Release 0.10.0

So far, this required using the client API to know it.

.travis.yml

@@ -47,7 +47,7 @@ notifications:
   email: false
   irc:
     channels:
-      - "irc.freenode.org#mpv-player-dev"
+      - "irc.freenode.org#mpv-devel"
     on_success: change
     on_failure: always
 

Copyright

@@ -2,9 +2,15 @@ mpv is a fork of mplayer2, which is a fork of MPlayer.
 
 mpv as a whole is licensed as GPL version 2 or later (see LICENSE). Most source
 files are GPLv2+, but some files are available under a more liberal license,
-such as LGPLv2+, BSD, MIT, ISC, and possibly others. Look at the copyright
+such as LGPLv2.1+, BSD, MIT, ISC, and possibly others. Look at the copyright
 header of each source file, and grep the sources for "Copyright" if you need
-to know details. Files without Copyright notice are licensed as LGPLv2+.
+to know details. Files without Copyright notice are licensed as LGPLv2.1+.
+
+All new contributions must be LGPLv2.1+ licensed, or if the changes are done on
+GPL code, must come with the implicit agreement that the project can relicense
+the code to LGPLv2.1+ at a later point without asking the contributor. (This
+is a safeguard for making potential relicensing of the project to LGPLv2.1+
+easier.) Using a more liberal license compatible to LGPLv2.1+ is also ok.
 
 For information about authors and contributors, consult the git log, which
 contains the complete SVN and CVS history as well.
@@ -17,3 +23,5 @@ MPlayer as a whole is licensed under GPLv2 (incompatible to GPLv3!), because
 some files are licensed to GPLv2 (and _not_ any later version of the license).
 In particular, this affects the file libmpdemux/demux_ty_osd.c. It is disabled
 under mplayer2, and has been removed from mpv.
+
+"v2.1+" in this context means "version 2.1 or later".

DOCS/client-api-changes.rst

@@ -4,9 +4,7 @@ Introduction
 This file lists all changes that can cause compatibility issues when using
 mpv through the client API (libmpv and ``client.h``). Since the client API
 interfaces to input handling (commands, properties) as well as command line
-options, this list is interesting for other uses of mpv, such as the Lua
-scripting interface, key bindings in ``input.rst``, or plain command line
-usage.
+options, you should also look at ``interface-changes.rst``.
 
 Normally, changes to the C API that are incompatible to previous iterations
 receive a major version bump (i.e. the first version number is increased),
@@ -20,15 +18,31 @@ The version number is the same as used for MPV_CLIENT_API_VERSION (see
 
 Also, read the section ``Compatibility`` in ``client.h``.
 
+Options, commands, properties
+=============================
+
+Changes to these are not listed here, but in ``interface-changes.rst``. (Before
+client API version 1.17, they were listed here partially.)
+
+This listing includes changes to the bare C API and behavior only, not what
+you can access with them.
+
 API changes
 ===========
 
 ::
 
- 1.17   - add MPV_FORMAT_BYTE_ARRAY
- --- mpv 0.9.0 will be released ---
+ ...    - add "GL_MP_D3D_interfaces" pseudo extension to make it possible to
+          use DXVA2 in OpenGL fullscreen mode in some situations
+ 1.19   - mpv_request_log_messages() now accepts "terminal-default" as parameter
+ 1.18   - add MPV_END_FILE_REASON_REDIRECT, and change behavior of
+          MPV_EVENT_END_FILE accordingly
+        - a bunch of interface-changes.rst changes
+ 1.17   - mpv_initialize() now blocks SIGPIPE (details see client.h)
+ --- mpv 0.9.0 is released ---
  1.16   - add mpv_opengl_cb_report_flip()
         - introduce mpv_opengl_cb_draw() and deprecate mpv_opengl_cb_render()
+        - add MPV_FORMAT_BYTE_ARRAY
  1.15   - mpv_initialize() will now load config files. This requires setting
           the "config" and "config-dir" options. In particular, it will load
           mpv.conf.

DOCS/client_api_examples/README.md

@@ -24,6 +24,11 @@ qml_direct
 Alternative example, which typically avoids a FBO indirection. Might be
 slightly faster, but is less flexible and harder to use.
 
