined(AFMT_S24_BE)
+#define AFMT_S24_NE MP_SELECT_LE_BE(AFMT_S24_LE, AFMT_S24_BE)
+#endif
+
+#if !defined(AFMT_U32_NE) && defined(AFMT_U32_LE) && defined(AFMT_U32_BE)
+#define AFMT_U32MP_SELECT_LE_BE(AFMT_U32_LE, AFMT_U32_BE)
+#endif
+
+#if !defined(AFMT_S32_NE) && defined(AFMT_S32_LE) && defined(AFMT_S32_BE)
+#define AFMT_S32MP_SELECT_LE_BE(AFMT_S32_LE, AFMT_S32_BE)
+#endif
+
 static const int format_table[][2] = {
     {AFMT_U8,           AF_FORMAT_U8},
     {AFMT_S8,           AF_FORMAT_S8},
-    {AFMT_U16_LE,       AF_FORMAT_U16_LE},
-    {AFMT_U16_BE,       AF_FORMAT_U16_BE},
-    {AFMT_S16_LE,       AF_FORMAT_S16_LE},
-    {AFMT_S16_BE,       AF_FORMAT_S16_BE},
-#ifdef AFMT_S24_PACKED
-    {AFMT_S24_PACKED,   AF_FORMAT_S24_LE},
+    {AFMT_U16_NE,       AF_FORMAT_U16},
+    {AFMT_S16_NE,       AF_FORMAT_S16},
+#ifdef AFMT_U24_NE
+    {AFMT_U24_NE,       AF_FORMAT_U24},
 #endif
-#ifdef AFMT_U24_LE
-    {AFMT_U24_LE,       AF_FORMAT_U24_LE},
+#ifdef AFMT_S24_NE
+    {AFMT_S24_NE,       AF_FORMAT_S24},
 #endif
-#ifdef AFMT_U24_BE
-    {AFMT_U24_BE,       AF_FORMAT_U24_BE},
+#ifdef AFMT_U32_NE
+    {AFMT_U32_NE,       AF_FORMAT_U32},
 #endif
-#ifdef AFMT_S24_LE
-    {AFMT_S24_LE,       AF_FORMAT_S24_LE},
-#endif
-#ifdef AFMT_S24_BE
-    {AFMT_S24_BE,       AF_FORMAT_S24_BE},
-#endif
-#ifdef AFMT_U32_LE
-    {AFMT_U32_LE,       AF_FORMAT_U32_LE},
-#endif
-#ifdef AFMT_U32_BE
-    {AFMT_U32_BE,       AF_FORMAT_U32_BE},
-#endif
-#ifdef AFMT_S32_LE
-    {AFMT_S32_LE,       AF_FORMAT_S32_LE},
-#endif
-#ifdef AFMT_S32_BE
-    {AFMT_S32_BE,       AF_FORMAT_S32_BE},
+#ifdef AFMT_S32_NE
+    {AFMT_S32_NE,       AF_FORMAT_S32},
 #endif
 #ifdef AFMT_FLOAT
     {AFMT_FLOAT,        AF_FORMAT_FLOAT},
 #endif
-    // SPECIALS
 #ifdef AFMT_MPEG
-    {AFMT_MPEG,         AF_FORMAT_MPEG2},
-#endif
-#ifdef AFMT_AC3
-    {AFMT_AC3,          AF_FORMAT_AC3},
+    {AFMT_MPEG,         AF_FORMAT_S_MP3},
 #endif
     {-1, -1}
 };
 
+#ifndef AFMT_AC3
+#define AFMT_AC3 -1
+#endif
+
 static int format2oss(int format)
 {
     for (int n = 0; format_table[n][0] != -1; n++) {
@@ -143,7 +159,7 @@ static int oss2format(int format)
         if (format_table[n][0] == format)
             return format_table[n][1];
     }
-    return -1;
+    return 0;
 }
 
 
@@ -178,8 +194,7 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
     struct priv *p = ao->priv;
     switch (cmd) {
     case AOCONTROL_GET_VOLUME:
-    case AOCONTROL_SET_VOLUME:
-    {
+    case AOCONTROL_SET_VOLUME: {
         ao_control_vol_t *vol = (ao_control_vol_t *)arg;
         int fd, v, devs;
 
@@ -189,7 +204,7 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
             return CONTROL_OK;
 #endif
 
-        if (AF_FORMAT_IS_AC3(ao->format))
+        if (AF_FORMAT_IS_SPECIAL(ao->format))
             return CONTROL_TRUE;
 
         if ((fd = open(p->oss_mixer_device, O_RDONLY)) != -1) {
@@ -210,22 +225,205 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
             close(fd);
             return CONTROL_OK;
         }
-    }
         return CONTROL_ERROR;
     }
+#ifdef SNDCTL_DSP_GETPLAYVOL
+    case AOCONTROL_HAS_SOFT_VOLUME:
+        return CONTROL_TRUE;
+#endif
+    }
     return CONTROL_UNKNOWN;
 }
 
+// 1: ok, 0: not writable, -1: error
+static int device_writable(struct ao *ao)
+{
+    struct priv *p = ao->priv;
+    struct pollfd fd = {.fd = p->audio_fd, .events = POLLOUT};
+    return poll(&fd, 1, 0);
+}
+
+static void close_device(struct ao *ao)
+{
+    struct priv *p = ao->priv;
+    p->device_failed = false;
+    if (p->audio_fd == -1)
+        return;
+#if defined(SNDCTL_DSP_RESET)
+    ioctl(p->audio_fd, SNDCTL_DSP_RESET, NULL);
+#endif
+    close(p->audio_fd);
+    p->audio_fd = -1;
+}
+
+// close audio device
+static void uninit(struct ao *ao)
+{
+    close_device(ao);
+}
+
+static bool try_format(struct ao *ao, int *format)
+{
+    struct priv *p = ao->priv;
+
+    int oss_format = format2oss(*format);
+    if (oss_format == -1 && AF_FORMAT_IS_IEC61937(*format))
+        oss_format = AFMT_AC3;
+
+    if (oss_format == -1) {
+        MP_VERBOSE(ao, "Unknown/not supported internal format: %s\n",
+                   af_fmt_to_str(*format));
+        *format = 0;
+        return false;
+    }
+
+    int actual_format = oss_format;
+    if (ioctl(p->audio_fd, SNDCTL_DSP_SETFMT, &actual_format) < 0)
+        actual_format = -1;
+
+    if (actual_format == oss_format)
+        return true;
+
+    MP_WARN(ao, "Can't set audio device to %s output.\n", af_fmt_to_str(*format));
+    *format = oss2format(actual_format);
+    if (actual_format != -1 && !*format)
+        MP_ERR(ao, "Unknown/Unsupported OSS format: 0x%x.\n", actual_format);
+    return false;
+}
+
+static int reopen_device(struct ao *ao, bool allow_format_changes)
+{
+    struct priv *p = ao->priv;
+
+    int samplerate = ao->samplerate;
+    int format = ao->format;
+    struct mp_chmap channels = ao->channels;
+
+#ifdef __linux__
+    p->audio_fd = open(p->dsp, O_WRONLY | O_NONBLOCK);
+#else
+    p->audio_fd = open(p->dsp, O_WRONLY);
+#endif
+    if (p->audio_fd < 0) {
+        MP_ERR(ao, "Can't open audio device %s: %s\n", p->dsp, strerror(errno));
+        goto fail;
+    }
+
+#ifdef __linux__
+    /* Remove the non-blocking flag */
+    if (fcntl(p->audio_fd, F_SETFL, 0) < 0) {
+        MP_ERR(ao, "Can't make file descriptor blocking: %s\n",  strerror(errno));
+        goto fail;
+    }
+#endif
+
+#if defined(FD_CLOEXEC) && defined(F_SETFD)
+    fcntl(p->audio_fd, F_SETFD, FD_CLOEXEC);
+#endif
+
+    if (AF_FORMAT_IS_IEC61937(format)) {
+        if (ioctl(p->audio_fd, SNDCTL_DSP_SPEED, &samplerate) == -1)
+            goto fail;
+        // Probably could be fixed by setting number of channels; needs testing.
+        if (channels.num != 2) {
+            MP_ERR(ao, "Format %s not implemented.\n", af_fmt_to_str(format));
+            goto fail;
+        }
+    }
+
+    int try_formats[] = {format, AF_FORMAT_S32, AF_FORMAT_S24, AF_FORMAT_S16, 0};
+    for (int n = 0; try_formats[n]; n++) {
+        format = try_formats[n];
+        if (try_format(ao, &format))
+            break;
+    }
+
+    if (!format) {
+        MP_ERR(ao, "Can't set sample format.\n");
+        goto fail;
+    }
+
+    MP_VERBOSE(ao, "sample format: %s\n", af_fmt_to_str(format));
+
+    if (!AF_FORMAT_IS_IEC61937(format)) {
+        struct mp_chmap_sel sel = {0};
+        for (int n = 0; n < MP_NUM_CHANNELS + 1; n++)
+            mp_chmap_sel_add_map(&sel, &oss_layouts[n]);
+        if (!ao_chmap_sel_adjust(ao, &sel, &channels))
+            goto fail;
+        int reqchannels = channels.num;
+        // We only use SNDCTL_DSP_CHANNELS for >2 channels, in case some drivers don't have it
+        if (reqchannels > 2) {
+            int nchannels = reqchannels;
+            if (ioctl(p->audio_fd, SNDCTL_DSP_CHANNELS, &nchannels) == -1 ||
+                nchannels != reqchannels)
+            {
+                MP_ERR(ao, "Failed to set audio device to %d channels.\n",
+                       reqchannels);
+                goto fail;
+            }
+        } else {
+            int c = reqchannels - 1;
+            if (ioctl(p->audio_fd, SNDCTL_DSP_STEREO, &c) == -1) {
+                MP_ERR(ao, "Failed to set audio device to %d channels.\n",
+                       reqchannels);
+                goto fail;
+            }
+            if (!ao_chmap_sel_get_def(ao, &sel, &channels, c + 1))
+                goto fail;
+        }
+        MP_VERBOSE(ao, "using %d channels (requested: %d)\n",
+                   channels.num, reqchannels);
+        // set rate
+        if (ioctl(p->audio_fd, SNDCTL_DSP_SPEED, &samplerate) == -1)
+            goto fail;
+        MP_VERBOSE(ao, "using %d Hz samplerate\n", samplerate);
+    }
+
+    audio_buf_info zz = {0};
+    if (ioctl(p->audio_fd, SNDCTL_DSP_GETOSPACE, &zz) == -1) {
+        int r = 0;
+        MP_WARN(ao, "driver doesn't support SNDCTL_DSP_GETOSPACE\n");
+        if (ioctl(p->audio_fd, SNDCTL_DSP_GETBLKSIZE, &r) == -1)
+            MP_VERBOSE(ao, "%d bytes/frag (config.h)\n", p->outburst);
+        else {
+            p->outburst = r;
+            MP_VERBOSE(ao, "%d bytes/frag (GETBLKSIZE)\n", p->outburst);
+        }
+    } else {
+        MP_VERBOSE(ao, "frags: %3d/%d  (%d bytes/frag)  free: %6d\n",
+                   zz.fragments, zz.fragstotal, zz.fragsize, zz.bytes);
+        p->buffersize = zz.bytes;
+        p->outburst = zz.fragsize;
+    }
+
+    if (allow_format_changes) {
+        ao->format = format;
+        ao->samplerate = samplerate;
+        ao->channels = channels;
+    } else {
+        if (format != ao->format || samplerate != ao->samplerate ||
+            !mp_chmap_equals(&channels, &ao->channels))
+        {
+            MP_ERR(ao, "Could not reselect previous audio format.\n");
+            goto fail;
+        }
+    }
+
+    p->outburst -= p->outburst % (channels.num * af_fmt2bps(format)); // round down
+
+    return 0;
+
+fail:
+    close_device(ao);
+    return -1;
+}
+
 // open & setup audio device
 // return: 0=success -1=fail
 static int init(struct ao *ao)
 {
     struct priv *p = ao->priv;
-    int oss_format;
-
-#ifdef SNDCTL_DSP_GETPLAYVOL
-    ao->no_persistent_volume = true;
-#endif
 
     const char *mchan = NULL;
     if (p->cfg_oss_mixer_channel && p->cfg_oss_mixer_channel[0])
@@ -266,178 +464,36 @@ static int init(struct ao *ao)
     MP_VERBOSE(ao, "using '%s' mixer device\n", p->oss_mixer_device);
     MP_VERBOSE(ao, "using '%s' mixer device\n", mixer_channels[p->oss_mixer_channel]);
 
-#ifdef __linux__
-    p->audio_fd = open(p->dsp, O_WRONLY | O_NONBLOCK);
-#else
-    p->audio_fd = open(p->dsp, O_WRONLY);
-#endif
-    if (p->audio_fd < 0) {
-        MP_ERR(ao, "Can't open audio device %s: %s\n", p->dsp, strerror(errno));
-        return -1;
-    }
-
-#ifdef __linux__
-    /* Remove the non-blocking flag */
-    if (fcntl(p->audio_fd, F_SETFL, 0) < 0) {
-        MP_ERR(ao, "Can't make file descriptor blocking: %s\n",  strerror(errno));
-        return -1;
-    }
-#endif
-
-#if defined(FD_CLOEXEC) && defined(F_SETFD)
-    fcntl(p->audio_fd, F_SETFD, FD_CLOEXEC);
-#endif
-
     ao->format = af_fmt_from_planar(ao->format);
 
-    if (AF_FORMAT_IS_AC3(ao->format)) {
-        ioctl(p->audio_fd, SNDCTL_DSP_SPEED, &ao->samplerate);
-    }
-
-ac3_retry:
-    if (AF_FORMAT_IS_AC3(ao->format))
-        ao->format = AF_FORMAT_AC3;
-    oss_format = format2oss(ao->format);
-    if (oss_format == -1) {
-        MP_VERBOSE(ao, "Unknown/not supported internal format: %s\n",
-                   af_fmt_to_str(ao->format));
-#if defined(AFMT_S32_LE) && defined(AFMT_S32_BE)
-#if BYTE_ORDER == BIG_ENDIAN
-        oss_format = AFMT_S32_BE;
-#else
-        oss_format = AFMT_S32_LE;
-#endif
-        ao->format = AF_FORMAT_S32;
-#elif defined(AFMT_S24_LE) && defined(AFMT_S24_BE)
-#if BYTE_ORDER == BIG_ENDIAN
-        oss_format = AFMT_S24_BE;
-#else
-        oss_format = AFMT_S24_LE;
-#endif
-        ao->format = AF_FORMAT_S24;
-#else
-#if BYTE_ORDER == BIG_ENDIAN
-        oss_format = AFMT_S16_BE;
-#else
-        oss_format = AFMT_S16_LE;
-#endif
-        ao->format = AF_FORMAT_S16;
-#endif
-    }
-    if (ioctl(p->audio_fd, SNDCTL_DSP_SETFMT, &oss_format) < 0 ||
-        oss_format != format2oss(ao->format))
-    {
-        MP_WARN(ao, "Can't set audio device %s to %s output, trying %s...\n",
-                p->dsp, af_fmt_to_str(ao->format),
-                af_fmt_to_str(AF_FORMAT_S16));
-        ao->format = AF_FORMAT_S16;
-        goto ac3_retry;
-    }
-
-    ao->format = oss2format(oss_format);
-    if (ao->format == -1) {
-        MP_ERR(ao, "Unknown/Unsupported OSS format: %x.\n", oss_format);
-        return -1;
-    }
-
-    MP_VERBOSE(ao, "sample format: %s\n", af_fmt_to_str(ao->format));
-
-    if (!AF_FORMAT_IS_AC3(ao->format)) {
-        struct mp_chmap_sel sel = {0};
-        for (int n = 0; n < MP_NUM_CHANNELS + 1; n++)
-            mp_chmap_sel_add_map(&sel, &oss_layouts[n]);
-        if (!ao_chmap_sel_adjust(ao, &sel, &ao->channels))
-            return -1;
-        int reqchannels = ao->channels.num;
-        // We only use SNDCTL_DSP_CHANNELS for >2 channels, in case some drivers don't have it
-        if (reqchannels > 2) {
-            int nchannels = reqchannels;
-            if (ioctl(p->audio_fd, SNDCTL_DSP_CHANNELS, &nchannels) == -1 ||
-                nchannels != reqchannels)
-            {
-                MP_ERR(ao, "Failed to set audio device to %d channels.\n",
-                       reqchannels);
-                return -1;
-            }
-        } else {
-            int c = reqchannels - 1;
-            if (ioctl(p->audio_fd, SNDCTL_DSP_STEREO, &c) == -1) {
-                MP_ERR(ao, "Failed to set audio device to %d channels.\n",
-                       reqchannels);
-                return -1;
-            }
-            if (!ao_chmap_sel_get_def(ao, &sel, &ao->channels, c + 1))
-                return -1;
-        }
-        MP_VERBOSE(ao, "using %d channels (requested: %d)\n",
-                   ao->channels.num, reqchannels);
-        // set rate
-        ioctl(p->audio_fd, SNDCTL_DSP_SPEED, &ao->samplerate);
-        MP_VERBOSE(ao, "using %d Hz samplerate\n", ao->samplerate);
-    }
-
-    if (ioctl(p->audio_fd, SNDCTL_DSP_GETOSPACE, &p->zz) == -1) {
-        int r = 0;
-        MP_WARN(ao, "driver doesn't support SNDCTL_DSP_GETOSPACE\n");
-        if (ioctl(p->audio_fd, SNDCTL_DSP_GETBLKSIZE, &r) == -1)
-            MP_VERBOSE(ao, "%d bytes/frag (config.h)\n", p->outburst);
-        else {
-            p->outburst = r;
-            MP_VERBOSE(ao, "%d bytes/frag (GETBLKSIZE)\n", p->outburst);
-        }
-    } else {
-        MP_VERBOSE(ao, "frags: %3d/%d  (%d bytes/frag)  free: %6d\n",
-                   p->zz.fragments, p->zz.fragstotal, p->zz.fragsize, p->zz.bytes);
-        p->buffersize = p->zz.bytes;
-        p->outburst = p->zz.fragsize;
-    }
+    if (reopen_device(ao, true) < 0)
+        goto fail;
 
     if (p->buffersize == -1) {
         // Measuring buffer size:
-        void *data;
+        void *data = malloc(p->outburst);
+        if (!data) {
+            MP_ERR(ao, "Out of memory, or broken outburst size.\n");
+            goto fail;
+        }
         p->buffersize = 0;
-#if HAVE_AUDIO_SELECT
-        data = malloc(p->outburst);
         memset(data, 0, p->outburst);
-        while (p->buffersize < 0x40000) {
-            fd_set rfds;
-            struct timeval tv;
-            FD_ZERO(&rfds);
-            FD_SET(p->audio_fd, &rfds);
-            tv.tv_sec = 0;
-            tv.tv_usec = 0;
-            if (!select(p->audio_fd + 1, NULL, &rfds, NULL, &tv))
-                break;
+        while (p->buffersize < 0x40000 && device_writable(ao) > 0) {
             write(p->audio_fd, data, p->outburst);
             p->buffersize += p->outburst;
         }
         free(data);
         if (p->buffersize == 0) {
-            MP_ERR(ao, "***  Your audio driver DOES NOT support select()  ***\n");
-            MP_ERR(ao, "Recompile mpv with #define HAVE_AUDIO_SELECT 0 in config.h!\n");
-            return -1;
+            MP_ERR(ao, "Your OSS audio driver DOES NOT support poll().\n");
+            goto fail;
         }
-#endif
     }
 