+sdl
+---
+
+Show how to embed the mpv OpenGL renderer in SDL.
+
 simple
 ------
 

DOCS/client_api_examples/cocoa-openglcb/cocoa-openglcb.m

@@ -0,0 +1,304 @@
+// Plays a video from the command line in an opengl view in its own window.
+
+// Build with: clang -o cocoa-openglcb cocoa-openglcb.m `pkg-config --libs --cflags mpv` -framework Cocoa -framework OpenGL
+
+#import <mpv/client.h>
+#import <mpv/opengl_cb.h>
+
+#import <stdio.h>
+#import <stdlib.h>
+#import <OpenGL/gl.h>
+
+#import <Cocoa/Cocoa.h>
+
+
+static inline void check_error(int status)
+{
+    if (status < 0) {
+        printf("mpv API error: %s\n", mpv_error_string(status));
+        exit(1);
+    }
+}
+
+static void *get_proc_address(void *ctx, const char *name)
+{
+    CFStringRef symbolName = CFStringCreateWithCString(kCFAllocatorDefault, name, kCFStringEncodingASCII);
+    void *addr = CFBundleGetFunctionPointerForName(CFBundleGetBundleWithIdentifier(CFSTR("com.apple.opengl")), symbolName);
+    CFRelease(symbolName);
+    return addr;
+}
+
+static void glupdate(void *ctx);
+
+@interface MpvClientOGLView : NSOpenGLView
+@property mpv_opengl_cb_context *mpvGL;
+- (instancetype)initWithFrame:(NSRect)frame;
+- (void)drawRect;
+- (void)fillBlack;
+@end
+
+@implementation MpvClientOGLView
+- (instancetype)initWithFrame:(NSRect)frame
+{
+    // make sure the pixel format is double buffered so we can use
+    // [[self openGLContext] flushBuffer].
+    NSOpenGLPixelFormatAttribute attributes[] = {
+        NSOpenGLPFADoubleBuffer,
+        0
+    };
+    self = [super initWithFrame:frame
+                    pixelFormat:[[NSOpenGLPixelFormat alloc] initWithAttributes:attributes]];
+
+    if (self) {
+        [self setAutoresizingMask:NSViewWidthSizable|NSViewHeightSizable];
+        // swap on vsyncs
+        GLint swapInt = 1;
+        [[self openGLContext] setValues:&swapInt forParameter:NSOpenGLCPSwapInterval];
+        [[self openGLContext] makeCurrentContext];
+        self.mpvGL = nil;
+    }
+    return self;
+}
+
+- (void)fillBlack
+{
+    glClearColor(0, 0, 0, 0);
+    glClear(GL_COLOR_BUFFER_BIT);
+}
+
+- (void)drawRect
+{
+    if (self.mpvGL)
+        mpv_opengl_cb_draw(self.mpvGL, 0, self.bounds.size.width, -self.bounds.size.height);
+    else
+        [self fillBlack];
+    [[self openGLContext] flushBuffer];
+}
+
+- (void)drawRect:(NSRect)dirtyRect
+{
+    [self drawRect];
+}
+@end
+
+@interface CocoaWindow : NSWindow
+@property(retain, readonly) MpvClientOGLView *glView;
+@property(retain, readonly) NSButton *pauseButton;
+@end
+
+@implementation CocoaWindow
+- (BOOL)canBecomeMainWindow { return YES; }
+- (BOOL)canBecomeKeyWindow { return YES; }
+- (void)initOGLView {
+    NSRect bounds = [[self contentView] bounds];
+    // window coordinate origin is bottom left
+    NSRect glFrame = NSMakeRect(bounds.origin.x, bounds.origin.y + 30, bounds.size.width, bounds.size.height - 30);
+    _glView = [[MpvClientOGLView alloc] initWithFrame:glFrame];
+    [self.contentView addSubview:_glView];
+
+    NSRect buttonFrame = NSMakeRect(bounds.origin.x, bounds.origin.y, 60, 30);
+    _pauseButton = [[NSButton alloc] initWithFrame:buttonFrame];
+    _pauseButton.buttonType = NSToggleButton;
+    // button target has to be the delegate (it holds the mpv context
+    // pointer), so that's set later.
+    _pauseButton.action = @selector(togglePause:);
+    _pauseButton.title = @"Pause";
+    _pauseButton.alternateTitle = @"Play";
+    [self.contentView addSubview:_pauseButton];
+}
+@end
+
+@interface AppDelegate : NSObject <NSApplicationDelegate>
+{
+    mpv_handle *mpv;
+    dispatch_queue_t queue;
+    CocoaWindow *window;
+}
+@end
+
+static void wakeup(void *);
+
+@implementation AppDelegate
+
+- (void)createWindow {
+
+    int mask = NSTitledWindowMask|NSClosableWindowMask|
+               NSMiniaturizableWindowMask|NSResizableWindowMask;
+
+    window = [[CocoaWindow alloc]
+        initWithContentRect:NSMakeRect(0, 0, 1280, 720)
+                  styleMask:mask
+                    backing:NSBackingStoreBuffered
+                      defer:NO];
+
+    // force a minimum size to stop opengl from exploding.
+    [window setMinSize:NSMakeSize(200, 200)];
+    [window initOGLView];
+    [window setTitle:@"cocoa-openglcb example"];
+    [window makeMainWindow];
+    [window 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: @"quit" action:@selector(terminate:) keyEquivalent:@"q"];
+    [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];
+    window.pauseButton.target = self;
+
+    mpv = mpv_create();
+    if (!mpv) {
+        printf("failed creating context\n");
+        exit(1);
+    }
+
+    check_error(mpv_set_option_string(mpv, "input-media-keys", "yes"));
+    // request important errors
+    check_error(mpv_request_log_messages(mpv, "warn"));
+
+    check_error(mpv_initialize(mpv));
+    check_error(mpv_set_option_string(mpv, "vo", "opengl-cb"));
+    mpv_opengl_cb_context *mpvGL = mpv_get_sub_api(mpv, MPV_SUB_API_OPENGL_CB);
+    if (!mpvGL) {
+        puts("libmpv does not have the opengl-cb sub-API.");
+        exit(1);
+    }
+    // pass the mpvGL context to our view
+    window.glView.mpvGL = mpvGL;
+    int r = mpv_opengl_cb_init_gl(mpvGL, NULL, get_proc_address, NULL);
+    if (r < 0) {
+        puts("gl init has failed.");
+        exit(1);
+    }
+    mpv_opengl_cb_set_update_callback(mpvGL, glupdate, (__bridge void *)window.glView);
+
+    // Deal with MPV in the background.
+    queue = dispatch_queue_create("mpv", DISPATCH_QUEUE_SERIAL);
+    dispatch_async(queue, ^{
+        // 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));
+    });
+}
+
+static void glupdate(void *ctx)
+{
+    MpvClientOGLView *glView = (__bridge MpvClientOGLView *)ctx;
+    // I'm still not sure what the best way to handle this is, but this
+    // works.
+    dispatch_async(dispatch_get_main_queue(), ^{
+        [glView drawRect];
+    });
+}
+
+- (void) handleEvent:(mpv_event *)event
+{
+    switch (event->event_id) {
+    case MPV_EVENT_SHUTDOWN: {
+        mpv_detach_destroy(mpv);
+        mpv_opengl_cb_uninit_gl(window.glView.mpvGL);
+        mpv = NULL;
+        printf("event: shutdown\n");
+        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);
+    }
+
+    default:
+        printf("event: %s\n", mpv_event_name(event->event_id));
+    }
+}
+
+- (void)togglePause:(NSButton *)button
+{
+    if (mpv) {
+        switch (button.state) {
+            case NSOffState:
+                {
+                    int pause = 0;
+                    mpv_set_property(mpv, "pause", MPV_FORMAT_FLAG, &pause);
+                }
+                break;
+            case NSOnState:
+                {
+                    int pause = 1;
+                    mpv_set_property(mpv, "pause", MPV_FORMAT_FLAG, &pause);
+                }
+                break;
+            default:
+                NSLog(@"This should never happen.");
+        }
+    }
+}
+
+- (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];
+}
+
+// quit when the window is closed.
+- (BOOL)applicationShouldTerminateAfterLastWindowClosed:(NSApplication *)app
+{
+    return YES;
+}
+
+- (NSApplicationTerminateReply)applicationShouldTerminate:(NSApplication *)sender
+{
+    NSLog(@"Terminating.");
+    const char *args[] = {"quit", NULL};
+    mpv_command(mpv, args);
+    [window.glView clearGLContext];
+    return NSTerminateNow;
+}
+
+@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 0;
+}