-    ao->bps = ao->channels.num * af_fmt2bps(ao->format);
-    p->outburst -= p->outburst % ao->bps; // round down
-    ao->bps *= ao->samplerate;
-
     return 0;
-}
 
-// close audio device
-static void uninit(struct ao *ao)
-{
-    struct priv *p = ao->priv;
-    if (p->audio_fd == -1)
-        return;
-#ifdef SNDCTL_DSP_RESET
-    ioctl(p->audio_fd, SNDCTL_DSP_RESET, NULL);
-#endif
-    close(p->audio_fd);
-    p->audio_fd = -1;
+fail:
+    uninit(ao);
+    return -1;
 }
 
 static void drain(struct ao *ao)
@@ -450,104 +506,39 @@ static void drain(struct ao *ao)
 #endif
 }
 
-#ifndef SNDCTL_DSP_RESET
-static void close_device(struct ao *ao)
-{
-    struct priv *p = ao->priv;
-    close(p->audio_fd);
-    p->audio_fd = -1;
-}
-#endif
-
 // stop playing and empty buffers (for seeking/pause)
 static void reset(struct ao *ao)
 {
+#if KEEP_DEVICE
     struct priv *p = ao->priv;
-    int oss_format;
-#ifdef SNDCTL_DSP_RESET
-    ioctl(p->audio_fd, SNDCTL_DSP_RESET, NULL);
-#else
-    close_device(ao);
-    p->audio_fd = open(p->dsp, O_WRONLY);
-    if (p->audio_fd < 0) {
-        MP_ERR(ao, "Fatal error: *** CANNOT "
-               "RE-OPEN / RESET AUDIO DEVICE *** %s\n", strerror(errno));
-        return;
-    }
-
-#if defined(FD_CLOEXEC) && defined(F_SETFD)
-    fcntl(p->audio_fd, F_SETFD, FD_CLOEXEC);
-#endif
-#endif
-
-    oss_format = format2oss(ao->format);
-    if (AF_FORMAT_IS_AC3(ao->format))
-        ioctl(p->audio_fd, SNDCTL_DSP_SPEED, &ao->samplerate);
-    ioctl(p->audio_fd, SNDCTL_DSP_SETFMT, &oss_format);
-    if (!AF_FORMAT_IS_AC3(ao->format)) {
-        int c = ao->channels.num;
-        if (ao->channels.num > 2)
-            ioctl(p->audio_fd, SNDCTL_DSP_CHANNELS, &c);
-        else {
-            c--;
-            ioctl(p->audio_fd, SNDCTL_DSP_STEREO, &c);
-        }
-        ioctl(p->audio_fd, SNDCTL_DSP_SPEED, &ao->samplerate);
-    }
-}
-
-// return: how many bytes can be played without blocking
-static int get_space(struct ao *ao)
-{
-    struct priv *p = ao->priv;
-    int playsize = p->outburst;
-
-#ifdef SNDCTL_DSP_GETOSPACE
-    if (ioctl(p->audio_fd, SNDCTL_DSP_GETOSPACE, &p->zz) != -1) {
-        // calculate exact buffer space:
-        playsize = p->zz.fragments * p->zz.fragsize;
-        return playsize / ao->sstride;
-    }
-#endif
-
-    // check buffer
-#if HAVE_AUDIO_SELECT
-    {
-        fd_set rfds;
-        struct timeval tv;
-        FD_ZERO(&rfds);
-        FD_SET(p->audio_fd, &rfds);
-        tv.tv_sec = 0;
-        tv.tv_usec = 0;
-        if (!select(p->audio_fd + 1, NULL, &rfds, NULL, &tv))
-            return 0;                                            // not block!
-    }
-#endif
-
-    return p->outburst / ao->sstride;
-}
-
-// stop playing, keep buffers (for pause)
-static void audio_pause(struct ao *ao)
-{
-    struct priv *p = ao->priv;
-    p->prepause_space = get_space(ao) * ao->sstride;
-#ifdef SNDCTL_DSP_RESET
     ioctl(p->audio_fd, SNDCTL_DSP_RESET, NULL);
 #else
     close_device(ao);
 #endif
 }
 
-// plays 'len' bytes of 'data'
+// plays 'len' samples of 'data'
 // it should round it down to outburst*n
-// return: number of bytes played
+// return: number of samples played
 static int play(struct ao *ao, void **data, int samples, int flags)
 {
     struct priv *p = ao->priv;
     int len = samples * ao->sstride;
     if (len == 0)
         return len;
+
+    if (p->audio_fd < 0 && !p->device_failed && reopen_device(ao, false) < 0)
+        MP_ERR(ao, "Fatal error: *** CANNOT RE-OPEN / RESET AUDIO DEVICE ***\n");
+    if (p->audio_fd < 0) {
+        // Let playback continue normally, even with a closed device.
+        p->device_failed = true;
+        double now = mp_time_sec();
+        if (p->audio_end < now)
+            p->audio_end = now;
+        p->audio_end += samples / (double)ao->samplerate;
+        return samples;
+    }
+
     if (len > p->outburst || !(flags & AOPLAY_FINAL_CHUNK)) {
         len /= p->outburst;
         len *= p->outburst;
@@ -556,40 +547,86 @@ static int play(struct ao *ao, void **data, int samples, int flags)
     return len / ao->sstride;
 }
 
-// resume playing, after audio_pause()
-static void audio_resume(struct ao *ao)
-{
-    struct priv *p = ao->priv;
-#ifndef SNDCTL_DSP_RESET
-    reset(ao);
-#endif
-    int fillframes = get_space(ao) - p->prepause_space / ao->sstride;
-    if (fillframes > 0)
-        ao_play_silence(ao, fillframes);
-}
-
 // return: delay in seconds between first and last sample in buffer
-static float get_delay(struct ao *ao)
+static double get_delay(struct ao *ao)
 {
     struct priv *p = ao->priv;
+    if (p->audio_fd < 0) {
+        double rest = p->audio_end - mp_time_sec();
+        if (rest > 0)
+            return rest;
+        return 0;
+    }
     /* Calculate how many bytes/second is sent out */
     if (p->audio_delay_method == 2) {
 #ifdef SNDCTL_DSP_GETODELAY
         int r = 0;
         if (ioctl(p->audio_fd, SNDCTL_DSP_GETODELAY, &r) != -1)
-            return ((float)r) / (float)ao->bps;
+            return r / (double)ao->bps;
 #endif
         p->audio_delay_method = 1; // fallback if not supported
     }
     if (p->audio_delay_method == 1) {
-        // SNDCTL_DSP_GETOSPACE
-        if (ioctl(p->audio_fd, SNDCTL_DSP_GETOSPACE, &p->zz) != -1) {
-            return ((float)(p->buffersize -
-                            p->zz.bytes)) / (float)ao->bps;
+        audio_buf_info zz = {0};
+        if (ioctl(p->audio_fd, SNDCTL_DSP_GETOSPACE, &zz) != -1) {
+            return (p->buffersize - zz.bytes) / (double)ao->bps;
         }
         p->audio_delay_method = 0; // fallback if not supported
     }
-    return ((float)p->buffersize) / (float)ao->bps;
+    return p->buffersize / (double)ao->bps;
+}
+
+
+// return: how many samples can be played without blocking
+static int get_space(struct ao *ao)
+{
+    struct priv *p = ao->priv;
+
+    audio_buf_info zz = {0};
+    if (ioctl(p->audio_fd, SNDCTL_DSP_GETOSPACE, &zz) != -1) {
+        // calculate exact buffer space:
+        return zz.fragments * zz.fragsize / ao->sstride;
+    }
+
+    if (p->audio_fd < 0 && p->device_failed && get_delay(ao) > 0.2)
+        return 0;
+
+    if (p->audio_fd < 0 || device_writable(ao) > 0)
+        return p->outburst / ao->sstride;
+
+    return 0;
+}
+
+// stop playing, keep buffers (for pause)
+static void audio_pause(struct ao *ao)
+{
+    struct priv *p = ao->priv;
+    p->prepause_samples = get_delay(ao) * ao->samplerate;
+#if KEEP_DEVICE
+    ioctl(p->audio_fd, SNDCTL_DSP_RESET, NULL);
+#else
+    close_device(ao);
+#endif
+}
+
+// resume playing, after audio_pause()
+static void audio_resume(struct ao *ao)
+{
+    struct priv *p = ao->priv;
+    p->audio_end = 0;
+    if (p->prepause_samples > 0)
+        ao_play_silence(ao, p->prepause_samples);
+}
+
+static int audio_wait(struct ao *ao, pthread_mutex_t *lock)
+{
+    struct priv *p = ao->priv;
+
+    struct pollfd fd = {.fd = p->audio_fd, .events = POLLOUT};
+    int r = ao_wait_poll(ao, &fd, 1, lock);
+    if (fd.revents & (POLLERR | POLLNVAL))
+        return -1;
+    return r;
 }
 
 #define OPT_BASE_STRUCT struct priv
@@ -607,6 +644,8 @@ const struct ao_driver audio_out_oss = {
     .resume    = audio_resume,
     .reset     = reset,
     .drain     = drain,
+    .wait      = audio_wait,
+    .wakeup    = ao_wakeup_poll,
     .priv_size = sizeof(struct priv),
     .priv_defaults = &(const struct priv) {
         .audio_fd = -1,

audio/out/ao_pcm.c

@@ -35,6 +35,7 @@
 #include "ao.h"
 #include "internal.h"
 #include "common/msg.h"
+#include "osdep/endian.h"
 
 #ifdef __MINGW32__
 // for GetFileType to detect pipes
@@ -71,40 +72,35 @@ static void fput32le(uint32_t val, FILE *fp)
 
 static void write_wave_header(struct ao *ao, FILE *fp, uint64_t data_length)
 {
-    bool use_waveex = true;
-    uint16_t fmt = ao->format == AF_FORMAT_FLOAT_LE ?
-        WAV_ID_FLOAT_PCM : WAV_ID_PCM;
-    uint32_t fmt_chunk_size = use_waveex ? 40 : 16;
+    uint16_t fmt = ao->format == AF_FORMAT_FLOAT ? WAV_ID_FLOAT_PCM : WAV_ID_PCM;
     int bits = af_fmt2bits(ao->format);
 
     // Master RIFF chunk
     fput32le(WAV_ID_RIFF, fp);
-    // RIFF chunk size: 'WAVE' + 'fmt ' + 4 + fmt_chunk_size +
+    // RIFF chunk size: 'WAVE' + 'fmt ' + 4 + 40 +
     // data chunk hdr (8) + data length
-    fput32le(12 + fmt_chunk_size + 8 + data_length, fp);
+    fput32le(12 + 40 + 8 + data_length, fp);
     fput32le(WAV_ID_WAVE, fp);
 
     // Format chunk
     fput32le(WAV_ID_FMT, fp);
-    fput32le(fmt_chunk_size, fp);
-    fput16le(use_waveex ? WAV_ID_FORMAT_EXTENSIBLE : fmt, fp);
+    fput32le(40, fp);
+    fput16le(WAV_ID_FORMAT_EXTENSIBLE, fp);
     fput16le(ao->channels.num, fp);
     fput32le(ao->samplerate, fp);
     fput32le(ao->bps, fp);
     fput16le(ao->channels.num * (bits / 8), fp);
     fput16le(bits, fp);
 
-    if (use_waveex) {
-        // Extension chunk
-        fput16le(22, fp);
-        fput16le(bits, fp);
-        fput32le(mp_chmap_to_waveext(&ao->channels), fp);
-        // 2 bytes format + 14 bytes guid
-        fput32le(fmt, fp);
-        fput32le(0x00100000, fp);
-        fput32le(0xAA000080, fp);
-        fput32le(0x719B3800, fp);
-    }
+    // Extension chunk
+    fput16le(22, fp);
+    fput16le(bits, fp);
+    fput32le(mp_chmap_to_waveext(&ao->channels), fp);
+    // 2 bytes format + 14 bytes guid
+    fput32le(fmt, fp);
+    fput32le(0x00100000, fp);
+    fput32le(0xAA000080, fp);
+    fput32le(0x719B3800, fp);
 
     // Data chunk
     fput32le(WAV_ID_DATA, fp);
@@ -124,16 +120,23 @@ static int init(struct ao *ao)
     if (priv->waveheader) {
         // WAV files must have one of the following formats
 
+        // And they don't work in big endian; fixing it would be simple, but
+        // nobody cares.
+        if (BYTE_ORDER == BIG_ENDIAN) {
+            MP_FATAL(ao, "Not supported on big endian.\n");
+            return -1;
+        }
+
         switch (ao->format) {
         case AF_FORMAT_U8:
-        case AF_FORMAT_S16_LE:
-        case AF_FORMAT_S24_LE:
-        case AF_FORMAT_S32_LE:
-        case AF_FORMAT_FLOAT_LE:
-        case AF_FORMAT_AC3_LE:
+        case AF_FORMAT_S16:
+        case AF_FORMAT_S24:
+        case AF_FORMAT_S32:
+        case AF_FORMAT_FLOAT:
              break;
         default:
-            ao->format = AF_FORMAT_S16_LE;
+            if (!AF_FORMAT_IS_IEC61937(ao->format))
+                ao->format = AF_FORMAT_S16;
             break;
         }
     }
@@ -203,7 +206,7 @@ static int play(struct ao *ao, void **data, int samples, int flags)
 
     fwrite(data[0], len, 1, priv->fp);
     priv->data_length += len;
-    return len / ao->sstride;
+    return samples;
 }
 
 #define OPT_BASE_STRUCT struct priv

audio/out/ao_portaudio.c

@@ -125,6 +125,24 @@ static int validate_device_opt(struct mp_log *log, const m_option_t *opt,
     return 0;
 }
 
+static void list_devs(struct ao *ao, struct ao_device_list *list)
+{
+    if (Pa_Initialize() < 0)
+        return;
+    int count = Pa_GetDeviceCount(); // negative on error
+    for (int n = 0; n < count; n++) {
+        const PaDeviceInfo* info = Pa_GetDeviceInfo(n);
+        char desc[1024];
+        snprintf(desc, sizeof(desc), "%s, %d channels, latency: %.2f "
+                 "ms, sample rate: %.0f", info->name,
+                 info->maxOutputChannels,
+                 info->defaultHighOutputLatency * 1000,
+                 info->defaultSampleRate);
+        ao_device_list_add(list, ao, &(struct ao_device_desc){info->name, desc});
+    }
+    Pa_Terminate();
+}
+
 static int stream_callback(const void *input,
                            void *output_v,
                            unsigned long frameCount,
@@ -167,8 +185,11 @@ static int init(struct ao *ao)
         return -1;
 
     int pa_device = Pa_GetDefaultOutputDevice();
-    if (priv->cfg_device && priv->cfg_device[0])
-        pa_device = find_device(ao->log, priv->cfg_device);
+    char *dev_name = priv->cfg_device;
+    if (!dev_name || !dev_name[0])
+        dev_name = ao->device;
+    if (dev_name && dev_name[0])
+        pa_device = find_device(ao->log, dev_name);
     if (pa_device == paNoDevice)
         goto error_exit;
 
@@ -246,6 +267,7 @@ const struct ao_driver audio_out_portaudio = {
     .uninit    = uninit,
     .reset     = reset,
     .resume    = resume,
+    .list_devs = list_devs,
     .priv_size = sizeof(struct priv),
     .options = (const struct m_option[]) {
         OPT_STRING_VALIDATE("device", cfg_device, 0, validate_device_opt),

audio/out/ao_pulse.c

@@ -35,8 +35,6 @@
 #include "ao.h"
 #include "internal.h"
 
-#define PULSE_CLIENT_NAME "mpv"
-
 #define VOL_PA2MP(v) ((v) * 100 / PA_VOLUME_NORM)
 #define VOL_MP2PA(v) ((v) * PA_VOLUME_NORM / 100)
 
@@ -169,16 +167,31 @@ static const struct format_map {
     int mp_format;
     pa_sample_format_t pa_format;
 } format_maps[] = {
-    {AF_FORMAT_S16_LE, PA_SAMPLE_S16LE},
-    {AF_FORMAT_S16_BE, PA_SAMPLE_S16BE},
-    {AF_FORMAT_S32_LE, PA_SAMPLE_S32LE},
-    {AF_FORMAT_S32_BE, PA_SAMPLE_S32BE},
-    {AF_FORMAT_FLOAT_LE, PA_SAMPLE_FLOAT32LE},
-    {AF_FORMAT_FLOAT_BE, PA_SAMPLE_FLOAT32BE},
+    {AF_FORMAT_S16, PA_SAMPLE_S16NE},
+    {AF_FORMAT_S32, PA_SAMPLE_S32NE},
+    {AF_FORMAT_FLOAT, PA_SAMPLE_FLOAT32NE},
     {AF_FORMAT_U8, PA_SAMPLE_U8},
     {AF_FORMAT_UNKNOWN, 0}
 };
 
+static pa_encoding_t map_digital_format(int format)
+{
+    switch (format) {
+    case AF_FORMAT_S_AC3:   return PA_ENCODING_AC3_IEC61937;
+    case AF_FORMAT_S_EAC3:  return PA_ENCODING_EAC3_IEC61937;
+    case AF_FORMAT_S_MP3:   return PA_ENCODING_MPEG_IEC61937;
+    case AF_FORMAT_S_DTS:
+    case AF_FORMAT_S_DTSHD: return PA_ENCODING_DTS_IEC61937;
+#ifdef PA_ENCODING_MPEG2_AAC_IEC61937
+    case AF_FORMAT_S_AAC:   return PA_ENCODING_MPEG2_AAC_IEC61937;
+#endif
+    default:
+        if (AF_FORMAT_IS_IEC61937(format))
+            return PA_ENCODING_ANY;
+        return PA_ENCODING_PCM;
+    }
+}
+
 static const int speaker_map[][2] = {
   {PA_CHANNEL_POSITION_FRONT_LEFT,              MP_SPEAKER_ID_FL},
   {PA_CHANNEL_POSITION_FRONT_RIGHT,             MP_SPEAKER_ID_FR},
@@ -272,67 +285,105 @@ static void uninit(struct ao *ao)
     pthread_mutex_destroy(&priv->wakeup_lock);
 }
 