DOCS/client_api_examples/cocoa/cocoabasic.m

@@ -1,9 +1,7 @@
-// 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.
+// Plays a video from the command line in a view provided by the client
+// application.
 
-// Build with: clang -o cocoabasic cocoabasic.m `pkg-config --libs --cflags mpv`
+// Build with: clang -o cocoabasic cocoabasic.m `pkg-config --libs --cflags mpv` -framework cocoa
 
 #include <mpv/client.h>
 
@@ -207,5 +205,5 @@ int main(int argc, const char * argv[]) {
         app.delegate = delegate;
         [app run];
     }
-    return EXIT_SUCCESS;
+    return 0;
 }

DOCS/client_api_examples/qml/mpvtest.pro

@@ -3,6 +3,7 @@ QT += qml quick
 HEADERS += main.h
 SOURCES += main.cpp
 
+QT_CONFIG -= no-pkg-config
 CONFIG += link_pkgconfig debug
 PKGCONFIG += mpv
 

DOCS/client_api_examples/qml_direct/mpvtest.pro

@@ -3,6 +3,7 @@ QT += qml quick
 HEADERS += main.h
 SOURCES += main.cpp
 
+QT_CONFIG -= no-pkg-config
 CONFIG += link_pkgconfig debug
 PKGCONFIG += mpv
 

DOCS/client_api_examples/qt/qtexample.pro

@@ -5,6 +5,7 @@ greaterThan(QT_MAJOR_VERSION, 4): QT += widgets
 TARGET = qtexample
 TEMPLATE = app
 
+QT_CONFIG -= no-pkg-config
 CONFIG += link_pkgconfig debug
 PKGCONFIG += mpv
 

DOCS/client_api_examples/sdl/main.c

@@ -0,0 +1,157 @@
+// Build with: gcc -o main main.c `pkg-config --libs --cflags mpv sdl2`
+
+#include <stddef.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include <SDL.h>
+
+#include <mpv/client.h>
+#include <mpv/opengl_cb.h>
+
+static Uint32 wakeup_on_mpv_redraw, wakeup_on_mpv_events;
+
+static void die(const char *msg)
+{
+    fprintf(stderr, "%s\n", msg);
+    exit(1);
+}
+
+static void *get_proc_address_mpv(void *fn_ctx, const char *name)
+{
+    return SDL_GL_GetProcAddress(name);
+}
+
+static void on_mpv_events(void *ctx)
+{
+    SDL_Event event = {.type = wakeup_on_mpv_events};
+    SDL_PushEvent(&event);
+}
+
+static void on_mpv_redraw(void *ctx)
+{
+    SDL_Event event = {.type = wakeup_on_mpv_redraw};
+    SDL_PushEvent(&event);
+}
+
+int main(int argc, char *argv[])
+{
+    if (argc != 2)
+        die("pass a single media file as argument");
+
+    mpv_handle *mpv = mpv_create();
+    if (!mpv)
+        die("context init failed");
+
+    // Some minor options can only be set before mpv_initialize().
+    if (mpv_initialize(mpv) < 0)
+        die("mpv init failed");
+
+    if (SDL_Init(SDL_INIT_VIDEO) < 0)
+        die("SDL init failed");
+
+    SDL_Window *window =
+        SDL_CreateWindow("hi", SDL_WINDOWPOS_CENTERED, SDL_WINDOWPOS_CENTERED,
+                         1000, 500, SDL_WINDOW_OPENGL | SDL_WINDOW_SHOWN);
+    if (!window)
+        die("failed to create SDL window");
+
+    // The OpenGL API is somewhat separate from the normal mpv API. This only
+    // returns NULL if no OpenGL support is compiled.
+    mpv_opengl_cb_context *mpv_gl = mpv_get_sub_api(mpv, MPV_SUB_API_OPENGL_CB);
+    if (!mpv_gl)
+        die("failed to create mpv GL API handle");
+
+    SDL_GLContext glcontext = SDL_GL_CreateContext(window);
+    if (!glcontext)
+        die("failed to create SDL GL context");
+
+    // This makes mpv use the currently set GL context. It will use the callback
+    // to resolve GL builtin functions, as well as extensions.
+    if (mpv_opengl_cb_init_gl(mpv_gl, NULL, get_proc_address_mpv, NULL) < 0)
+        die("failed to initialize mpv GL context");
+
+    // Actually using the opengl_cb state has to be explicitly requested.
+    // Otherwise, mpv will create a separate platform window.