-static int init(struct ao *ao)
+static int pa_init_boilerplate(struct ao *ao)
 {
-    struct pa_sample_spec ss;
-    struct pa_channel_map map;
-    pa_proplist *proplist = NULL;
     struct priv *priv = ao->priv;
     char *host = priv->cfg_host && priv->cfg_host[0] ? priv->cfg_host : NULL;
-    char *sink = priv->cfg_sink && priv->cfg_sink[0] ? priv->cfg_sink : NULL;
+    bool locked = false;
 
     pthread_mutex_init(&priv->wakeup_lock, NULL);
     pthread_cond_init(&priv->wakeup, NULL);
 
-    ao->per_application_mixer = true;
-
     if (!(priv->mainloop = pa_threaded_mainloop_new())) {
         MP_ERR(ao, "Failed to allocate main loop\n");
         goto fail;
     }
 
+    if (pa_threaded_mainloop_start(priv->mainloop) < 0)
+        goto fail;
+
+    pa_threaded_mainloop_lock(priv->mainloop);
+    locked = true;
+
     if (!(priv->context = pa_context_new(pa_threaded_mainloop_get_api(
-                                 priv->mainloop), PULSE_CLIENT_NAME))) {
+                                         priv->mainloop), ao->client_name)))
+    {
         MP_ERR(ao, "Failed to allocate context\n");
         goto fail;
     }
 
+    MP_VERBOSE(ao, "Library version: %s\n", pa_get_library_version());
+    MP_VERBOSE(ao, "Proto: %lu\n",
+        (long)pa_context_get_protocol_version(priv->context));
+    MP_VERBOSE(ao, "Server proto: %lu\n",
+        (long)pa_context_get_server_protocol_version(priv->context));
+
     pa_context_set_state_callback(priv->context, context_state_cb, ao);
 
     if (pa_context_connect(priv->context, host, 0, NULL) < 0)
         goto fail;
 
+    /* Wait until the context is ready */
+    while (1) {
+        int state = pa_context_get_state(priv->context);
+        if (state == PA_CONTEXT_READY)
+            break;
+        if (!PA_CONTEXT_IS_GOOD(state))
+            goto fail;
+        pa_threaded_mainloop_wait(priv->mainloop);
+    }
+
+    pa_threaded_mainloop_unlock(priv->mainloop);
+    return 0;
+
+fail:
+    if (locked)
+        pa_threaded_mainloop_unlock(priv->mainloop);
+
+    if (priv->context) {
+        pa_threaded_mainloop_lock(priv->mainloop);
+        if (!(pa_context_errno(priv->context) == PA_ERR_CONNECTIONREFUSED
+              && ao->probing))
+            GENERIC_ERR_MSG("Init failed");
+        pa_threaded_mainloop_unlock(priv->mainloop);
+    }
+    uninit(ao);
+    return -1;
+}
+
+static int init(struct ao *ao)
+{
+    struct pa_channel_map map;
+    pa_proplist *proplist = NULL;
+    pa_format_info *format = NULL;
+    struct priv *priv = ao->priv;
+    char *sink = priv->cfg_sink && priv->cfg_sink[0] ? priv->cfg_sink : ao->device;
+
+    if (pa_init_boilerplate(ao) < 0)
+        return -1;
+
     pa_threaded_mainloop_lock(priv->mainloop);
 
-    if (pa_threaded_mainloop_start(priv->mainloop) < 0)
+    if (!(format = pa_format_info_new()))
         goto unlock_and_fail;
 
-    /* Wait until the context is ready */
-    pa_threaded_mainloop_wait(priv->mainloop);
-
-    if (pa_context_get_state(priv->context) != PA_CONTEXT_READY)
-        goto unlock_and_fail;
-
-    ss.channels = ao->channels.num;
-    ss.rate = ao->samplerate;
-
     ao->format = af_fmt_from_planar(ao->format);
 
-    const struct format_map *fmt_map = format_maps;
-    while (fmt_map->mp_format != ao->format) {
-        if (fmt_map->mp_format == AF_FORMAT_UNKNOWN) {
-            MP_VERBOSE(ao, "Unsupported format, using default\n");
-            fmt_map = format_maps;
-            break;
-        }
-        fmt_map++;
-    }
-    ao->format = fmt_map->mp_format;
-    ss.format = fmt_map->pa_format;
+    format->encoding = map_digital_format(ao->format);
+    if (format->encoding == PA_ENCODING_PCM) {
+        const struct format_map *fmt_map = format_maps;
 
-    if (!pa_sample_spec_valid(&ss)) {
-        MP_ERR(ao, "Invalid sample spec\n");
-        goto unlock_and_fail;
+        while (fmt_map->mp_format != ao->format) {
+            if (fmt_map->mp_format == AF_FORMAT_UNKNOWN) {
+                MP_VERBOSE(ao, "Unsupported format, using default\n");
+                fmt_map = format_maps;
+                break;
+            }
+            fmt_map++;
+        }
+        ao->format = fmt_map->mp_format;
+
+        pa_format_info_set_sample_format(format, fmt_map->pa_format);
     }
 
     if (!select_chmap(ao, &map))
@@ -343,12 +394,19 @@ static int init(struct ao *ao)
         goto unlock_and_fail;
     }
     (void)pa_proplist_sets(proplist, PA_PROP_MEDIA_ROLE, "video");
-    (void)pa_proplist_sets(proplist, PA_PROP_MEDIA_ICON_NAME,
-                           PULSE_CLIENT_NAME);
+    (void)pa_proplist_sets(proplist, PA_PROP_MEDIA_ICON_NAME, ao->client_name);
 
-    if (!(priv->stream = pa_stream_new_with_proplist(priv->context,
-                                                     "audio stream", &ss,
-                                                     &map, proplist)))
+    pa_format_info_set_rate(format, ao->samplerate);
+    pa_format_info_set_channels(format, ao->channels.num);
+    pa_format_info_set_channel_map(format, &map);
+
+    if (!pa_format_info_valid(format)) {
+        MP_ERR(ao, "Invalid audio format\n");
+        goto unlock_and_fail;
+    }
+
+    if (!(priv->stream = pa_stream_new_extended(priv->context, "audio stream",
+                                                &format, 1, proplist)))
         goto unlock_and_fail;
 
     pa_proplist_free(proplist);
@@ -358,10 +416,11 @@ static int init(struct ao *ao)
     pa_stream_set_write_callback(priv->stream, stream_request_cb, ao);
     pa_stream_set_latency_update_callback(priv->stream,
                                           stream_latency_update_cb, ao);
+    int buf_size = af_fmt_seconds_to_bytes(ao->format, priv->cfg_buffer / 1000.0,
+                                           ao->channels.num, ao->samplerate);
     pa_buffer_attr bufattr = {
         .maxlength = -1,
-        .tlength = priv->cfg_buffer > 0 ?
-            pa_usec_to_bytes(priv->cfg_buffer * 1000, &ss) : (uint32_t)-1,
+        .tlength = buf_size > 0 ? buf_size : (uint32_t)-1,
         .prebuf = -1,
         .minreq = -1,
         .fragsize = -1,
@@ -376,26 +435,28 @@ static int init(struct ao *ao)
         goto unlock_and_fail;
 
     /* Wait until the stream is ready */
-    pa_threaded_mainloop_wait(priv->mainloop);
+    while (1) {
+        int state = pa_stream_get_state(priv->stream);
+        if (state == PA_STREAM_READY)
+            break;
+        if (!PA_STREAM_IS_GOOD(state))
+            goto unlock_and_fail;
+        pa_threaded_mainloop_wait(priv->mainloop);
+    }
 
-    if (pa_stream_get_state(priv->stream) != PA_STREAM_READY)
+    if (pa_stream_is_suspended(priv->stream)) {
+        MP_ERR(ao, "The stream is suspended. Bailing out.\n");
         goto unlock_and_fail;
+    }
 
     pa_threaded_mainloop_unlock(priv->mainloop);
-
     return 0;
 
 unlock_and_fail:
+    pa_threaded_mainloop_unlock(priv->mainloop);
 
-    if (priv->mainloop)
-        pa_threaded_mainloop_unlock(priv->mainloop);
-
-fail:
-    if (priv->context) {
-        if (!(pa_context_errno(priv->context) == PA_ERR_CONNECTIONREFUSED
-              && ao->probing))
-            GENERIC_ERR_MSG("Init failed");
-    }
+    if (format)
+        pa_format_info_free(format);
 
     if (proplist)
         pa_proplist_free(proplist);
@@ -469,7 +530,7 @@ static int get_space(struct ao *ao)
     return space / ao->sstride;
 }
 
-static float get_delay_hackfixed(struct ao *ao)
+static double get_delay_hackfixed(struct ao *ao)
 {
     /* This code basically does what pa_stream_get_latency() _should_
      * do, but doesn't due to multiple known bugs in PulseAudio (at
@@ -524,7 +585,7 @@ static float get_delay_hackfixed(struct ao *ao)
     return latency / 1e6;
 }
 
-static float get_delay_pulse(struct ao *ao)
+static double get_delay_pulse(struct ao *ao)
 {
     struct priv *priv = ao->priv;
     pa_usec_t latency = (pa_usec_t) -1;
@@ -542,7 +603,7 @@ static float get_delay_pulse(struct ao *ao)
 }
 
 // Return the current latency in seconds
-static float get_delay(struct ao *ao)
+static double get_delay(struct ao *ao)
 {
     struct priv *priv = ao->priv;
     if (priv->cfg_latency_hacks) {
@@ -581,7 +642,7 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
         pa_threaded_mainloop_lock(priv->mainloop);
         if (!waitop(priv, pa_context_get_sink_input_info(priv->context, devidx,
                                                          info_func, ao))) {
-            GENERIC_ERR_MSG("pa_stream_get_sink_input_info() failed");
+            GENERIC_ERR_MSG("pa_context_get_sink_input_info() failed");
             return CONTROL_ERROR;
         }
         // Warning: some information in pi might be unaccessible, because
@@ -644,6 +705,9 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
         return CONTROL_OK;
     }
 
+    case AOCONTROL_HAS_PER_APP_VOLUME:
+        return CONTROL_TRUE;
+
     case AOCONTROL_UPDATE_STREAM_TITLE: {
         char *title = (char *)arg;
         pa_threaded_mainloop_lock(priv->mainloop);
@@ -661,6 +725,42 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
     }
 }
 
+struct sink_cb_ctx {
+    struct ao *ao;
+    struct ao_device_list *list;
+};
+
+static void sink_info_cb(pa_context *c, const pa_sink_info *i, int eol, void *ud)
+{
+    struct sink_cb_ctx *ctx = ud;
+    struct priv *priv = ctx->ao->priv;
+
+    if (eol) {
+        pa_threaded_mainloop_signal(priv->mainloop, 0); // wakeup waitop()
+        return;
+    }
+
+    struct ao_device_desc entry = {.name = i->name, .desc = i->description};
+    ao_device_list_add(ctx->list, ctx->ao, &entry);
+}
+
+static void list_devs(struct ao *ao, struct ao_device_list *list)
+{
+    struct priv *priv = ao->priv;
+    bool need_uninit = !priv->mainloop;
+    struct sink_cb_ctx ctx = {ao, list};
+
+    if (need_uninit && pa_init_boilerplate(ao) < 0)
+        return;
+
+    pa_threaded_mainloop_lock(priv->mainloop);
+    waitop(priv, pa_context_get_sink_info_list(priv->context, sink_info_cb, &ctx));
+
+    if (need_uninit)
+        uninit(ao);
+}
+
+
 #define OPT_BASE_STRUCT struct priv
 
 const struct ao_driver audio_out_pulse = {
@@ -678,6 +778,7 @@ const struct ao_driver audio_out_pulse = {
     .drain     = drain,
     .wait      = wait_audio,
     .wakeup    = wakeup,
+    .list_devs = list_devs,
     .priv_size = sizeof(struct priv),
     .priv_defaults = &(const struct priv) {
         .cfg_buffer = 250,
@@ -686,7 +787,7 @@ const struct ao_driver audio_out_pulse = {
     .options = (const struct m_option[]) {
         OPT_STRING("host", cfg_host, 0),
         OPT_STRING("sink", cfg_sink, 0),
-        OPT_CHOICE_OR_INT("buffer", cfg_buffer, 0, 1, 2000, ({"native", -1})),
+        OPT_CHOICE_OR_INT("buffer", cfg_buffer, 0, 1, 2000, ({"native", 0})),
         OPT_FLAG("latency-hacks", cfg_latency_hacks, 0),
         {0}
     },

audio/out/ao_rsound.c

@@ -52,40 +52,26 @@ static int set_format(struct ao *ao)
     case AF_FORMAT_S8:
         rsd_format = RSD_S8;
         break;
-    case AF_FORMAT_S16_LE:
-        rsd_format = RSD_S16_LE;
+    case AF_FORMAT_S16:
+        rsd_format = RSD_S16_NE;
         break;
-    case AF_FORMAT_S16_BE:
-        rsd_format = RSD_S16_BE;
+    case AF_FORMAT_U16:
+        rsd_format = RSD_U16_NE;
         break;
-    case AF_FORMAT_U16_LE:
-        rsd_format = RSD_U16_LE;
+    case AF_FORMAT_S24:
+    case AF_FORMAT_U24:
+        rsd_format = RSD_S32_NE;
+        ao->format = AF_FORMAT_S32;
         break;
-    case AF_FORMAT_U16_BE:
-        rsd_format = RSD_U16_BE;
+    case AF_FORMAT_S32:
+        rsd_format = RSD_S32_NE;
         break;
-    case AF_FORMAT_S24_LE:
-    case AF_FORMAT_S24_BE:
-    case AF_FORMAT_U24_LE:
-    case AF_FORMAT_U24_BE:
-        rsd_format = RSD_S32_LE;
-        ao->format = AF_FORMAT_S32_LE;
-        break;
-    case AF_FORMAT_S32_LE:
-        rsd_format = RSD_S32_LE;
-        break;
-    case AF_FORMAT_S32_BE:
-        rsd_format = RSD_S32_BE;
-        break;
-    case AF_FORMAT_U32_LE:
-        rsd_format = RSD_U32_LE;
-        break;
-    case AF_FORMAT_U32_BE:
-        rsd_format = RSD_U32_BE;
+    case AF_FORMAT_U32:
+        rsd_format = RSD_U32_NE;
         break;
     default:
-        rsd_format = RSD_S16_LE;
-        ao->format = AF_FORMAT_S16_LE;
+        rsd_format = RSD_S16_NE;
+        ao->format = AF_FORMAT_S16;
     }
 
     return rsd_format;
@@ -167,7 +153,7 @@ static int play(struct ao *ao, void **data, int samples, int flags)
     return rsd_write(priv->rd, data[0], samples * ao->sstride) / ao->sstride;
 }
 
-static float get_delay(struct ao *ao)
+static double get_delay(struct ao *ao)
 {
     struct priv *priv = ao->priv;
     return rsd_delay_ms(priv->rd) / 1000.0;

audio/out/ao_sdl.c

@@ -38,6 +38,20 @@ struct priv
     float buflen;
 };
 
+static const int fmtmap[][2] = {
+    {AF_FORMAT_U8,      AUDIO_U8},
+    {AF_FORMAT_S8,      AUDIO_S8},
+    {AF_FORMAT_U16,     AUDIO_U16SYS},
+    {AF_FORMAT_S16,     AUDIO_S16SYS},
+#ifdef AUDIO_S32SYS
+    {AF_FORMAT_S32,     AUDIO_S32SYS},
+#endif
+#ifdef AUDIO_F32SYS
+    {AF_FORMAT_FLOAT,   AUDIO_F32SYS},
+#endif
+    {0}
+};
+
 static void audio_callback(void *userdata, Uint8 *stream, int len)
 {
     struct ao *ao = userdata;
@@ -104,26 +118,12 @@ static int init(struct ao *ao)
 
     SDL_AudioSpec desired, obtained;
 
-    switch (ao->format) {
-        case AF_FORMAT_U8: desired.format = AUDIO_U8; break;
-        case AF_FORMAT_S8: desired.format = AUDIO_S8; break;
-        case AF_FORMAT_U16_LE: desired.format = AUDIO_U16LSB; break;
-        case AF_FORMAT_U16_BE: desired.format = AUDIO_U16MSB; break;
-        default:
-        case AF_FORMAT_S16_LE: desired.format = AUDIO_S16LSB; break;
-        case AF_FORMAT_S16_BE: desired.format = AUDIO_S16MSB; break;
-#ifdef AUDIO_S32LSB
-        case AF_FORMAT_S32_LE: desired.format = AUDIO_S32LSB; break;
-#endif
-#ifdef AUDIO_S32MSB
-        case AF_FORMAT_S32_BE: desired.format = AUDIO_S32MSB; break;
-#endif
-#ifdef AUDIO_F32LSB
-        case AF_FORMAT_FLOAT_LE: desired.format = AUDIO_F32LSB; break;
-#endif
-#ifdef AUDIO_F32MSB
-        case AF_FORMAT_FLOAT_BE: desired.format = AUDIO_F32MSB; break;
-#endif
+    desired.format = AUDIO_S16SYS;
+    for (int n = 0; fmtmap[n][0]; n++) {
+        if (ao->format == fmtmap[n][0]) {
+            desired.format = fmtmap[n][1];
+            break;
+        }
     }
     desired.freq = ao->samplerate;
     desired.channels = ao->channels.num;
@@ -156,30 +156,18 @@ static int init(struct ao *ao)
     // large, this will help.
     ao->device_buffer = 3 * obtained.samples;
 
-    switch (obtained.format) {
-        case AUDIO_U8: ao->format = AF_FORMAT_U8; break;
-        case AUDIO_S8: ao->format = AF_FORMAT_S8; break;
-        case AUDIO_S16LSB: ao->format = AF_FORMAT_S16_LE; break;
-        case AUDIO_S16MSB: ao->format = AF_FORMAT_S16_BE; break;
-        case AUDIO_U16LSB: ao->format = AF_FORMAT_U16_LE; break;
-        case AUDIO_U16MSB: ao->format = AF_FORMAT_U16_BE; break;
-#ifdef AUDIO_S32LSB
-        case AUDIO_S32LSB: ao->format = AF_FORMAT_S32_LE; break;
-#endif
-#ifdef AUDIO_S32MSB
-        case AUDIO_S32MSB: ao->format = AF_FORMAT_S32_BE; break;
-#endif
-#ifdef AUDIO_F32LSB
-        case AUDIO_F32LSB: ao->format = AF_FORMAT_FLOAT_LE; break;
-#endif
-#ifdef AUDIO_F32MSB
-        case AUDIO_F32MSB: ao->format = AF_FORMAT_FLOAT_BE; break;
-#endif
-        default:
-            if (!ao->probing)
-                MP_ERR(ao, "could not find matching format\n");
-            uninit(ao);
-            return -1;
+    ao->format = 0;
+    for (int n = 0; fmtmap[n][0]; n++) {
+        if (obtained.format == fmtmap[n][1]) {
+            ao->format = fmtmap[n][0];
+            break;
+        }
+    }
+    if (!ao->format) {
+        if (!ao->probing)
+            MP_ERR(ao, "could not find matching format\n");
+        uninit(ao);
+        return -1;
     }
 
     if (!ao_chmap_sel_get_def(ao, &sel, &ao->channels, obtained.channels)) {

audio/out/ao_sndio.c

@@ -33,6 +33,7 @@ struct priv {
     struct sio_hdl *hdl;
     struct sio_par par;
     int delay;
+    bool playing;
     int vol;
     int havevol;
 #define SILENCE_NMAX 0x1000
@@ -60,6 +61,8 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
             return CONTROL_FALSE;
         sio_setvol(p->hdl, vol->left * SIO_MAXVOL / 100);
         break;
+    case AOCONTROL_HAS_SOFT_VOLUME:
+        return CONTROL_TRUE;
     default:
         return CONTROL_UNKNOWN;
     }
@@ -72,7 +75,7 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
 static void movecb(void *addr, int delta)
 {
     struct priv *p = addr;
-    p->delay -= delta * (int)(p->par.bps * p->par.pchan);
+    p->delay -= delta;
 }
 