+    if (mpv_set_option_string(mpv, "vo", "opengl-cb") < 0)
+        die("failed to set VO");
+
+    // We use events for thread-safe notification of the SDL main loop.
+    // Generally, the wakeup callbacks (set further below) should do as least
+    // work as possible, and merely wake up another thread to do actual work.
+    // On SDL, waking up the mainloop is the ideal course of action. SDL's
+    // SDL_PushEvent() is thread-safe, so we use that.
+    wakeup_on_mpv_redraw = SDL_RegisterEvents(1);
+    wakeup_on_mpv_events = SDL_RegisterEvents(1);
+    if (wakeup_on_mpv_redraw == (Uint32)-1 || wakeup_on_mpv_events == (Uint32)-1)
+        die("could not register events");
+
+    // When normal mpv events are available.
+    mpv_set_wakeup_callback(mpv, on_mpv_events, NULL);
+
+    // When a new frame should be drawn with mpv_opengl_cb_draw().
+    // (Separate from the normal event handling mechanism for the sake of
+    //  users which run OpenGL on a different thread.)
+    mpv_opengl_cb_set_update_callback(mpv_gl, on_mpv_redraw, NULL);
+
+    // Play this file. Note that this starts playback asynchronously.
+    const char *cmd[] = {"loadfile", argv[1], NULL};
+    mpv_command(mpv, cmd);
+
+    while (1) {
+        SDL_Event event;
+        if (SDL_WaitEvent(&event) != 1)
+            die("event loop error");
+        int redraw = 0;
+        switch (event.type) {
+        case SDL_QUIT:
+            goto done;
+        case SDL_WINDOWEVENT:
+            if (event.window.event == SDL_WINDOWEVENT_EXPOSED)
+                redraw = 1;
+            break;
+        case SDL_KEYDOWN:
+            if (event.key.keysym.sym == SDLK_SPACE)
+                mpv_command_string(mpv, "cycle pause");
+            break;
+        default:
+            // Happens when a new video frame should be rendered, or if the
+            // current frame has to be redrawn e.g. due to OSD changes.
+            if (event.type == wakeup_on_mpv_redraw)
+                redraw = 1;
+            // Happens when at least 1 new event is in the mpv event queue.
+            if (event.type == wakeup_on_mpv_events) {
+                // Handle all remaining mpv events.
+                while (1) {
+                    mpv_event *mp_event = mpv_wait_event(mpv, 0);
+                    if (mp_event->event_id == MPV_EVENT_NONE)
+                        break;
+                    printf("event: %s\n", mpv_event_name(mp_event->event_id));
+                }
+            }
+        }
+        if (redraw) {
+            int w, h;
+            SDL_GetWindowSize(window, &w, &h);
+            // Note:
+            // - The 0 is the FBO to use; 0 is the default framebuffer (i.e.
+            //   render to the window directly.
+            // - The negative height tells mpv to flip the coordinate system.
+            // - If you do not want the video to cover the whole screen, or want
+            //   to apply any form of fancy transformation, you will have to
+            //   render to a FBO.
+            // - See opengl_cb.h on what OpenGL environment mpv expects, and
+            //   other API details.
+            mpv_opengl_cb_draw(mpv_gl, 0, w, -h);
+            SDL_GL_SwapWindow(window);
+        }
+    }
+done:
+
+    // Destroy the GL renderer and all of the GL objects it allocated. If video
+    // is still running, the video track will be deselected.
+    mpv_opengl_cb_uninit_gl(mpv_gl);
+
+    mpv_terminate_destroy(mpv);
+    return 0;
+}