 /*
@@ -107,26 +110,22 @@ static int init(struct ao *ao)
     struct priv *p = ao->priv;
 
     struct af_to_par {
-        int format, bits, sig, le;
-    } static const af_to_par[] = {
-        {AF_FORMAT_U8,      8, 0, 0},
-        {AF_FORMAT_S8,      8, 1, 0},
-        {AF_FORMAT_U16_LE, 16, 0, 1},
-        {AF_FORMAT_U16_BE, 16, 0, 0},
-        {AF_FORMAT_S16_LE, 16, 1, 1},
-        {AF_FORMAT_S16_BE, 16, 1, 0},
-        {AF_FORMAT_U24_LE, 16, 0, 1},
-        {AF_FORMAT_U24_BE, 24, 0, 0},
-        {AF_FORMAT_S24_LE, 24, 1, 1},
-        {AF_FORMAT_S24_BE, 24, 1, 0},
-        {AF_FORMAT_U32_LE, 32, 0, 1},
-        {AF_FORMAT_U32_BE, 32, 0, 0},
-        {AF_FORMAT_S32_LE, 32, 1, 1},
-        {AF_FORMAT_S32_BE, 32, 1, 0}
-    }, *ap;
+        int format, bits, sig;
+    };
+    static const struct af_to_par af_to_par[] = {
+        {AF_FORMAT_U8,   8, 0},
+        {AF_FORMAT_S8,   8, 1},
+        {AF_FORMAT_U16, 16, 0},
+        {AF_FORMAT_S16, 16, 1},
+        {AF_FORMAT_U24, 24, 0},
+        {AF_FORMAT_S24, 24, 1},
+        {AF_FORMAT_U32, 32, 0},
+        {AF_FORMAT_S32, 32, 1},
+    };
+    const struct af_to_par *ap;
     int i;
 
-    p->hdl = sio_open(p->dev, SIO_PLAY, 0);
+    p->hdl = sio_open(p->dev, SIO_PLAY, 1);
     if (p->hdl == NULL) {
         MP_ERR(ao, "can't open sndio %s\n", p->dev);
         goto error;
@@ -147,7 +146,7 @@ static int init(struct ao *ao)
             p->par.bits = ap->bits;
             p->par.sig = ap->sig;
             if (ap->bits > 8)
-                p->par.le = ap->le;
+                p->par.le = SIO_LE_NATIVE;
             if (ap->bits != SIO_BPS(ap->bits))
                 p->par.bps = ap->bits / 8;
             break;
@@ -173,30 +172,25 @@ static int init(struct ao *ao)
         MP_ERR(ao, "couldn't get params\n");
         goto error;
     }
+    if (p->par.bps > 1 && p->par.le != SIO_LE_NATIVE) {
+        MP_ERR(ao, "swapped endian output not supported\n");
+        goto error;
+    }
     if (p->par.bits == 8 && p->par.bps == 1) {
         ao->format = p->par.sig ? AF_FORMAT_S8 : AF_FORMAT_U8;
     } else if (p->par.bits == 16 && p->par.bps == 2) {
-        ao->format = p->par.sig ?
-            (p->par.le ? AF_FORMAT_S16_LE : AF_FORMAT_S16_BE) :
-            (p->par.le ? AF_FORMAT_U16_LE : AF_FORMAT_U16_BE);
+        ao->format = p->par.sig ? AF_FORMAT_S16 : AF_FORMAT_U16;
     } else if ((p->par.bits == 24 || p->par.msb) && p->par.bps == 3) {
-        ao->format = p->par.sig ?
-            (p->par.le ? AF_FORMAT_S24_LE : AF_FORMAT_S24_BE) :
-            (p->par.le ? AF_FORMAT_U24_LE : AF_FORMAT_U24_BE);
+        ao->format = p->par.sig ? AF_FORMAT_S24 : AF_FORMAT_U24;
     } else if ((p->par.bits == 32 || p->par.msb) && p->par.bps == 4) {
-        ao->format = p->par.sig ?
-            (p->par.le ? AF_FORMAT_S32_LE : AF_FORMAT_S32_BE) :
-            (p->par.le ? AF_FORMAT_U32_LE : AF_FORMAT_U32_BE);
+        ao->format = p->par.sig ? AF_FORMAT_S32 : AF_FORMAT_U32;
     } else {
         MP_ERR(ao, "couldn't set format\n");
         goto error;
     }
 
-    ao->bps = p->par.bps * p->par.pchan * p->par.rate;
-    ao->no_persistent_volume = true;
     p->havevol = sio_onvol(p->hdl, volcb, p);
     sio_onmove(p->hdl, movecb, p);
-    p->delay = 0;
     if (!sio_start(p->hdl))
         MP_ERR(ao, "init: couldn't start\n");
 
@@ -233,55 +227,68 @@ static void reset(struct ao *ao)
 {
     struct priv *p = ao->priv;
 
-    if (!sio_stop(p->hdl))
-        MP_ERR(ao, "reset: couldn't stop\n");
-    p->delay = 0;
-    if (!sio_start(p->hdl))
-        MP_ERR(ao, "reset: couldn't start\n");
+    if (p->playing) {
+        MP_WARN(ao, "Blocking until remaining audio is played... (sndio design bug).\n");
+
+        p->playing = false;
+
+        if (!sio_stop(p->hdl))
+            MP_ERR(ao, "reset: couldn't stop\n");
+        p->delay = 0;
+        if (!sio_start(p->hdl))
+            MP_ERR(ao, "reset: couldn't start\n");
+    }
 }
 
 /*
- * play given number of bytes until sio_write() blocks
+ * play given number of samples until sio_write() blocks
  */
 static int play(struct ao *ao, void **data, int samples, int flags)
 {
     struct priv *p = ao->priv;
     int n;
 
-    n = sio_write(p->hdl, data[0], samples * ao->sstride);
+    n = sio_write(p->hdl, data[0], samples * ao->sstride) / ao->sstride;
     p->delay += n;
-    if (flags & AOPLAY_FINAL_CHUNK)
-        reset(ao);
-    return n / ao->sstride;
+    p->playing = true;
+    /* on AOPLAY_FINAL_CHUNK, just let it underrun */
+    return n;
 }
 
 /*
- * how many bytes can be played without blocking
+ * make libsndio call movecb()
+ */
+static void update(struct ao *ao)
+{
+    struct priv *p = ao->priv;
+    int n = sio_pollfd(p->hdl, p->pfd, POLLOUT);
+    while (poll(p->pfd, n, 0) < 0 && errno == EINTR) {}
+    sio_revents(p->hdl, p->pfd);
+}
+
+/*
+ * how many samples can be played without blocking
  */
 static int get_space(struct ao *ao)
 {
     struct priv *p = ao->priv;
-    int n;
 
-    /*
-     * call poll() and sio_revents(), so the
-     * delay counter is updated
-     */
-    n = sio_pollfd(p->hdl, p->pfd, POLLOUT);
-    while (poll(p->pfd, n, 0) < 0 && errno == EINTR)
-        ; /* nothing */
-    sio_revents(p->hdl, p->pfd);
+    update(ao);
 
-    return (p->par.bufsz * p->par.pchan * p->par.bps - p->delay) / ao->sstride;
+    int samples = p->par.bufsz - p->delay;
+    return samples / p->par.round * p->par.round;
 }
 
 /*
  * return: delay in seconds between first and last sample in buffer
  */
-static float get_delay(struct ao *ao)
+static double get_delay(struct ao *ao)
 {
     struct priv *p = ao->priv;
-    return (float)p->delay / (p->par.bps * p->par.pchan * p->par.rate);
+
+    update(ao);
+
+    return p->delay / (double)p->par.rate;
 }
 
 /*
@@ -301,8 +308,8 @@ static void audio_resume(struct ao *ao)
     int n, count, todo;
 
     /*
-     * we want to start with buffers full, because mplayer uses
-     * get_space() pointer as clock, which would cause video to
+     * we want to start with buffers full, because mpv uses
+     * get_delay() as clock, which would cause video to
      * accelerate while buffers are filled.
      */
     todo = p->par.bufsz * p->par.pchan * p->par.bps;

audio/out/ao_wasapi.c

@@ -36,11 +36,6 @@
 #include "osdep/timer.h"
 #include "osdep/io.h"
 
-#define EXIT_ON_ERROR(hres)  \
-              do { if (FAILED(hres)) { goto exit_label; } } while(0)
-#define SAFE_RELEASE(unk, release) \
-              do { if ((unk) != NULL) { release; (unk) = NULL; } } while(0)
-
 static double get_device_delay(struct wasapi_state *state) {
     UINT64 sample_count = atomic_load(&state->sample_count);
     UINT64 position, qpc_position;
@@ -50,7 +45,8 @@ static double get_device_delay(struct wasapi_state *state) {
         case S_OK: case S_FALSE:
             break;
         default:
-            MP_ERR(state, "IAudioClock::GetPosition returned %s\n", wasapi_explain_err(hr));
+            MP_ERR(state, "IAudioClock::GetPosition returned %s (0x%"PRIx32")\n",
+                   wasapi_explain_err(hr), (uint32_t)hr);
     }
 
     LARGE_INTEGER qpc_count;
@@ -65,7 +61,7 @@ static double get_device_delay(struct wasapi_state *state) {
     double diff = sample_count - position;
     double delay = diff / state->format.Format.nSamplesPerSec;
 
-    MP_TRACE(state, "device delay: %g samples (%g ms)\n", diff, delay * 1000);
+    MP_TRACE(state, "Device delay: %g samples (%g ms)\n", diff, delay * 1000);
 
     return delay;
 }
@@ -83,6 +79,7 @@ static void thread_feed(struct ao *ao)
         EXIT_ON_ERROR(hr);
 
         frame_count -= padding;
+        MP_TRACE(ao, "Frame to fill: %"PRIu32". Padding: %"PRIu32"\n", frame_count, padding);
     }
 
     BYTE *pData;
@@ -103,7 +100,10 @@ static void thread_feed(struct ao *ao)
 
     return;
 exit_label:
-    MP_ERR(state, "thread_feed fails with %"PRIx32"!\n", (uint32_t)hr);
+    MP_ERR(state, "Error feeding audio: %s (0x%"PRIx32")\n",
+           wasapi_explain_err(hr), (uint32_t)hr);
+    MP_VERBOSE(ao, "Requesting ao reload\n");
+    ao_request_reload(ao);
     return;
 }
 
@@ -113,20 +113,27 @@ static DWORD __stdcall ThreadLoop(void *lpParameter)
     if (!ao || !ao->priv)
         return -1;
     struct wasapi_state *state = (struct wasapi_state *)ao->priv;
-    if (wasapi_thread_init(ao))
-        goto exit_label;
+    int thread_ret;
+    CoInitializeEx(NULL, COINIT_APARTMENTTHREADED);
 
-    MSG msg;
-    DWORD waitstatus = WAIT_FAILED;
+    state->init_ret = wasapi_thread_init(ao);
+    SetEvent(state->init_done);
+    if (state->init_ret != S_OK) {
+        thread_ret = -1;
+        goto exit_label;
+    }
+
+
+    DWORD waitstatus;
     HANDLE playcontrol[] =
         {state->hUninit, state->hFeed, state->hForceFeed, NULL};
-    MP_VERBOSE(ao, "Entering dispatch loop!\n");
+    MP_DBG(ao, "Entering dispatch loop\n");
     while (1) { /* watch events */
         waitstatus = MsgWaitForMultipleObjects(3, playcontrol, FALSE, INFINITE,
                                                QS_POSTMESSAGE | QS_SENDMESSAGE);
         switch (waitstatus) {
         case WAIT_OBJECT_0: /*shutdown*/
-            wasapi_thread_uninit(state);
+            thread_ret = 0;
             goto exit_label;
         case (WAIT_OBJECT_0 + 1): /* feed */
             thread_feed(ao);
@@ -136,51 +143,52 @@ static DWORD __stdcall ThreadLoop(void *lpParameter)
             SetEvent(state->hFeedDone);
             break;
         case (WAIT_OBJECT_0 + 3): /* messages to dispatch (COM marshalling) */
-            while (PeekMessage(&msg, NULL, 0, 0, PM_REMOVE)) {
-                DispatchMessage(&msg);
-            }
+            wasapi_dispatch();
             break;
-        case WAIT_FAILED: /* ??? */
-            return -1;
+        default:
+            MP_ERR(ao, "Unhandled case in thread loop");
+            thread_ret = -1;
+            goto exit_label;
         }
     }
 exit_label:
-    /* dispatch any possible pending messages */
-    while (PeekMessage(&msg, NULL, 0, 0, PM_REMOVE)) {
-        DispatchMessage(&msg);
-    }
-    return state->init_ret;
+    wasapi_thread_uninit(ao);
+
+    CoUninitialize();
+    return thread_ret;
 }
 
 static void closehandles(struct ao *ao)
 {
     struct wasapi_state *state = (struct wasapi_state *)ao->priv;
-    if (state->init_done)
-        CloseHandle(state->init_done);
-    if (state->hUninit)
-        CloseHandle(state->hUninit);
-    if (state->hFeed)
-        CloseHandle(state->hFeed);
+    if (state->init_done)  CloseHandle(state->init_done);
+    if (state->hUninit)    CloseHandle(state->hUninit);
+    if (state->hFeed)      CloseHandle(state->hFeed);
+    if (state->hForceFeed) CloseHandle(state->hForceFeed);
+    if (state->hFeedDone)  CloseHandle(state->hFeedDone);
+    if (state->threadLoop) CloseHandle(state->threadLoop);
 }
 
 static void uninit(struct ao *ao)
 {
-    MP_VERBOSE(ao, "uninit!\n");
+    MP_DBG(ao, "Uninit wasapi\n");
     struct wasapi_state *state = (struct wasapi_state *)ao->priv;
     wasapi_release_proxies(state);
     SetEvent(state->hUninit);
     /* wait up to 10 seconds */
-    if (WaitForSingleObject(state->threadLoop, 10000) == WAIT_TIMEOUT)
-        MP_ERR(ao, "audio loop thread refuses to abort!");
+    if (WaitForSingleObject(state->threadLoop, 10000) == WAIT_TIMEOUT){
+        MP_ERR(ao, "Audio loop thread refuses to abort");
+        return;
+    }
     if (state->VistaBlob.hAvrt)
         FreeLibrary(state->VistaBlob.hAvrt);
     closehandles(ao);
-    MP_VERBOSE(ao, "uninit END!\n");
+    MP_DBG(ao, "Uninit wasapi done\n");
 }
 
 static int init(struct ao *ao)
 {
-    MP_VERBOSE(ao, "init!\n");
+    MP_DBG(ao, "Init wasapi\n");
     ao->format = af_fmt_from_planar(ao->format);
     struct mp_chmap_sel sel = {0};
     mp_chmap_sel_add_waveext(&sel);
@@ -191,10 +199,9 @@ static int init(struct ao *ao)
     wasapi_fill_VistaBlob(state);
 
     if (state->opt_list) {
-        wasapi_enumerate_devices(state->log);
+        wasapi_enumerate_devices(state->log, NULL, NULL);
     }
 
-    ao->per_application_mixer = true;
     if (state->opt_exclusive) {
         state->share_mode = AUDCLNT_SHAREMODE_EXCLUSIVE;
     } else {
@@ -213,23 +220,26 @@ static int init(struct ao *ao)
         /* failed to init events */
         return -1;
     }
-    state->init_ret = -1;
+
+    state->init_ret = E_FAIL;
     state->threadLoop = (HANDLE)CreateThread(NULL, 0, &ThreadLoop, ao, 0, NULL);
     if (!state->threadLoop) {
         /* failed to init thread */
-        MP_ERR(ao, "fail to create thread!\n");
+        MP_ERR(ao, "Failed to create thread\n");
         return -1;
     }
+
     WaitForSingleObject(state->init_done, INFINITE); /* wait on init complete */
-    if (state->init_ret) {
-        if (!ao->probing) {
-            MP_ERR(ao, "thread_init failed!\n");
-        }
-    } else
-        MP_VERBOSE(ao, "Init Done!\n");
+    if (state->init_ret != S_OK) {
+        if (!ao->probing)
+            MP_ERR(ao, "Received failure from audio thread\n");
+        uninit(ao);
+        return -1;
+    }
 
     wasapi_setup_proxies(state);
-    return state->init_ret;
+    MP_DBG(ao, "Init wasapi done\n");
+    return 0;
 }
 
 static int control(struct ao *ao, enum aocontrol cmd, void *arg)
@@ -250,7 +260,7 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
         /* check to see if user manually changed volume through mixer;
            this information is used in exclusive mode for restoring the mixer volume on uninit */
         if (state->audio_volume != state->previous_volume) {
-            MP_VERBOSE(state, "mixer difference: %.2g now, expected %.2g\n",
+            MP_VERBOSE(state, "Mixer difference: %.2g now, expected %.2g\n",
                        state->audio_volume, state->previous_volume);
             state->initial_volume = state->audio_volume;
         }
@@ -284,6 +294,8 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
             ISimpleAudioVolume_SetMute(state->pAudioVolumeProxy, mute, NULL);
 
         return CONTROL_OK;
+    case AOCONTROL_HAS_PER_APP_VOLUME:
+        return CONTROL_TRUE;
     case AOCONTROL_UPDATE_STREAM_TITLE: {
         MP_VERBOSE(state, "Updating stream title to \"%s\"\n", (char*)arg);
         wchar_t *title = mp_from_utf8(NULL, (char*)arg);
@@ -328,6 +340,11 @@ static void audio_resume(struct ao *ao)
     IAudioClient_Start(state->pAudioClientProxy);
 }
 