DOCS/compile-windows.md

@@ -134,10 +134,10 @@ pacman -S git pkg-config python3 mingw-w64-x86_64-gcc
 
 # Install the most important MinGW-w64 dependencies. libass, libbluray and
 # lcms2 are also pulled in as dependencies of ffmpeg.
-pacman -S mingw-w64-x86_64-ffmpeg mingw-w64-x86_64-libjpeg-turbo mingw-w64-x86_64-lua
+pacman -S mingw-w64-x86_64-ffmpeg mingw-w64-x86_64-libjpeg-turbo mingw-w64-x86_64-lua51
 
 # Install additional (optional) dependencies
-pacman -S mingw-w64-x86_64-libdvdnav mingw-w64-x86_64-mpg123 mingw-w64-x86_64-libguess
+pacman -S mingw-w64-x86_64-libdvdnav mingw-w64-x86_64-libguess
 ```
 
 For a 32-bit build, install ``mingw-w64-i686-*`` packages instead.

DOCS/contribute.md

@@ -4,8 +4,8 @@ How to contribute
 General
 -------
 
-The main contact for mpv development is IRC, specifically #mpv-player
-and #mpv-player-dev on Freenode.
+The main contact for mpv development is IRC, specifically #mpv
+and #mpv-devel on Freenode.
 
 Sending patches
 ---------------
@@ -54,6 +54,9 @@ Sending patches
   change in the same commit.
 - If you add a new command line option, document it in options.rst. If you
   add a new input property, document it in input.rst.
+- All new code must be LGPLv2.1+ licensed, or come with the implicit agreement
+  that it will be relicensed to LGPLv2.1+ later (see ``Copyright`` in the
+  repository root directory).
 
 Code formatting
 ---------------

DOCS/interface-changes.rst

@@ -0,0 +1,88 @@
+Introduction
+============
+
+mpv provides access to its internal via the following means:
+
+- options
+- commands
+- properties
+- events
+
+All of these are important for interfacing both with end users and API users
+(which include Lua scripts, libmpv, and the JSON IPC). As such, they constitute
+a large part of the user interface and APIs.
+
+This document lists changes to them. New changes are added to the top.
+
+Interface changes
+=================
+
+::
+
+ --- mpv 0.10.0 will be released ---
+    - add --video-aspect-method option
+    - add --playlist-pos option
+    - add --video-sync* options
+      "display-sync-active" property
+      "vo-missed-frame-count" property
+      "audio-speed-correction" and "video-speed-correction" properties
+    - remove --demuxer-readahead-packets and --demuxer-readahead-bytes
+      add --demuxer-max-packets and --demuxer-max-bytes
+      (the new options are not replacement and have very different semantics)
+    - change "video-aspect" property: always settable, even if no video is
+      running; always return the override - if no override is set, return
+      the video's aspect ratio
+    - remove disc-nav (DVD, BD) related properties and commands
+    - add "option-info/<name>/set-locally" property
+    - add --cache-backbuffer; change --cache-default default to 75MB
+      the new total cache size is the sum of backbuffer and the cache size
+      specified by --cache-default or --cache
+    - add ``track-list/N/audio-channels`` property
+    - change --screenshot-tag-colorspace default value
+    - add --stretch-image-subs-to-screen
+    - add "playlist/N/title" property
+    - add --video-stereo-mode=no to disable auto-conversions
+    - add --force-seekable, and change default seekability in some cases
+    - add vf yadif/vavpp/vdpaupp interlaced-only suboptions
+      Also, the option is enabled by default (Except vf_yadif, which has
+      it enabled only if it's inserted by the deinterlace property.)
+    - add --hwdec-preload
+    - add ao coreaudio exclusive suboption
+    - add ``track-list/N/forced`` property
+    - add audio-params/channel-count and ``audio-params-out/channel-count props.
+    - add af volume replaygain-fallback suboption
+    - add video-params/stereo-in property
+    - add "keypress", "keydown", and "keyup" commands
+    - deprecate --ad-spdif-dtshd and enabling passthrough via --ad
+      add --audio-spdif as replacement
+    - remove "get_property" command
+    - remove --slave-broken
+    - add vo opengl custom shader suboptions (source-shader, scale-shader,
+      pre-shaders, post-shaders)
+    - completely change how the hwdec properties work:
+        - "hwdec" now reflects the --hwdec option
+        - "hwdec-detected" does partially what the old "hwdec" property did
+          (and also, "detected-hwdec" is removed)
+        - "hwdec-active" is added
+    - add protocol-list property
+    - deprecate audio-samplerate and audio-channels properties
+      (audio-params sub-properties are the replacement)
+    - add audio-params and audio-out-params properties
+    - deprecate "audio-format" property, replaced with "audio-codec-name"
+    - deprecate --media-title, replaced with --force-media-title
+    - deprecate "length" property, replaced with "duration"
+    - change volume property:
+        - the value 100 is now always "unchanged volume" - with softvol, the
+          range is 0 to --softvol-max, without it is 0-100
+        - the minimum value of --softvol-max is raised to 100
+    - remove vo opengl npot suboption
+    - add relative seeking by percentage to "seek" command
+    - add playlist_shuffle command
+    - add --force-window=immediate
+    - add ao coreaudio change-physical-format suboption
+    - remove vo opengl icc-cache suboption, add icc-cache-dir suboption
+    - add --screenshot-directory
+    - add --screenshot-high-bit-depth
+    - add --screenshot-jpeg-source-chroma
+    - default action for "rescan_external_files" command changes