+static void list_devs(struct ao *ao, struct ao_device_list *list)
+{
+    wasapi_enumerate_devices(mp_null_log, ao, list);
+}
+
 #define OPT_BASE_STRUCT struct wasapi_state
 
 const struct ao_driver audio_out_wasapi = {
@@ -336,8 +353,9 @@ const struct ao_driver audio_out_wasapi = {
     .init      = init,
     .uninit    = uninit,
     .control   = control,
-    //.reset     = audio_reset, <- doesn't wait for audio callback to return
+    .reset     = audio_reset,
     .resume    = audio_resume,
+    .list_devs = list_devs,
     .priv_size = sizeof(wasapi_state),
     .options   = (const struct m_option[]) {
         OPT_FLAG("exclusive", opt_exclusive, 0),

audio/out/ao_wasapi.h

@@ -28,14 +28,28 @@
 #include <mmdeviceapi.h>
 #include <avrt.h>
 
-#include "compat/atomics.h"
+#include "osdep/atomics.h"
+
+typedef struct change_notify {
+    IMMNotificationClient client; /* this must be first in the structure! */
+    LPWSTR monitored; /* Monitored device */
+    struct ao *ao;
+} change_notify;
+
+HRESULT wasapi_change_init(struct ao* ao);
+void wasapi_change_uninit(struct ao* ao);
+
+#define EXIT_ON_ERROR(hres)  \
+              do { if (FAILED(hres)) { goto exit_label; } } while(0)
+#define SAFE_RELEASE(unk, release) \
+              do { if ((unk) != NULL) { release; (unk) = NULL; } } while(0)
 
 typedef struct wasapi_state {
     struct mp_log *log;
     HANDLE threadLoop;
 
     /* Init phase */
-    int init_ret;
+    HRESULT init_ret;
     HANDLE init_done;
     int share_mode;
 
@@ -49,10 +63,6 @@ typedef struct wasapi_state {
 
     /* Buffers */
     size_t buffer_block_size; /* Size of each block in bytes */
-    REFERENCE_TIME
-        minRequestedDuration; /* minimum wasapi buffer block size, in 100-nanosecond units */
-    REFERENCE_TIME
-        defaultRequestedDuration; /* default wasapi default block size, in 100-nanosecond units */
     UINT32 bufferFrameCount; /* wasapi buffer block size, number of frames, frame size at format.nBlockAlign */
 
     /* WASAPI handles, owned by other thread */
@@ -62,6 +72,8 @@ typedef struct wasapi_state {
     ISimpleAudioVolume *pAudioVolume;
     IAudioEndpointVolume *pEndpointVolume;
     IAudioSessionControl *pSessionControl;
+    IMMDeviceEnumerator *pEnumerator;
+
     HANDLE hFeed; /* wasapi event */
     HANDLE hForceFeed; /* forces writing a buffer (e.g. before audio_resume) */
     HANDLE hFeedDone; /* set only after a hForceFeed */
@@ -100,6 +112,8 @@ typedef struct wasapi_state {
         HANDLE (WINAPI *pAvSetMmThreadCharacteristicsW)(LPCWSTR, LPDWORD);
         WINBOOL (WINAPI *pAvRevertMmThreadCharacteristics)(HANDLE);
     } VistaBlob;
+
+    change_notify change;
 } wasapi_state;
 
 #endif

audio/out/ao_wasapi_changenotify.c

@@ -0,0 +1,284 @@
+/*
+ * This file is part of mpv.
+ *
+ * Original author: Jonathan Yong <10walls@gmail.com>
+ *
+ * mpv is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * mpv is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with mpv.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#define COBJMACROS 1
+#define _WIN32_WINNT 0x600
+
+#include <initguid.h>
+#include <audioclient.h>
+#include <endpointvolume.h>
+#include <mmdeviceapi.h>
+#include <wchar.h>
+#include <stdlib.h>
+
+#include "ao_wasapi.h"
+#include "ao_wasapi_utils.h"
+
+static int GUID_compare(const GUID *l, const GUID *r)
+{
+    unsigned int i;
+    if (l->Data1 != r->Data1) return 1;
+    if (l->Data2 != r->Data2) return 1;
+    if (l->Data3 != r->Data3) return 1;
+    for (i = 0; i < 8; i++) {
+        if (l->Data4[i] != r->Data4[i]) return 1;
+    }
+    return 0;
+}
+
+static int PKEY_compare(const PROPERTYKEY *l, const PROPERTYKEY *r)
+{
+    if (GUID_compare(&l->fmtid, &r->fmtid)) return 1;
+    if (l->pid != r->pid) return 1;
+    return 0;
+}
+
+static char *GUID_to_str_buf(char *buf, size_t buf_size, const GUID *guid)
+{
+    snprintf(buf, buf_size,
+             "{%8.8x-%4.4x-%4.4x-%2.2x%2.2x-%2.2x%2.2x%2.2x%2.2x%2.2x%2.2x}",
+             (unsigned) guid->Data1, guid->Data2, guid->Data3,
+             guid->Data4[0], guid->Data4[1],
+             guid->Data4[2], guid->Data4[3],
+             guid->Data4[4], guid->Data4[5],
+             guid->Data4[6], guid->Data4[7]);
+    return buf;
+}
+
+static char *PKEY_to_str_buf(char *buf, size_t buf_size, const PROPERTYKEY *pkey)
+{
+    buf = GUID_to_str_buf(buf, buf_size, &pkey->fmtid);
+    size_t guid_len = strnlen(buf, buf_size);
+    snprintf(buf + guid_len, buf_size - guid_len, ",%"PRIu32, (uint32_t)pkey->pid );
+    return buf;
+}
+
+#define PKEY_to_str(pkey) PKEY_to_str_buf((char[42]){0}, 42, (pkey))
+
+static char* ERole_to_str(ERole role)
+{
+    switch(role){
+    case eConsole:        return "console";
+    case eMultimedia:     return "multimedia";
+    case eCommunications: return "communications";
+    default:              return "<Unknown>";
+    }
+}
+
+static char* EDataFlow_to_str(EDataFlow flow)
+{
+    switch(flow){
+    case eRender:  return "render";
+    case eCapture: return "capture";
+    case eAll:     return "all";
+    default:       return "<Unknown>";
+    }
+}
+
+static HRESULT STDMETHODCALLTYPE sIMMNotificationClient_QueryInterface(
+    IMMNotificationClient* This, REFIID riid, void **ppvObject)
+{
+    /* Compatible with IMMNotificationClient and IUnknown */
+    if (!GUID_compare(&IID_IMMNotificationClient, riid) ||
+        !GUID_compare(&IID_IUnknown, riid)) {
+        *ppvObject = (void *)This;
+        return S_OK;
+    } else {
+        *ppvObject = NULL;
+        return E_NOINTERFACE;
+    }
+}
+
+/* these are required, but not actually used */
+static ULONG STDMETHODCALLTYPE sIMMNotificationClient_AddRef(
+    IMMNotificationClient *This)
+{
+    return 1;
+}
+
+/* MSDN says it should free itself, but we're static */
+static ULONG STDMETHODCALLTYPE sIMMNotificationClient_Release(
+    IMMNotificationClient *This)
+{
+    return 1;
+}
+
+static HRESULT STDMETHODCALLTYPE sIMMNotificationClient_OnDeviceStateChanged(
+    IMMNotificationClient *This,
+    LPCWSTR pwstrDeviceId,
+    DWORD dwNewState)
+{
+    change_notify *change = (change_notify *)This;
+    struct ao *ao = change->ao;
+
+    if (pwstrDeviceId && !wcscmp(change->monitored, pwstrDeviceId)){
+        switch (dwNewState) {
+        case DEVICE_STATE_DISABLED:
+        case DEVICE_STATE_NOTPRESENT:
+        case DEVICE_STATE_UNPLUGGED:
+            MP_VERBOSE(ao,
+                       "OnDeviceStateChange triggered - requesting ao reload\n");
+            ao_request_reload(ao);
+        case DEVICE_STATE_ACTIVE:
+        default:
+            return S_OK;
+        }
+    }
+    return S_OK;
+}
+
+static HRESULT STDMETHODCALLTYPE sIMMNotificationClient_OnDeviceAdded(
+    IMMNotificationClient *This,
+    LPCWSTR pwstrDeviceId)
+{
+    change_notify *change = (change_notify *)This;
+    struct ao *ao = change->ao;
+
+    MP_VERBOSE(ao, "OnDeviceAdded triggered\n");
+    if(pwstrDeviceId)
+        MP_VERBOSE(ao, "New device %S\n",pwstrDeviceId);
+    return S_OK;
+}
+
+/* maybe MPV can go over to the prefered device once it is plugged in? */
+static HRESULT STDMETHODCALLTYPE sIMMNotificationClient_OnDeviceRemoved(
+    IMMNotificationClient *This,
+    LPCWSTR pwstrDeviceId)
+{
+    change_notify *change = (change_notify *)This;
+    struct ao *ao = change->ao;
+
+    if (pwstrDeviceId && !wcscmp(change->monitored, pwstrDeviceId)) {
+        MP_VERBOSE(ao, "OnDeviceRemoved triggered - requesting ao reload\n");
+        ao_request_reload(ao);
+    }
+    return S_OK;
+}
+
+static HRESULT STDMETHODCALLTYPE sIMMNotificationClient_OnDefaultDeviceChanged(
+    IMMNotificationClient *This,
+    EDataFlow flow,
+    ERole role,
+    LPCWSTR pwstrDeviceId)
+{
+    change_notify *change = (change_notify *)This;
+    struct ao *ao = change->ao;
+    struct wasapi_state *state = (struct wasapi_state *)ao->priv;
+
+    MP_VERBOSE(ao, "OnDefaultDeviceChanged triggered for role:%s flow:%s\n",
+               ERole_to_str(role), EDataFlow_to_str(flow));
+    if(pwstrDeviceId)
+        MP_VERBOSE(ao, "New default device %S\n", pwstrDeviceId);
+
+    /* don't care about "eCapture" or non-"eMultimedia" roles  */
+    if ( flow == eCapture ||
+         role != eMultimedia ) return S_OK;
+
+    /* stay on the device the user specified */
+    if (state->opt_device) {
+        MP_VERBOSE(ao, "Staying on specified device \"%s\"", state->opt_device);
+        return S_OK;
+    }
+
+    /* don't reload if already on the new default */
+    if ( pwstrDeviceId && !wcscmp(change->monitored, pwstrDeviceId) ){
+        MP_VERBOSE(ao, "Already using default device, no reload required\n");
+        return S_OK;
+    }
+
+    /* if we got here, we need to reload */
+    ao_request_reload(ao);
+    MP_VERBOSE(ao, "Requesting ao reload\n");
+    return S_OK;
+}
+
+static HRESULT STDMETHODCALLTYPE sIMMNotificationClient_OnPropertyValueChanged(
+    IMMNotificationClient *This,
+    LPCWSTR pwstrDeviceId,
+    const PROPERTYKEY key)
+{
+    change_notify *change = (change_notify *)This;
+    struct ao *ao = change->ao;
+
+    if (pwstrDeviceId && !wcscmp(change->monitored, pwstrDeviceId)) {
+        MP_VERBOSE(ao, "OnPropertyValueChanged triggered\n");
+        MP_VERBOSE(ao, "Changed property: ");
+        if (!PKEY_compare(&PKEY_AudioEngine_DeviceFormat, &key)) {
+            MP_VERBOSE(change->ao,
+                       "PKEY_AudioEngine_DeviceFormat - requesting ao reload\n");
+            ao_request_reload(change->ao);
+        } else {
+            MP_VERBOSE(ao, "%s\n", PKEY_to_str(&key));
+        }
+    }
+    return S_OK;
+}
+
+static CONST_VTBL IMMNotificationClientVtbl sIMMDeviceEnumeratorVtbl_vtbl = {
+    .QueryInterface = sIMMNotificationClient_QueryInterface,
+    .AddRef = sIMMNotificationClient_AddRef,
+    .Release = sIMMNotificationClient_Release,
+    .OnDeviceStateChanged = sIMMNotificationClient_OnDeviceStateChanged,
+    .OnDeviceAdded = sIMMNotificationClient_OnDeviceAdded,
+    .OnDeviceRemoved = sIMMNotificationClient_OnDeviceRemoved,
+    .OnDefaultDeviceChanged = sIMMNotificationClient_OnDefaultDeviceChanged,
+    .OnPropertyValueChanged = sIMMNotificationClient_OnPropertyValueChanged,
+};
+
+
+HRESULT wasapi_change_init(struct ao *ao)
+{
+    MP_DBG(ao, "Setting up monitoring on playback device\n");
+    struct wasapi_state *state = (struct wasapi_state *)ao->priv;
+    struct change_notify *change = &state->change;
+    /* COM voodoo to emulate c++ class */
+    change->client.lpVtbl = &sIMMDeviceEnumeratorVtbl_vtbl;
+
+    /* so the callbacks can access the ao */
+    change->ao = ao;
+
+    /* get the device string to compare with the pwstrDeviceId argument in callbacks */
+    HRESULT hr = IMMDevice_GetId(state->pDevice, &change->monitored);
+    EXIT_ON_ERROR(hr);
+
+    /* register the change notification client */
+    hr = IMMDeviceEnumerator_RegisterEndpointNotificationCallback(
+        state->pEnumerator, (IMMNotificationClient *)change);
+    EXIT_ON_ERROR(hr);
+
+    MP_VERBOSE(state, "Monitoring changes in device: %S\n", state->change.monitored);
+    return hr;
+exit_label:
+    MP_ERR(state, "Error setting up device change monitoring: %s\n",
+           wasapi_explain_err(hr));
+    wasapi_change_uninit(ao);
+    return hr;
+}
+
+void wasapi_change_uninit(struct ao *ao)
+{
+    struct wasapi_state *state = (struct wasapi_state *)ao->priv;
+    struct change_notify *change = &state->change;
+
+    if( state->pEnumerator && change->client.lpVtbl )
+        IMMDeviceEnumerator_UnregisterEndpointNotificationCallback(
+            state->pEnumerator, (IMMNotificationClient *)change);
+
+    if (change->monitored) CoTaskMemFree(change->monitored);
+}

audio/out/ao_wasapi_utils.c

@@ -20,6 +20,7 @@
 #define COBJMACROS 1
 #define _WIN32_WINNT 0x600
 
+#include <math.h>
 #include <initguid.h>
 #include <audioclient.h>
 #include <endpointvolume.h>
@@ -30,14 +31,10 @@
 
 #include "audio/format.h"
 #include "osdep/io.h"
+#include "osdep/timer.h"
 
 #define MIXER_DEFAULT_LABEL L"mpv - video player"
 
-#define EXIT_ON_ERROR(hres)  \
-              do { if (FAILED(hres)) { goto exit_label; } } while(0)
-#define SAFE_RELEASE(unk, release) \
-              do { if ((unk) != NULL) { release; (unk) = NULL; } } while(0)
-
 #ifndef PKEY_Device_FriendlyName
 DEFINE_PROPERTYKEY(PKEY_Device_FriendlyName,
                    0xa45c254e, 0xdf1c, 0x4efd, 0x80, 0x20,
@@ -85,6 +82,10 @@ const char *wasapi_explain_err(const HRESULT hr)
 #define E(x) case x : return # x ;
     switch (hr) {
     E(S_OK)
+    E(E_FAIL)
+    E(E_OUTOFMEMORY)
+    E(E_POINTER)
+    E(E_INVALIDARG)
     E(AUDCLNT_E_NOT_INITIALIZED)
     E(AUDCLNT_E_ALREADY_INITIALIZED)
     E(AUDCLNT_E_WRONG_ENDPOINT_TYPE)
@@ -142,46 +143,35 @@ static void set_format(WAVEFORMATEXTENSIBLE *wformat, WORD bytepersample,
     wformat->dwChannelMask = chanmask;
 }
 
-static int format_set_bits(int old_format, int bits, int fp) {
-    int format = old_format;
-    format &= (~AF_FORMAT_BITS_MASK) & (~AF_FORMAT_POINT_MASK) & (~AF_FORMAT_SIGN_MASK);
-    format |= AF_FORMAT_SI;
-
-    switch (bits) {
-    case 32:
-        format |= AF_FORMAT_32BIT;
-        break;
-    case 24:
-        format |= AF_FORMAT_24BIT;
-        break;
-    case 16:
-        format |= AF_FORMAT_16BIT;
-        break;
-    default:
-        abort(); // (should be) unreachable
-    }
-
+static int format_set_bits(int old_format, int bits, int fp)
+{
     if (fp) {
-        format |= AF_FORMAT_F;
-    } else {
-        format |= AF_FORMAT_I;
+        switch (bits) {
+        case 64: return AF_FORMAT_DOUBLE;
+        case 32: return AF_FORMAT_FLOAT;
+        default: return 0;
+        }
     }
 
-    return format;
+    return af_fmt_change_bits(old_format, bits);
 }
 
 static int set_ao_format(struct wasapi_state *state,
                          struct ao *const ao,
-                         WAVEFORMATEXTENSIBLE wformat) {
+                         WAVEFORMATEXTENSIBLE wformat)
+{
+    if (wformat.SubFormat.Data1 != 1 && wformat.SubFormat.Data1 != 3) {
+        MP_ERR(ao, "Unknown SubFormat %"PRIu32"\n",
+               (uint32_t)wformat.SubFormat.Data1);
+        return 0;
+    }
+
     // .Data1 == 1 is PCM, .Data1 == 3 is IEEE_FLOAT
     int format = format_set_bits(ao->format,
         wformat.Format.wBitsPerSample, wformat.SubFormat.Data1 == 3);
 
-    if (wformat.SubFormat.Data1 != 1 && wformat.SubFormat.Data1 != 3) {
-        MP_ERR(ao, "unknown SubFormat %"PRIu32"\n",
-               (uint32_t)wformat.SubFormat.Data1);
+    if (!format)
         return 0;
-    }
 
     ao->samplerate = wformat.Format.nSamplesPerSec;
     ao->bps = wformat.Format.nAvgBytesPerSec;
@@ -204,8 +194,10 @@ static int try_format(struct wasapi_state *state,
     set_format(&wformat, bits / 8, samplerate, channels.num, mp_chmap_to_waveext(&channels));
 
     int af_format = format_set_bits(ao->format, bits, bits == 32);
+    if (!af_format)
+        return 0;
 
-    MP_VERBOSE(ao, "trying %dch %s @ %dhz\n",
+    MP_VERBOSE(ao, "Trying %dch %s @ %dhz\n",
                channels.num, af_fmt_to_str(af_format), samplerate);
 
     union WAVEFMT u;
@@ -229,7 +221,7 @@ static int try_format(struct wasapi_state *state,
 
     if (hr == S_FALSE) {
         if (set_ao_format(state, ao, wformat)) {
-            MP_VERBOSE(ao, "accepted as %dch %s @ %dhz\n",
+            MP_VERBOSE(ao, "Accepted as %dch %s @ %dhz\n",
                        ao->channels.num, af_fmt_to_str(ao->format), ao->samplerate);
 
             return 1;
@@ -292,14 +284,13 @@ static int try_passthrough(struct wasapi_state *state,
     union WAVEFMT u;
     u.extensible = &wformat;
 