+ --- mpv 0.9.0 is released ---

DOCS/man/af.rst

@@ -263,13 +263,6 @@ Available filters are:
     used to do conversion itself, unlike this one which lets the filter system
     handle the conversion.
 
-``convert24``
-    Filter for internal use only. Converts between 24-bit and 32-bit sample
-    formats.
-
-``convertsign``
-    Filter for internal use only. Converts between signed/unsigned formats.
-
 ``volume[=<volumedb>[:...]]``
     Implements software volume control. Use this filter with caution since it
     can reduce the signal to noise ratio of the sound. In most cases it is
@@ -294,6 +287,10 @@ Available filters are:
     ``replaygain-clip=yes|no``
         Prevent clipping caused by replaygain by automatically lowering the
         gain (default). Use ``replaygain-clip=no`` to disable this.
+    ``replaygain-fallback``
+        Gain in dB to apply if the file has no replay gain tags. This option
+        is always applied if the replaygain logic is somehow inactive. If this
+        is applied, no other replaygain options are applied.
     ``softclip``
         Turns soft clipping on. Soft-clipping can make the
         sound more smooth if very high volume levels are used. Enable this

DOCS/man/ao.rst

@@ -55,7 +55,7 @@ Available audio output drivers are:
         Allow output of non-interleaved formats (if the audio decoder uses
         this format). Currently disabled by default, because some popular
         ALSA plugins are utterly broken with non-interleaved formats.
-    ``ingore-chmap``
+    ``ignore-chmap``
         Don't read or set the channel map of the ALSA device - only request the
         required number of channels, and then pass the audio as-is to it. This
         option most likely should not be used. It can be useful for debugging,
@@ -162,12 +162,23 @@ Available audio output drivers are:
     Automatically redirects to ``coreaudio_exclusive`` when playing compressed
     formats.
 
+    ``change-physical-format=<yes|no>``
+        Change the physical format to one similar to the requested audio format
+        (default: no). This has the advantage that multichannel audio output
+        will actually work. The disadvantage is that it will change the
+        system-wide audio settings. This is equivalent to changing the ``Format``
+        setting in the ``Audio Devices`` dialog in the ``Audio MIDI Setup``
+        utility. Note that this does not effect the selected speaker setup.
+
+    ``exclusive``
+        Use exclusive mode access. This merely redirects to
+        ``coreaudio_exclusive``, but should be preferred over using that AO
+        directly.
+
 ``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).
-
 ``openal``
     Experimental OpenAL audio output driver
 
@@ -260,6 +271,10 @@ Available audio output drivers are:
     ``broken-delay``
         Simulate broken audio drivers, which don't report latency correctly.
 
+    ``channel-layouts``
+        If not empty, this is a ``,`` separated list of channel layouts the
+        AO allows. This can be used to test channel layout selection.
+
 ``pcm``
     Raw PCM/WAVE file writer audio output
 

DOCS/man/encode.rst

@@ -111,7 +111,7 @@ You can encode files from one format/codec to another using this facility.
 
     .. admonition:: Examples
 
-        ``"--ovc=mpeg4 --oacopts=qscale=5"``
+        ``"--ovc=mpeg4 --ovcopts=qscale=5"``
             selects constant quantizer scale 5 for MPEG-4 encoding.
 
         ``"--ovc=libx264 --ovcopts=crf=23"``

DOCS/man/input.rst

@@ -65,11 +65,11 @@ C-style escaping can be used.
 
 You can bind multiple commands to one key. For example:
 
-| a show_text "command 1" ; show_text "command 2"
+| a show-text "command 1" ; show-text "command 2"
 
 It's also possible to bind a command to a sequence of keys:
 
-| a-b-c show_text "command run after a, b, c have been pressed"
+| a-b-c show-text "command run after a, b, c have been pressed"
 
 (This is not shown in the general command syntax.)
 
@@ -86,7 +86,7 @@ List of Input Commands
     disabling default bindings, without disabling all bindings with
     ``--no-input-default-bindings``.
 
-``seek <seconds> [relative|absolute|absolute-percent|exact|keyframes]``
+``seek <seconds> [relative|absolute|absolute-percent|relative-percent|exact|keyframes]``
     Change the playback position. By default, seeks by a relative amount of
     seconds.
 
@@ -98,6 +98,8 @@ List of Input Commands
         Seek to a given time.
     absolute-percent
         Seek to a given percent position.
+    relative-percent
+        Seek relative to current position in percent.
     keyframes
         Always restart playback at keyframe boundaries (fast).
     exact
@@ -112,25 +114,25 @@ List of Input Commands
     3rd parameter (essentially using a space instead of ``+``). The 3rd
     parameter is still parsed, but is considered deprecated.
 
-``revert_seek [mode]``
+``revert-seek [mode]``
     Undoes the ``seek`` command, and some other commands that seek (but not
     necessarily all of them). Calling this command once will jump to the
     playback position before the seek. Calling it a second time undoes the
-    ``revert_seek`` command itself. This only works within a single file.
+    ``revert-seek`` command itself. This only works within a single file.
 
     The first argument is optional, and can change the behavior:
 
     mark
-        Mark the current time position. The next normal ``revert_seek`` command