-    MP_VERBOSE(ao, "trying passthrough for %s...\n",
-               af_fmt_to_str((ao->format&~AF_FORMAT_END_MASK) | AF_FORMAT_LE));
+    MP_VERBOSE(ao, "Trying passthrough for %s...\n", af_fmt_to_str(ao->format));
 
     HRESULT hr = IAudioClient_IsFormatSupported(state->pAudioClient,
                                                 state->share_mode,
                                                 u.ex, NULL);
     if (!FAILED(hr)) {
-        ao->format = (ao->format&~AF_FORMAT_END_MASK) | AF_FORMAT_LE;
+        ao->format = ao->format;
         state->format = wformat;
         return 1;
     }
@@ -310,11 +301,11 @@ static int find_formats(struct ao *const ao)
 {
     struct wasapi_state *state = (struct wasapi_state *)ao->priv;
 
-    if (AF_FORMAT_IS_IEC61937(ao->format) || ao->format == AF_FORMAT_MPEG2) {
+    if (AF_FORMAT_IS_IEC61937(ao->format)) {
         if (try_passthrough(state, ao))
             return 0;
 
-        MP_ERR(ao, "couldn't use passthrough!");
+        MP_ERR(ao, "Couldn't use passthrough");
         if (!state->opt_exclusive)
             MP_ERR(ao, " (try exclusive mode)");
         MP_ERR(ao, "\n");
@@ -336,7 +327,7 @@ static int find_formats(struct ao *const ao)
             return 0;
         }
 
-        MP_WARN(ao, "couldn't use default mix format!\n");
+        MP_WARN(ao, "Couldn't use default mix format\n");
     }
 
     /* Exclusive mode, we have to guess. */
@@ -407,13 +398,13 @@ static int find_formats(struct ao *const ao)
             bits = start_bits;
             mp_chmap_from_channels(&ao->channels, 2);
         } else {
-            MP_ERR(ao, "couldn't find acceptable audio format!\n");
+            MP_ERR(ao, "Couldn't find acceptable audio format\n");
             return -1;
         }
     }
 }
 
-static int init_clock(struct wasapi_state *state) {
+static HRESULT init_clock(struct wasapi_state *state) {
     HRESULT hr;
 
     hr = IAudioClient_GetService(state->pAudioClient,
@@ -427,16 +418,17 @@ static int init_clock(struct wasapi_state *state) {
 
     atomic_store(&state->sample_count, 0);
 
-    MP_VERBOSE(state, "IAudioClock::GetFrequency gave a frequency of %"PRIu64".\n", (uint64_t) state->clock_frequency);
+    MP_VERBOSE(state, "IAudioClock::GetFrequency gave a frequency of %"PRIu64".\n",
+               (uint64_t) state->clock_frequency);
 
-    return 0;
+    return S_OK;
 exit_label:
-    MP_ERR(state, "init_clock failed with %s, unable to obtain the audio device's timing!\n",
-           wasapi_explain_err(hr));
-    return 1;
+    MP_ERR(state, "Error obtaining the audio device's timing: %s (0x%"PRIx32")\n",
+           wasapi_explain_err(hr), (uint32_t)hr);
+    return hr;
 }
 
-static int init_session_display(struct wasapi_state *state) {
+static HRESULT init_session_display(struct wasapi_state *state) {
     HRESULT hr;
     wchar_t path[MAX_PATH+12] = {0};
 
@@ -453,44 +445,59 @@ static int init_session_display(struct wasapi_state *state) {
     hr = IAudioSessionControl_SetIconPath(state->pSessionControl, path, NULL);
     EXIT_ON_ERROR(hr);
 
-    return 0;
-
+    return S_OK;
 exit_label:
-    MP_ERR(state, "init_session_display failed with %s.\n",
-           wasapi_explain_err(hr));
-    return 1;
+    MP_ERR(state, "Error setting audio session display name: %s (0x%"PRIx32")\n",
+           wasapi_explain_err(hr), (uint32_t)hr);
+    return hr;
 }
 
-static int fix_format(struct wasapi_state *state)
+static HRESULT fix_format(struct ao *ao)
 {
+    struct wasapi_state *state = (struct wasapi_state *)ao->priv;
     HRESULT hr;
-    double offset = 0.5;
 
-    /* cargo cult code to negotiate buffer block size, affected by hardware/drivers combinations,
-       gradually grow it to 10s, by 0.5s, consider failure if it still doesn't work
-     */
-    hr = IAudioClient_GetDevicePeriod(state->pAudioClient,
-                                      &state->defaultRequestedDuration,
-                                      &state->minRequestedDuration);
+    REFERENCE_TIME devicePeriod, bufferDuration, bufferPeriod;
+    MP_DBG(state, "IAudioClient::GetDevicePeriod\n");
+    hr = IAudioClient_GetDevicePeriod(state->pAudioClient,&devicePeriod, NULL);
+    MP_VERBOSE(state, "Device period: %.2g ms\n", (double) devicePeriod / 10000.0 );
+
+    /* integer multiple of device period close to 50ms */
+    bufferPeriod = bufferDuration = ceil( 50.0 * 10000.0 / devicePeriod ) * devicePeriod;
+    if (state->share_mode == AUDCLNT_SHAREMODE_SHARED)
+        bufferPeriod = 0;
+
+    /* handle unsupported buffer size */
+    /* hopefully this shouldn't happen because of the above integer device period */
+    /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd370875%28v=vs.85%29.aspx */
+    int retries=0;
 reinit:
+    MP_DBG(state, "IAudioClient::Initialize\n");
     hr = IAudioClient_Initialize(state->pAudioClient,
                                  state->share_mode,
                                  AUDCLNT_STREAMFLAGS_EVENTCALLBACK,
-                                 state->defaultRequestedDuration,
-                                 state->defaultRequestedDuration,
+                                 bufferDuration,
+                                 bufferPeriod,
                                  &(state->format.Format),
                                  NULL);
-    /* something about buffer sizes on Win7, fixme might loop forever */
+    /* something about buffer sizes on Win7 */
     if (hr == AUDCLNT_E_BUFFER_SIZE_NOT_ALIGNED) {
-        MP_VERBOSE(state, "IAudioClient::Initialize negotiation failed with %s, used %lld * 100ns\n",
-                   wasapi_explain_err(hr), state->defaultRequestedDuration);
-        if (offset > 10.0)
-            goto exit_label;                /* is 10 enough to break out of the loop?*/
+        MP_VERBOSE(state, "IAudioClient::Initialize negotiation failed with %s (0x%"PRIx32"), used %lld * 100ns\n",
+                   wasapi_explain_err(hr), (uint32_t)hr, bufferDuration);
+        if (retries > 0) {
+            hr = E_FAIL;
+            EXIT_ON_ERROR(hr);
+        } else {
+            retries ++;
+        }
+
         IAudioClient_GetBufferSize(state->pAudioClient, &state->bufferFrameCount);
-        state->defaultRequestedDuration =
+        bufferPeriod = bufferDuration =
             (REFERENCE_TIME)((10000.0 * 1000 / state->format.Format.nSamplesPerSec *
-                              state->bufferFrameCount) + offset);
-        offset += 0.5;
+                              state->bufferFrameCount) + 0.5);
+        if (state->share_mode == AUDCLNT_SHAREMODE_SHARED)
+            bufferPeriod = 0;
+
         IAudioClient_Release(state->pAudioClient);
         state->pAudioClient = NULL;
         hr = IMMDeviceActivator_Activate(state->pDevice,
@@ -499,40 +506,54 @@ reinit:
         goto reinit;
     }
     EXIT_ON_ERROR(hr);
+
+    MP_DBG(state, "IAudioClient::Initialize pRenderClient\n");
     hr = IAudioClient_GetService(state->pAudioClient,
                                  &IID_IAudioRenderClient,
                                  (void **)&state->pRenderClient);
     EXIT_ON_ERROR(hr);
+
+    MP_DBG(state, "IAudioClient::Initialize pAudioVolume\n");
     hr = IAudioClient_GetService(state->pAudioClient,
                                  &IID_ISimpleAudioVolume,
                                  (void **) &state->pAudioVolume);
     EXIT_ON_ERROR(hr);
 
-    if (!state->hFeed)
-        goto exit_label;
+    MP_DBG(state, "IAudioClient::Initialize IAudioClient_SetEventHandle\n");
     hr = IAudioClient_SetEventHandle(state->pAudioClient, state->hFeed);
     EXIT_ON_ERROR(hr);
+
+    MP_DBG(state, "IAudioClient::Initialize IAudioClient_GetBufferSize\n");
     hr = IAudioClient_GetBufferSize(state->pAudioClient,
                                     &state->bufferFrameCount);
     EXIT_ON_ERROR(hr);
+
+    ao->device_buffer = state->bufferFrameCount;
     state->buffer_block_size = state->format.Format.nChannels *
                                state->format.Format.wBitsPerSample / 8 *
                                state->bufferFrameCount;
+    bufferDuration =
+        (REFERENCE_TIME)((10000.0 * 1000 / state->format.Format.nSamplesPerSec *
+                          state->bufferFrameCount) + 0.5);
+    MP_VERBOSE(state, "Buffer frame count: %"PRIu32" (%.2g ms)\n",
+               state->bufferFrameCount, (double) bufferDuration / 10000.0 );
 
-    if (init_clock(state))
-        return 1;
-    if (init_session_display(state))
-        return 1;
+    hr = init_clock(state);
+    EXIT_ON_ERROR(hr);
+
+    hr = init_session_display(state);
+    EXIT_ON_ERROR(hr);
 
     state->hTask =
         state->VistaBlob.pAvSetMmThreadCharacteristicsW(L"Pro Audio", &state->taskIndex);
-    MP_VERBOSE(state, "fix_format OK, using %lld byte buffer block size!\n",
+    MP_VERBOSE(state, "Format fixed. Using %lld byte buffer block size\n",
                (long long) state->buffer_block_size);
-    return 0;
+
+    return S_OK;
 exit_label:
-    MP_ERR(state, "fix_format fails with %s, failed to determine buffer block size!\n",
-           wasapi_explain_err(hr));
-    return 1;
+    MP_ERR(state, "Error initializing device: %s (0x%"PRIx32")\n",
+           wasapi_explain_err(hr), (uint32_t)hr);
+    return hr;
 }
 
 static char* get_device_id(IMMDevice *pDevice) {
@@ -629,8 +650,10 @@ end:
     return found;
 }
 
-static HRESULT enumerate_with_state(struct mp_log *log, char *header,
-                                    int status, int with_id)
+// Warning: ao and list are NULL in the "--ao=wasapi:device=help" path!
+static HRESULT enumerate_with_state(struct mp_log *log, struct ao *ao,
+                                    struct ao_device_list *list,
+                                    char *header, int status, int with_id)
 {
     HRESULT hr;
     IMMDeviceEnumerator *pEnumerator = NULL;
@@ -643,7 +666,7 @@ static HRESULT enumerate_with_state(struct mp_log *log, char *header,
     EXIT_ON_ERROR(hr);
 
     hr = IMMDeviceEnumerator_GetDefaultAudioEndpoint(pEnumerator,
-                                                     eRender, eConsole,
+                                                     eRender, eMultimedia,
                                                      &pDevice);
     EXIT_ON_ERROR(hr);
 
@@ -678,6 +701,12 @@ static HRESULT enumerate_with_state(struct mp_log *log, char *header,
             mp_info(log, "%s, ID: %s%s\n", name, id, mark);
         }
 
+        if (ao) {
+            char *desc = talloc_asprintf(NULL, "%s, ID: %s%s", name, id, mark);
+            struct ao_device_desc e = {id, desc};
+            ao_device_list_add(list, ao, &e);
+        }
+
         talloc_free(name);
         talloc_free(id);
         SAFE_RELEASE(pDevice, IMMDevice_Release(pDevice));
@@ -685,8 +714,8 @@ static HRESULT enumerate_with_state(struct mp_log *log, char *header,
     talloc_free(defid);
     SAFE_RELEASE(pDevices, IMMDeviceCollection_Release(pDevices));
     SAFE_RELEASE(pEnumerator, IMMDeviceEnumerator_Release(pEnumerator));
-    return hr;
 
+    return S_OK;
 exit_label:
     talloc_free(defid);
     SAFE_RELEASE(pDevice, IMMDevice_Release(pDevice));
@@ -695,29 +724,50 @@ exit_label:
     return hr;
 }
 
-int wasapi_enumerate_devices(struct mp_log *log)
+int wasapi_enumerate_devices(struct mp_log *log, struct ao *ao,
+                             struct ao_device_list *list)
 {
     HRESULT hr;
     CoInitializeEx(NULL, COINIT_APARTMENTTHREADED);
 
-    hr = enumerate_with_state(log, "Active devices:", DEVICE_STATE_ACTIVE, 1);
+    hr = enumerate_with_state(log, ao, list, "Active devices:",
+                              DEVICE_STATE_ACTIVE, 1);
     EXIT_ON_ERROR(hr);
-    hr = enumerate_with_state(log, "Unplugged devices:", DEVICE_STATE_UNPLUGGED, 0);
+    hr = enumerate_with_state(log, ao, list, "Unplugged devices:",
+                              DEVICE_STATE_UNPLUGGED, 0);
     EXIT_ON_ERROR(hr);
     CoUninitialize();
     return 0;
 exit_label:
-    mp_err(log, "Error enumerating devices: HRESULT %08"PRIx32" \"%s\"\n",
-           (uint32_t)hr, wasapi_explain_err(hr));
+    mp_err(log, "Error enumerating devices: %s (0x%"PRIx32")\n",
+           wasapi_explain_err(hr), (uint32_t)hr);
     CoUninitialize();
     return 1;
 }
 
-static HRESULT find_and_load_device(struct ao *ao, IMMDevice **ppDevice,
-                                    char *search)
+static HRESULT load_default_device(struct ao *ao, IMMDeviceEnumerator* pEnumerator,
+                                   IMMDevice **ppDevice)
+{
+    HRESULT hr = IMMDeviceEnumerator_GetDefaultAudioEndpoint(pEnumerator,
+                                                             eRender, eMultimedia,
+                                                             ppDevice);
+    EXIT_ON_ERROR(hr);
+
+    char *id = get_device_id(*ppDevice);
+    MP_VERBOSE(ao, "Default device ID: %s\n", id);
+    talloc_free(id);
+
+    return S_OK;
+exit_label:
+    MP_ERR(ao , "Error loading default device: %s (0x%"PRIx32")\n",
+           wasapi_explain_err(hr), (uint32_t)hr);
+    return hr;
+}
+
+static HRESULT find_and_load_device(struct ao *ao, IMMDeviceEnumerator* pEnumerator,
+                                    IMMDevice **ppDevice, char *search)
 {
     HRESULT hr;
-    IMMDeviceEnumerator *pEnumerator = NULL;
     IMMDeviceCollection *pDevices = NULL;
     IMMDevice *pTempDevice = NULL;
     LPWSTR deviceID = NULL;
@@ -730,10 +780,6 @@ static HRESULT find_and_load_device(struct ao *ao, IMMDevice **ppDevice,
         devid = search;
     }
 
-    hr = CoCreateInstance(&CLSID_MMDeviceEnumerator, NULL, CLSCTX_ALL,
-                          &IID_IMMDeviceEnumerator, (void**)&pEnumerator);
-    EXIT_ON_ERROR(hr);
-
     int search_err = 0;
 
     if (devid == NULL) {
@@ -745,16 +791,16 @@ static HRESULT find_and_load_device(struct ao *ao, IMMDevice **ppDevice,
         IMMDeviceCollection_GetCount(pDevices, &count);
 
         if (devno >= count) {
-            MP_ERR(ao, "no device #%d!\n", devno);
+            MP_ERR(ao, "No device #%d\n", devno);
         } else {
-            MP_VERBOSE(ao, "finding device #%d\n", devno);
+            MP_VERBOSE(ao, "Finding device #%d\n", devno);
             hr = IMMDeviceCollection_Item(pDevices, devno, &pTempDevice);
             EXIT_ON_ERROR(hr);
 
             hr = IMMDevice_GetId(pTempDevice, &deviceID);
             EXIT_ON_ERROR(hr);
 
-            MP_VERBOSE(ao, "found device #%d\n", devno);
+            MP_VERBOSE(ao, "Found device #%d\n", devno);
         }
     } else {
         hr = IMMDeviceEnumerator_EnumAudioEndpoints(pEnumerator, eRender,
@@ -765,7 +811,7 @@ static HRESULT find_and_load_device(struct ao *ao, IMMDevice **ppDevice,
         int count;
         IMMDeviceCollection_GetCount(pDevices, &count);
 
-        MP_VERBOSE(ao, "finding device %s\n", devid);
+        MP_VERBOSE(ao, "Finding device %s\n", devid);
 
         IMMDevice *prevDevice = NULL;
 
@@ -783,7 +829,7 @@ static HRESULT find_and_load_device(struct ao *ao, IMMDevice **ppDevice,
                 if (deviceID) {
                     char *name;
                     if (!search_err) {
-                        MP_ERR(ao, "multiple matching devices found!\n");
+                        MP_ERR(ao, "Multiple matching devices found\n");
                         name = get_device_name(prevDevice);
                         MP_ERR(ao, "%s\n", name);
                         talloc_free(name);
@@ -802,7 +848,7 @@ static HRESULT find_and_load_device(struct ao *ao, IMMDevice **ppDevice,
         }
 
         if (deviceID == NULL) {
-            MP_ERR(ao, "could not find device %s!\n", devid);
+            MP_ERR(ao, "Could not find device %s\n", devid);
         }
     }
 
@@ -812,19 +858,20 @@ static HRESULT find_and_load_device(struct ao *ao, IMMDevice **ppDevice,
     if (deviceID == NULL || search_err) {
         hr = E_NOTFOUND;
     } else {
-        MP_VERBOSE(ao, "loading device %S\n", deviceID);
+        MP_VERBOSE(ao, "Loading device %S\n", deviceID);
 
         hr = IMMDeviceEnumerator_GetDevice(pEnumerator, deviceID, ppDevice);
 
         if (FAILED(hr)) {
-            MP_ERR(ao, "could not load requested device!\n");
+            MP_ERR(ao, "Could not load requested device\n");
         }
     }
 
 exit_label:
     SAFE_RELEASE(pTempDevice, IMMDevice_Release(pTempDevice));
     SAFE_RELEASE(pDevices, IMMDeviceCollection_Release(pDevices));
-    SAFE_RELEASE(pEnumerator, IMMDeviceEnumerator_Release(pEnumerator));
+
+    CoTaskMemFree(deviceID);
     return hr;
 }
 
@@ -832,11 +879,11 @@ int wasapi_validate_device(struct mp_log *log, const m_option_t *opt,
                            struct bstr name, struct bstr param)
 {
     if (bstr_equals0(param, "help")) {
-        wasapi_enumerate_devices(log);
+        wasapi_enumerate_devices(log, NULL, NULL);
         return M_OPT_EXIT;
     }
 
-    mp_dbg(log, "validating device=%s\n", param.start);
+    mp_dbg(log, "Validating device=%s\n", param.start);
 
     char *end;
     int devno = (int) strtol(param.start, &end, 10);
@@ -869,8 +916,8 @@ HRESULT wasapi_setup_proxies(struct wasapi_state *state) {
 
 exit_label:
     if (hr != S_OK) {
-        MP_ERR(state, "error reading COM proxy: %08x %s\n", (unsigned int)hr,
-               wasapi_explain_err(hr));
+        MP_ERR(state, "Error reading COM proxy: %s (0x%"PRIx32")\n",
+               wasapi_explain_err(hr), (uint32_t)hr);
     }
     return hr;
 }
@@ -905,80 +952,129 @@ exit_label:
     return hr;
 }
 
-int wasapi_thread_init(struct ao *ao)
+static void destroy_proxies(struct wasapi_state *state) {
+    SAFE_RELEASE(state->sAudioClient, IUnknown_Release(state->sAudioClient));
+    SAFE_RELEASE(state->sAudioVolume, IUnknown_Release(state->sAudioVolume));
+    SAFE_RELEASE(state->sEndpointVolume, IUnknown_Release(state->sEndpointVolume));
+    SAFE_RELEASE(state->sSessionControl, IUnknown_Release(state->sSessionControl));
+}
+
+void wasapi_dispatch(void)
+{
+    /* dispatch any possible pending messages */
+    MSG msg;
+    while (PeekMessage(&msg, NULL, 0, 0, PM_REMOVE))
+        DispatchMessage(&msg);
+}
+
+HRESULT wasapi_thread_init(struct ao *ao)
 {
     struct wasapi_state *state = (struct wasapi_state *)ao->priv;
     HRESULT hr;
-    CoInitializeEx(NULL, COINIT_APARTMENTTHREADED);
+    MP_DBG(ao, "Init wasapi thread\n");
+    int64_t retry_wait = 1;
+retry:
+    state->initial_volume = -1.0;
 
-    if (!state->opt_device) {
-        IMMDeviceEnumerator *pEnumerator;
-        hr = CoCreateInstance(&CLSID_MMDeviceEnumerator, NULL, CLSCTX_ALL,
-                              &IID_IMMDeviceEnumerator, (void**)&pEnumerator);
-        EXIT_ON_ERROR(hr);
+    hr = CoCreateInstance(&CLSID_MMDeviceEnumerator, NULL, CLSCTX_ALL,
+                          &IID_IMMDeviceEnumerator, (void**)&state->pEnumerator);
+    EXIT_ON_ERROR(hr);
 
-        hr = IMMDeviceEnumerator_GetDefaultAudioEndpoint(pEnumerator,
-                                                         eRender, eConsole,
-                                                         &state->pDevice);
-        SAFE_RELEASE(pEnumerator, IMMDeviceEnumerator_Release(pEnumerator));
+    char *device = state->opt_device;
+    if (!device || !device[0])
+        device = ao->device;
 
-        char *id = get_device_id(state->pDevice);
-        MP_VERBOSE(ao, "default device ID: %s\n", id);
-        talloc_free(id);
-    } else {
-        hr = find_and_load_device(ao, &state->pDevice, state->opt_device);
-    }
+
+    if (!device || !device[0])
+        hr = load_default_device(ao, state->pEnumerator, &state->pDevice);
+    else
+        hr = find_and_load_device(ao, state->pEnumerator, &state->pDevice, device);
     EXIT_ON_ERROR(hr);
 
     char *name = get_device_name(state->pDevice);
-    MP_VERBOSE(ao, "device loaded: %s\n", name);
+    MP_VERBOSE(ao, "Device loaded: %s\n", name);
     talloc_free(name);
 
+    MP_DBG(ao, "Activating pAudioClient interface\n");
     hr = IMMDeviceActivator_Activate(state->pDevice, &IID_IAudioClient,
                                      CLSCTX_ALL, NULL, (void **)&state->pAudioClient);
     EXIT_ON_ERROR(hr);
 
+    MP_DBG(ao, "Activating pEndpointVolume interface\n");
     hr = IMMDeviceActivator_Activate(state->pDevice, &IID_IAudioEndpointVolume,
                                      CLSCTX_ALL, NULL,
                                      (void **)&state->pEndpointVolume);
     EXIT_ON_ERROR(hr);
-    IAudioEndpointVolume_QueryHardwareSupport(state->pEndpointVolume,
-                                              &state->vol_hw_support);
 
-    state->init_ret = find_formats(ao); /* Probe support formats */
-    if (state->init_ret)
-        goto exit_label;
-    if (!fix_format(state)) { /* now that we're sure what format to use */
-        EXIT_ON_ERROR(create_proxies(state));
+    MP_DBG(ao, "Query hardware volume support\n");
+    hr = IAudioEndpointVolume_QueryHardwareSupport(state->pEndpointVolume,
+                                                   &state->vol_hw_support);
+    if ( hr != S_OK )
+        MP_WARN(ao, "Query hardware volume control: %s (0x%"PRIx32")\n",
+                wasapi_explain_err(hr), (uint32_t)hr);
 
-        if (state->opt_exclusive)
-            IAudioEndpointVolume_GetMasterVolumeLevelScalar(state->pEndpointVolume,
-                                                            &state->initial_volume);
-        else
-            ISimpleAudioVolume_GetMasterVolume(state->pAudioVolume,
-                                               &state->initial_volume);
-
-        state->previous_volume = state->initial_volume;
-
-        MP_VERBOSE(ao, "thread_init OK!\n");
-        SetEvent(state->init_done);
-        return state->init_ret;
+    MP_DBG(ao, "Probing formats\n");
+    if (find_formats(ao)){
+        hr = E_FAIL;
+        EXIT_ON_ERROR(hr);
     }
+
+    MP_DBG(ao, "Fixing format\n");
+    hr = fix_format(ao);
+    if ( (hr == AUDCLNT_E_DEVICE_IN_USE ||
+          hr == AUDCLNT_E_DEVICE_INVALIDATED) &&
+         retry_wait <= 8 ) {
+        wasapi_thread_uninit(ao);
+        MP_WARN(ao, "Retrying in %"PRId64" us\n", retry_wait);
+        mp_sleep_us(retry_wait);
+        retry_wait *= 2;
+        goto retry;
+    }
+    EXIT_ON_ERROR(hr);
+
+    MP_DBG(ao, "Creating proxies\n");
+    hr = create_proxies(state);
+    EXIT_ON_ERROR(hr);
+
+    MP_DBG(ao, "Read volume levels\n");
+    if (state->opt_exclusive)
+        IAudioEndpointVolume_GetMasterVolumeLevelScalar(state->pEndpointVolume,
+                                                        &state->initial_volume);
+    else
+        ISimpleAudioVolume_GetMasterVolume(state->pAudioVolume,
+                                           &state->initial_volume);
+
+    state->previous_volume = state->initial_volume;
+
+    wasapi_change_init(ao);
+
+    MP_DBG(ao, "Init wasapi thread done\n");
+    return S_OK;
 exit_label:
-    state->init_ret = -1;
-    SetEvent(state->init_done);
-    return -1;
+    MP_ERR(state, "Error setting up audio thread: %s (0x%"PRIx32")\n",
+           wasapi_explain_err(hr), (uint32_t)hr);
+    return hr;
 }
 
-void wasapi_thread_uninit(wasapi_state *state)
+void wasapi_thread_uninit(struct ao *ao)
 {
+    struct wasapi_state *state = (struct wasapi_state *)ao->priv;
+
+    wasapi_dispatch();
+
     if (state->pAudioClient)
         IAudioClient_Stop(state->pAudioClient);
 
-    if (state->opt_exclusive)
+    wasapi_change_uninit(ao);
+
+    if (state->opt_exclusive &&
+        state->pEndpointVolume &&
+        state->initial_volume > 0 )
         IAudioEndpointVolume_SetMasterVolumeLevelScalar(state->pEndpointVolume,
                                                         state->initial_volume, NULL);
 
+    destroy_proxies(state);
+
     SAFE_RELEASE(state->pRenderClient,   IAudioRenderClient_Release(state->pRenderClient));
     SAFE_RELEASE(state->pAudioClock,     IAudioClock_Release(state->pAudioClock));
     SAFE_RELEASE(state->pAudioVolume,    ISimpleAudioVolume_Release(state->pAudioVolume));
@@ -986,9 +1082,9 @@ void wasapi_thread_uninit(wasapi_state *state)
     SAFE_RELEASE(state->pSessionControl, IAudioSessionControl_Release(state->pSessionControl));
     SAFE_RELEASE(state->pAudioClient,    IAudioClient_Release(state->pAudioClient));
     SAFE_RELEASE(state->pDevice,         IMMDevice_Release(state->pDevice));
+    SAFE_RELEASE(state->pEnumerator,     IMMDeviceEnumerator_Release(state->pEnumerator));
 
     if (state->hTask)
         state->VistaBlob.pAvRevertMmThreadCharacteristics(state->hTask);
-    CoUninitialize();
-    ExitThread(0);
+    MP_DBG(ao, "Thread uninit done\n");
 }

audio/out/ao_wasapi_utils.h

@@ -31,13 +31,16 @@ int wasapi_fill_VistaBlob(wasapi_state *state);
 
 const char *wasapi_explain_err(const HRESULT hr);
 
-int wasapi_enumerate_devices(struct mp_log *log);
+int wasapi_enumerate_devices(struct mp_log *log, struct ao *ao,
+                             struct ao_device_list *list);
 
 int wasapi_validate_device(struct mp_log *log, const m_option_t *opt,
                            struct bstr name, struct bstr param);
 
-int wasapi_thread_init(struct ao *ao);
-void wasapi_thread_uninit(wasapi_state *state);
+
+void wasapi_dispatch(void);
+HRESULT wasapi_thread_init(struct ao *ao);
+void wasapi_thread_uninit(struct ao *ao);
 
 HRESULT wasapi_setup_proxies(wasapi_state *state);
 void wasapi_release_proxies(wasapi_state *state);

audio/out/internal.h

@@ -22,16 +22,8 @@
 #include <stdbool.h>
 #include <pthread.h>
 
-#include "audio/chmap.h"
-#include "audio/chmap_sel.h"
-
-// Minimum buffer size in seconds.
-#define MIN_BUFFER 0.2
-
-// If ao_get_delay() reaches this value after ao_play() was called with the
-// AOPLAY_FINAL_CHUNK flag set, the playback core expects that the audio has
-// all been played.
-#define AO_EOF_DELAY 0.05
+#include "osdep/atomics.h"
+#include "audio/out/ao.h"
 
 /* global data used by ao.c and ao drivers */
 struct ao {
@@ -44,8 +36,6 @@ struct ao {
     int num_planes;
     bool probing;               // if true, don't fail loudly on init
     bool untimed;               // don't assume realtime playback
-    bool no_persistent_volume;  // the AO does the equivalent of af_volume
-    bool per_application_mixer; // like above, but volume persists (per app)
     int device_buffer;          // device buffer in samples (guessed by
                                 // common init code if not set by driver)
     const struct ao_driver *api; // entrypoints to the wrapper (push.c/pull.c)
@@ -55,7 +45,21 @@ struct ao {
     struct input_ctx *input_ctx;
     struct mp_log *log; // Using e.g. "[ao/coreaudio]" as prefix
 
+    // The device as selected by the user, usually using ao_device_desc.name
+    // from an entry from the list returned by driver->list_devices. If the
+    // default device should be used, this is set to NULL.
+    char *device;
+
+    // Application name to report to the audio API.
+    char *client_name;
+
+    // Used during init: if init fails, redirect to this ao
+    char *redirect;
+
+    atomic_bool request_reload;
+
     int buffer;
+    double def_buffer;
     void *api_priv;
 };
 
@@ -131,7 +135,7 @@ struct ao_driver {
     // push based: see ao_play()
     int (*play)(struct ao *ao, void **data, int samples, int flags);
     // push based: see ao_get_delay()
-    float (*get_delay)(struct ao *ao);
+    double (*get_delay)(struct ao *ao);
     // push based: block until all queued audio is played (optional)
     void (*drain)(struct ao *ao);
     // Optional. Return true if audio has stopped in any way.
@@ -151,6 +155,20 @@ struct ao_driver {
     // In combination with wait(). Lock may or may not be held.
     void (*wakeup)(struct ao *ao);
 
+    // Return the list of devices currently available in the system. Use
+    // ao_device_list_add() to add entries. The selected device will be set as
+    // ao->device (using ao_device_desc.name).
+    // Warning: the ao struct passed doesn't necessarily have ao_driver->init()
+    //          called on it - in this case, ->uninit() won't be called either
+    //          after this function. The idea is that list_devs can be called
+    //          both when no audio or when audio is active. the latter can
+    //          happen if the audio config change at runtime, and in this case
+    //          we don't want to force a new connection to the audio server
+    //          just to update the device list. For runtime updates, ->init()
+    //          will have been called. In both cases, ao->priv is properly
+    //          allocated. (Runtime updates are not used/supported yet.)
+    void (*list_devs)(struct ao *ao, struct ao_device_list *list);
+
     // For option parsing (see vo.h)
     int priv_size;
     const void *priv_defaults;
@@ -171,4 +189,9 @@ bool ao_chmap_sel_adjust(struct ao *ao, const struct mp_chmap_sel *s,
 bool ao_chmap_sel_get_def(struct ao *ao, const struct mp_chmap_sel *s,
                           struct mp_chmap *map, int num);
 
+// Add a deep copy of e to the list.
+// Call from ao_driver->list_devs callback only.
+void ao_device_list_add(struct ao_device_list *list, struct ao *ao,
+                        struct ao_device_desc *e);
+
 #endif

audio/out/pull.c

@@ -30,7 +30,7 @@
 
 #include "osdep/timer.h"
 #include "osdep/threads.h"
-#include "compat/atomics.h"
+#include "osdep/atomics.h"
 #include "misc/ring.h"
 
 /*
@@ -43,8 +43,11 @@ enum {
                     // finished, but device is open)
     AO_STATE_WAIT,  // wait for callback to go into AO_STATE_NONE state
     AO_STATE_PLAY,  // play the buffer
+    AO_STATE_BUSY,  // like AO_STATE_PLAY, but ao_read_data() is being called
 };
 
+#define IS_PLAYING(st) ((st) == AO_STATE_PLAY || (st) == AO_STATE_BUSY)
+
 struct ao_pull_state {
     // Be very careful with the order when accessing planes.
     struct mp_ring *buffers[MP_NUM_CHANNELS];
@@ -56,6 +59,21 @@ struct ao_pull_state {
     atomic_llong end_time_us;
 };
 
+static void set_state(struct ao *ao, int new_state)
+{
+    struct ao_pull_state *p = ao->api_priv;
+    while (1) {
+        int old = atomic_load(&p->state);
+        if (old == AO_STATE_BUSY) {
+            // A spinlock, because some audio APIs don't want us to use mutexes.
+            mp_sleep_us(1);
+            continue;
+        }
+        if (atomic_compare_exchange_strong(&p->state, &old, new_state))
+            break;
+    }
+}
+
 static int get_space(struct ao *ao)
 {
     struct ao_pull_state *p = ao->api_priv;
@@ -79,8 +97,10 @@ static int play(struct ao *ao, void **data, int samples, int flags)
         int r = mp_ring_write(p->buffers[n], data[n], write_bytes);
         assert(r == write_bytes);
     }
-    if (atomic_load(&p->state) != AO_STATE_PLAY) {
-        atomic_store(&p->state, AO_STATE_PLAY);
+
+    int state = atomic_load(&p->state);
+    if (!IS_PLAYING(state)) {
+        set_state(ao, AO_STATE_PLAY);
         ao->driver->resume(ao);
     }
 
@@ -101,14 +121,13 @@ int ao_read_data(struct ao *ao, void **data, int samples, int64_t out_time_us)
 
     struct ao_pull_state *p = ao->api_priv;
     int full_bytes = samples * ao->sstride;
-    int state = atomic_load(&p->state);
+    bool need_wakeup = false;
     int bytes = 0;
 
-    if (state != AO_STATE_PLAY) {
-        if (state == AO_STATE_WAIT)
-            atomic_store(&p->state, AO_STATE_NONE);
+    // Play silence in states other than AO_STATE_PLAY.
+    if (!atomic_compare_exchange_strong(&p->state, &(int){AO_STATE_PLAY},
+                                        AO_STATE_BUSY))
         goto end;
-    }
 
     // Since the writer will write the first plane last, its buffered amount
     // of data is the minimum amount across all planes.
@@ -124,10 +143,16 @@ int ao_read_data(struct ao *ao, void **data, int samples, int64_t out_time_us)
     }
 
     // Half of the buffer played -> request more.
-    if (buffered_bytes - bytes <= mp_ring_size(p->buffers[0]) / 2)
-        mp_input_wakeup_nolock(ao->input_ctx);
+    need_wakeup = buffered_bytes - bytes <= mp_ring_size(p->buffers[0]) / 2;
+
+    // Should never fail.
+    atomic_compare_exchange_strong(&p->state, &(int){AO_STATE_BUSY}, AO_STATE_PLAY);
 
 end:
+
+    if (need_wakeup)
+        mp_input_wakeup_nolock(ao->input_ctx);
+
     // pad with silence (underflow/paused/eof)
     for (int n = 0; n < ao->num_planes; n++)
         af_fill_silence(data[n], full_bytes - bytes, ao->format);
@@ -147,7 +172,7 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
 // it takes until the last sample in the buffer reaches the speakers. This is
 // used for audio/video synchronization, so it's very important to implement
 // this correctly.
-static float get_delay(struct ao *ao)
+static double get_delay(struct ao *ao)
 {
     struct ao_pull_state *p = ao->api_priv;
 
@@ -160,19 +185,9 @@ static float get_delay(struct ao *ao)
 static void reset(struct ao *ao)
 {
     struct ao_pull_state *p = ao->api_priv;
-    if (ao->driver->reset) {
+    if (ao->driver->reset)
         ao->driver->reset(ao); // assumes the audio callback thread is stopped
-        atomic_store(&p->state, AO_STATE_NONE);
-    } else {
-        // The thread keeps running. Wait until the audio callback gets into
-        // a defined state where it won't touch the ringbuffer. We must do
-        // this, because emptying the ringbuffer is not an atomic operation.
-        if (atomic_load(&p->state) != AO_STATE_NONE) {
-            atomic_store(&p->state, AO_STATE_WAIT);
-            while (atomic_load(&p->state) != AO_STATE_NONE)
-                mp_sleep_us(1);
-        }
-    }
+    set_state(ao, AO_STATE_NONE);
     for (int n = 0; n < ao->num_planes; n++)
         mp_ring_reset(p->buffers[n]);
     atomic_store(&p->end_time_us, 0);
@@ -180,23 +195,22 @@ static void reset(struct ao *ao)
 
 static void pause(struct ao *ao)
 {
-    struct ao_pull_state *p = ao->api_priv;
     if (ao->driver->reset)
         ao->driver->reset(ao);
-    atomic_store(&p->state, AO_STATE_NONE);
+    set_state(ao, AO_STATE_NONE);
 }
 
 static void resume(struct ao *ao)
 {
-    struct ao_pull_state *p = ao->api_priv;
-    atomic_store(&p->state, AO_STATE_PLAY);
+    set_state(ao, AO_STATE_PLAY);
     ao->driver->resume(ao);
 }
 
 static void drain(struct ao *ao)
 {
     struct ao_pull_state *p = ao->api_priv;
-    if (atomic_load(&p->state) == AO_STATE_PLAY)
+    int state = atomic_load(&p->state);
+    if (IS_PLAYING(state))
         mp_sleep_us(get_delay(ao) * 1000000);
     reset(ao);
 }

audio/out/push.c

@@ -35,7 +35,7 @@
 
 #include "osdep/threads.h"
 #include "osdep/timer.h"
-#include "compat/atomics.h"
+#include "osdep/atomics.h"
 
 #include "audio/audio.h"
 #include "audio/audio_buffer.h"
@@ -44,19 +44,15 @@ struct ao_push_state {
     pthread_t thread;
     pthread_mutex_t lock;
     pthread_cond_t wakeup;
-    pthread_cond_t wakeup_drain;
 
     // --- protected by lock
 
     struct mp_audio_buffer *buffer;
 
     bool terminate;
-    bool drain;
-    bool buffers_full;
-    bool avoid_ao_wait;
+    bool wait_on_ao;
     bool still_playing;
     bool need_wakeup;
-    bool requested_data;
     bool paused;
 
     // Whether the current buffer contains the complete audio.
@@ -94,22 +90,14 @@ static double unlocked_get_delay(struct ao *ao)
     double driver_delay = 0;
     if (ao->driver->get_delay)
         driver_delay = ao->driver->get_delay(ao);
-    double delay = driver_delay + mp_audio_buffer_seconds(p->buffer);
-    if (delay >= AO_EOF_DELAY && p->expected_end_time) {
-        if (mp_time_sec() > p->expected_end_time) {
-            MP_ERR(ao, "Audio device EOF reporting is broken!\n");
-            MP_ERR(ao, "Please report this problem.\n");
-            delay = 0;
-        }
-    }
-    return delay;
+    return driver_delay + mp_audio_buffer_seconds(p->buffer);
 }
 
-static float get_delay(struct ao *ao)
+static double get_delay(struct ao *ao)
 {
     struct ao_push_state *p = ao->api_priv;
     pthread_mutex_lock(&p->lock);
-    float delay = unlocked_get_delay(ao);
+    double delay = unlocked_get_delay(ao);
     pthread_mutex_unlock(&p->lock);
     return delay;
 }
@@ -122,8 +110,9 @@ static void reset(struct ao *ao)
         ao->driver->reset(ao);
     mp_audio_buffer_clear(p->buffer);
     p->paused = false;
+    if (p->still_playing)
+        wakeup_playthread(ao);
     p->still_playing = false;
-    wakeup_playthread(ao);
     pthread_mutex_unlock(&p->lock);
 }
 
@@ -154,20 +143,27 @@ static void drain(struct ao *ao)
 {
     struct ao_push_state *p = ao->api_priv;
 
+    MP_VERBOSE(ao, "draining...\n");
+
     pthread_mutex_lock(&p->lock);
-    if (p->paused) {
-        pthread_mutex_unlock(&p->lock);
-        return;
-    }
+    if (p->paused)
+        goto done;
+
     p->final_chunk = true;
-    p->drain = true;
     wakeup_playthread(ao);
-    while (p->drain)
-        pthread_cond_wait(&p->wakeup_drain, &p->lock);
+    while (p->still_playing && mp_audio_buffer_samples(p->buffer) > 0)
+        pthread_cond_wait(&p->wakeup, &p->lock);
+
+    if (ao->driver->drain) {
+        ao->driver->drain(ao);
+    } else {
+        double time = unlocked_get_delay(ao);
+        mp_sleep_us(MPMIN(time, ao->buffer / (double)ao->samplerate + 1) * 1e6);
+    }
+
+done:
     pthread_mutex_unlock(&p->lock);
 
-    if (!ao->driver->drain)
-        mp_sleep_us(get_delay(ao) * 1000000);
     reset(ao);
 }
 
@@ -177,13 +173,17 @@ static int unlocked_get_space(struct ao *ao)
     int space = mp_audio_buffer_get_write_available(p->buffer);
     if (ao->driver->get_space) {
         // The following code attempts to keep the total buffered audio to
-        // MIN_BUFFER/2+device_buffer in order to improve latency.
+        // ao->buffer in order to improve latency.
         int device_space = ao->driver->get_space(ao);
         int device_buffered = ao->device_buffer - device_space;
         int soft_buffered = mp_audio_buffer_samples(p->buffer);
-        int min_buffer = MIN_BUFFER / 2 * ao->samplerate + ao->device_buffer;
-        int total_buffer = device_buffered + soft_buffered;
-        int missing = min_buffer - total_buffer;
+        // The extra margin helps avoiding too many wakeups if the AO is fully
+        // byte based and doesn't do proper chunked processing.
+        int min_buffer = ao->buffer + 64;
+        int missing = min_buffer - device_buffered - soft_buffered;
+        // But always keep the device's buffer filled as much as we can.
+        int device_missing = device_space - soft_buffered;
+        missing = MPMAX(missing, device_missing);
         space = MPMIN(space, missing);
         space = MPMAX(0, space);
     }
@@ -217,6 +217,8 @@ static int play(struct ao *ao, void **data, int samples, int flags)
     int write_samples = mp_audio_buffer_get_write_available(p->buffer);
     write_samples = MPMIN(write_samples, samples);
 
+    MP_TRACE(ao, "samples=%d flags=%d r=%d\n", samples, flags, write_samples);
+
     if (write_samples < samples)
         flags = flags & ~AOPLAY_FINAL_CHUNK;
     bool is_final = flags & AOPLAY_FINAL_CHUNK;
@@ -230,17 +232,17 @@ static int play(struct ao *ao, void **data, int samples, int flags)
 
     bool got_data = write_samples > 0 || p->paused || p->final_chunk != is_final;
 
-    p->expected_end_time = 0;
     p->final_chunk = is_final;
     p->paused = false;
-    p->still_playing |= write_samples > 0;
+    if (got_data) {
+        p->still_playing = true;
+        p->expected_end_time = 0;
+    }
 
     // If we don't have new data, the decoder thread basically promises it
     // will send new data as soon as it's available.
-    if (got_data) {
-        p->requested_data = false;
+    if (got_data)
         wakeup_playthread(ao);
-    }
     pthread_mutex_unlock(&p->lock);
     return write_samples;
 }
@@ -270,68 +272,46 @@ static void ao_play_data(struct ao *ao)
     }
     r = MPMAX(r, 0);
     // Probably can't copy the rest of the buffer due to period alignment.
-    bool stuck = r <= 0 && space >= max && data.samples > 0;
-    if ((flags & AOPLAY_FINAL_CHUNK) && stuck) {
+    bool stuck_eof = r <= 0 && space >= max && data.samples > 0;
+    if ((flags & AOPLAY_FINAL_CHUNK) && stuck_eof) {
         MP_ERR(ao, "Audio output driver seems to ignore AOPLAY_FINAL_CHUNK.\n");
         r = max;
     }
     mp_audio_buffer_skip(p->buffer, r);
-    if (p->final_chunk && mp_audio_buffer_samples(p->buffer) == 0) {
-        p->expected_end_time = mp_time_sec() + AO_EOF_DELAY + 0.25; // + margin
-        if (ao->driver->get_delay)
-            p->expected_end_time += ao->driver->get_delay(ao);
-    }
-    // In both cases, we have to account for space!=0, but the AO not accepting
-    // any new data (due to rounding to period boundaries).
-    p->buffers_full = max >= space && r <= 0;
-    p->avoid_ao_wait = (max == 0 && space > 0) || p->paused || stuck;
+    if (r > 0)
+        p->expected_end_time = 0;
+    // Nothing written, but more input data than space - this must mean the
+    // AO's get_space() doesn't do period alignment correctly.
+    bool stuck = r == 0 && max >= space && space > 0;
+    if (stuck)
+        MP_ERR(ao, "Audio output is reporting incorrect buffer status.\n");
+    // Wait until space becomes available. Also wait if we actually wrote data,
+    // so the AO wakes us up properly if it needs more data.
+    p->wait_on_ao = space == 0 || r > 0 || stuck;
     p->still_playing |= r > 0;
-    MP_TRACE(ao, "in=%d, space=%d r=%d flags=%d aw=%d full=%d f=%d\n", max,
-             space, r, flags, p->avoid_ao_wait, p->buffers_full, p->final_chunk);
-}
-
-// Estimate when the AO needs data again.
-static double ao_estimate_timeout(struct ao *ao)
-{
-    struct ao_push_state *p = ao->api_priv;
-    double timeout = 0;
-    if (p->buffers_full && ao->driver->get_delay) {
-        timeout = ao->driver->get_delay(ao) - 0.050;
-        // Keep extra safety margin if the buffers are large
-        if (timeout > 0.100)
-            timeout = MPMAX(timeout - 0.200, 0.100);
-    }
-    return MPMAX(timeout, ao->device_buffer * 0.75 / ao->samplerate);
+    // If we just filled the AO completely (r == space), don't refill for a
+    // while. Prevents wakeup feedback with byte-granular AOs.
+    int needed = unlocked_get_space(ao);
+    bool more = needed >= (r == space ? ao->device_buffer / 4 : 1) && !stuck;
+    if (more)
+        mp_input_wakeup(ao->input_ctx); // request more data
+    MP_TRACE(ao, "in=%d flags=%d space=%d r=%d wa=%d needed=%d more=%d\n",
+             max, flags, space, r, p->wait_on_ao, needed, more);
 }
 
 static void *playthread(void *arg)
 {
     struct ao *ao = arg;
     struct ao_push_state *p = ao->api_priv;
+    mpthread_set_name("ao");
     pthread_mutex_lock(&p->lock);
     while (!p->terminate) {
         if (!p->paused)
             ao_play_data(ao);
 
-        // Request new data from decoder if buffer goes below "full".
-        // Allow a small margin of missing data for AOs that use timeouts.
-        double margin = ao->driver->wait ? 0 : ao->device_buffer / 8;
-        if (!p->buffers_full && unlocked_get_space(ao) > margin) {
-            if (!p->requested_data)
-                mp_input_wakeup(ao->input_ctx);
-            p->requested_data = true;
-        }
-
-        if (p->drain && (p->avoid_ao_wait || p->paused)) {
-            if (ao->driver->drain)
-                ao->driver->drain(ao);
-            p->drain = false;
-            pthread_cond_signal(&p->wakeup_drain);
-        }
-
         if (!p->need_wakeup) {
             MP_STATS(ao, "start audio wait");
-            if (p->avoid_ao_wait || p->paused) {
+            if (!p->wait_on_ao || p->paused) {
                 // Avoid busy waiting, because the audio API will still report
                 // that it needs new data, even if we're not ready yet, or if
                 // get_space() decides that the amount of audio buffered in the
@@ -342,14 +322,22 @@ static void *playthread(void *arg)
 
                 bool was_playing = p->still_playing;
                 double timeout = -1;
-                if (p->still_playing && !p->paused) {
-                    timeout = unlocked_get_delay(ao);
-                    if (timeout < AO_EOF_DELAY)
+                if (p->still_playing && !p->paused && p->final_chunk &&
+                    !mp_audio_buffer_samples(p->buffer))
+                {
+                    double now = mp_time_sec();
+                    if (!p->expected_end_time)
+                        p->expected_end_time = now + unlocked_get_delay(ao);
+                    if (p->expected_end_time < now) {
                         p->still_playing = false;
+                    } else {
+                        timeout = p->expected_end_time - now;
+                    }
                 }
 
-                if (!p->requested_data || (was_playing && !p->still_playing))
+                if (was_playing && !p->still_playing)
                     mp_input_wakeup(ao->input_ctx);
+                pthread_cond_signal(&p->wakeup); // for draining
 
                 if (p->still_playing && timeout > 0) {
                     mpthread_cond_timedwait_rel(&p->wakeup, &p->lock, timeout);
@@ -357,10 +345,15 @@ static void *playthread(void *arg)
                     pthread_cond_wait(&p->wakeup, &p->lock);
                 }
             } else {
+                // Wait until the device wants us to write more data to it.
                 if (!ao->driver->wait || ao->driver->wait(ao, &p->lock) < 0) {
                     // Fallback to guessing.
-                    double timeout = ao_estimate_timeout(ao);
-                    mpthread_cond_timedwait_rel(&p->wakeup, &p->lock, timeout);
+                    double timeout = 0;
+                    if (ao->driver->get_delay)
+                        timeout = ao->driver->get_delay(ao);
+                    timeout *= 0.25; // wake up if 25% played
+                    if (!p->need_wakeup)
+                        mpthread_cond_timedwait_rel(&p->wakeup, &p->lock, timeout);
                 }
             }
             MP_STATS(ao, "end audio wait");
@@ -371,6 +364,19 @@ static void *playthread(void *arg)
     return NULL;
 }
 
+static void destroy_no_thread(struct ao *ao)
+{
+    struct ao_push_state *p = ao->api_priv;
+
+    ao->driver->uninit(ao);
+
+    for (int n = 0; n < 2; n++)
+        close(p->wakeup_pipe[n]);
+
+    pthread_cond_destroy(&p->wakeup);
+    pthread_mutex_destroy(&p->lock);
+}
+
 static void uninit(struct ao *ao)
 {
     struct ao_push_state *p = ao->api_priv;
@@ -382,14 +388,7 @@ static void uninit(struct ao *ao)
 
     pthread_join(p->thread, NULL);
 
-    ao->driver->uninit(ao);
-
-    for (int n = 0; n < 2; n++)
-        close(p->wakeup_pipe[n]);
-
-    pthread_cond_destroy(&p->wakeup);
-    pthread_cond_destroy(&p->wakeup_drain);
-    pthread_mutex_destroy(&p->lock);
+    destroy_no_thread(ao);
 }
 
 static int init(struct ao *ao)
@@ -398,9 +397,13 @@ static int init(struct ao *ao)
 
     pthread_mutex_init(&p->lock, NULL);
     pthread_cond_init(&p->wakeup, NULL);
-    pthread_cond_init(&p->wakeup_drain, NULL);
     mp_make_wakeup_pipe(p->wakeup_pipe);
 
+    if (ao->device_buffer <= 0) {
+        MP_FATAL(ao, "Couldn't probe device buffer size.\n");
+        goto err;
+    }
+
     p->buffer = mp_audio_buffer_create(ao);
     mp_audio_buffer_reinit_fmt(p->buffer, ao->format,
                                &ao->channels, ao->samplerate);
@@ -409,7 +412,7 @@ static int init(struct ao *ao)
         goto err;
     return 0;
 err:
-    ao->driver->uninit(ao);
+    destroy_no_thread(ao);
     return -1;
 }
 
@@ -461,7 +464,7 @@ int ao_wait_poll(struct ao *ao, struct pollfd *fds, int num_fds,
     assert(ao->api == &ao_api_push);
     assert(&p->lock == lock);
 
-    if (num_fds > MAX_POLL_FDS || p->wakeup_pipe[0] < 0)
+    if (num_fds >= MAX_POLL_FDS || p->wakeup_pipe[0] < 0)
         return -1;
 
     struct pollfd p_fds[MAX_POLL_FDS];

bootstrap.py

@@ -5,9 +5,9 @@
 from __future__ import print_function
 import os, sys, stat, hashlib, subprocess
 
-WAFRELEASE = "waf-1.7.16"
+WAFRELEASE = "waf-1.8.1"
 WAFURL     = "http://ftp.waf.io/pub/release/" + WAFRELEASE
-SHA256HASH = "b64dc26c882572415fd450b745006107965f3fe17b357e3eb43d6676c9635a61"
+SHA256HASH = "ec658116ba0b96629d91fde0b32321849e866e0819f1e835c4c2c7f7ffe1a21d"
 
 if os.path.exists("waf"):
     wafver = subprocess.check_output(['./waf', '--version']).decode()
@@ -31,6 +31,8 @@ if SHA256HASH == hashlib.sha256(waf).hexdigest():
     print("Checksum verified.")
 else:
     print("The checksum of the downloaded file does not match!")
+    print(" - got:      {}".format(hashlib.sha256(waf).hexdigest()))
+    print(" - expected: {}".format(SHA256HASH))
     print("Please download and verify the file manually.")
 
     sys.exit(1)

common/av_common.c

@@ -117,6 +117,8 @@ void mp_set_av_packet(AVPacket *dst, struct demux_packet *mpkt, AVRational *tb)
     if (mpkt && mpkt->avpacket) {
         dst->side_data = mpkt->avpacket->side_data;
         dst->side_data_elems = mpkt->avpacket->side_data_elems;
+        if (dst->data == mpkt->avpacket->data)
+            dst->buf = mpkt->avpacket->buf;
     }
     if (mpkt && tb && tb->num > 0 && tb->den > 0)
         dst->duration = mpkt->duration / av_q2d(*tb);

common/codecs.c

@@ -17,7 +17,7 @@
 
 #include <assert.h>
 #include "talloc.h"
-#include "bstr/bstr.h"
+#include "misc/bstr.h"
 #include "common/msg.h"
 #include "codecs.h"
 

common/common.c

@@ -22,7 +22,7 @@
 #include <libavutil/common.h>
 
 #include "talloc.h"
-#include "bstr/bstr.h"
+#include "misc/bstr.h"
 #include "common/common.h"
 
 #define appendf(ptr, ...) \
@@ -171,7 +171,7 @@ static bool mp_parse_escape(void *talloc_ctx, bstr *dst, bstr *code)
     if (code->start[0] == 'x' && code->len >= 3) {
         bstr num = bstr_splice(*code, 1, 3);
         char c = bstrtoll(num, &num, 16);
-        if (!num.len)
+        if (num.len)
             return false;
         bstr_xappend(talloc_ctx, dst, (bstr){&c, 1});
         *code = bstr_cut(*code, 3);

common/common.h

@@ -24,7 +24,7 @@
 #include <stdbool.h>
 #include <stdint.h>
 
-#include "compat/compiler.h"
+#include "osdep/compiler.h"
 #include "talloc.h"
 
 // double should be able to represent this exactly
@@ -60,8 +60,8 @@ enum stream_type {
     STREAM_TYPE_COUNT,
 };
 
-extern const char *mpv_version;
-extern const char *mpv_builddate;
+extern const char *const mpv_version;
+extern const char *const mpv_builddate;
 
 char *mp_format_time(double time, bool fractions);
 char *mp_format_time_fmt(const char *fmt, double time);