+        Mark the current time position. The next normal ``revert-seek`` command
         will seek back to this point, no matter how many seeks happened since
         last time.
 
     Using it without any arguments gives you the default behavior.
 
-``frame_step``
+``frame-step``
     Play one frame, then pause. Does nothing with audio-only playback.
 
-``frame_back_step``
+``frame-back-step``
     Go back by one frame, then pause. Note that this can be very slow (it tries
     to be precise, not fast), and sometimes fails to behave as expected. How
     well this works depends on whether precise seeking works correctly (e.g.
@@ -180,7 +182,7 @@ List of Input Commands
         frame was dropped. This flag can be combined with the other flags,
         e.g. ``video+each-frame``.
 
-``screenshot_to_file "<filename>" [subtitles|video|window]``
+``screenshot-to-file "<filename>" [subtitles|video|window]``
     Take a screenshot and save it to a given file. The format of the file will
     be guessed by the extension (and ``--screenshot-format`` is ignored - the
     behavior when the extension is missing or unknown is arbitrary).
@@ -192,7 +194,7 @@ List of Input Commands
     Like all input command parameters, the filename is subject to property
     expansion as described in `Property Expansion`_.
 
-``playlist_next [weak|force]``
+``playlist-next [weak|force]``
     Go to the next entry on the playlist.
 
     weak (default)
@@ -200,7 +202,7 @@ List of Input Commands
     force
         Terminate playback if there are no more files on the playlist.
 
-``playlist_prev [weak|force]``
+``playlist-prev [weak|force]``
     Go to the previous entry on the playlist.
 
     weak (default)
@@ -230,22 +232,26 @@ List of Input Commands
 ``loadlist "<playlist>" [replace|append]``
     Load the given playlist file (like ``--playlist``).
 
-``playlist_clear``
+``playlist-clear``
     Clear the playlist, except the currently played file.
 
-``playlist_remove current|<index>``
+``playlist-remove current|<index>``
     Remove the playlist entry at the given index. Index values start counting
     with 0. The special value ``current`` removes the current entry. Note that
     removing the current entry also stops playback and starts playing the next
     entry.
 
-``playlist_move <index1> <index2>``
+``playlist-move <index1> <index2>``
     Move the playlist entry at index1, so that it takes the place of the
     entry index2. (Paradoxically, the moved playlist entry will not have
     the index value index2 after moving if index1 was lower than index2,
     because index2 refers to the target entry, not the index the entry
     will have after moving.)
 
+``playlist-shuffle``
+    Shuffle the playlist. This is similar to what is done on start if the
+    ``--shuffle`` option is used.
+
 ``run "command" "arg1" "arg2" ...``
     Run the given command. Unlike in MPlayer/mplayer2 and earlier versions of
     mpv (0.2.x and older), this doesn't call the shell. Instead, the command
@@ -271,12 +277,12 @@ List of Input Commands
 ``quit [<code>]``
     Exit the player. If an argument is given, it's used as process exit code.
 
-``quit_watch_later [<code>]``
+``quit-watch-later [<code>]``
     Exit player, and store current playback position. Playing that file later
     will seek to the previous position on start. The (optional) argument is
     exactly as in the ``quit`` command.
 
-``sub_add "<file>" [<flags> [<title> [<lang>]]]``
+``sub-add "<file>" [<flags> [<title> [<lang>]]]``
     Load the given subtitle file. It is selected as current subtitle after
     loading.
 
@@ -303,27 +309,27 @@ List of Input Commands
     The ``lang`` argument sets the track language, and can also influence
     stream selection with ``flags`` set to ``auto``.
 
-``sub_remove [<id>]``
+``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>]``
+``sub-reload [<id>]``
     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.
 
-``sub_step <skip>``
+``sub-step <skip>``
     Change subtitle timing such, that the subtitle event after the next
     ``<skip>`` subtitle events is displayed. ``<skip>`` can be negative to step
     backwards.
 
-``sub_seek <skip>``
+``sub-seek <skip>``
     Seek to the next (skip set to 1) or the previous (skip set to -1) subtitle.
-    This is similar to ``sub_step``, except that it seeks video and audio
+    This is similar to ``sub-step``, except that it seeks video and audio
     instead of adjusting the subtitle delay.
 
-    Like with ``sub_step``, this works with external text subtitles only. For
+    Like with ``sub-step``, this works with external text subtitles only. For
     embedded text subtitles (like with Matroska), this works only with subtitle
     events that have already been displayed.
 
@@ -331,11 +337,11 @@ List of Input Commands
     Toggle OSD level. If ``<level>`` is specified, set the OSD mode
     (see ``--osd-level`` for valid values).
 
-``print_text "<string>"``
+``print-text "<string>"``
     Print text to stdout. The string can contain properties (see
     `Property Expansion`_).
 
-``show_text "<string>" [<duration>|- [<level>]]``
+``show-text "<string>" [<duration>|- [<level>]]``
     Show text on the OSD. The string can contain properties, which are expanded
     as described in `Property Expansion`_. This can be used to show playback
     time, filename, and so on.
@@ -347,24 +353,12 @@ List of Input Commands
     <level>
         The minimum OSD level to show the text at (see ``--osd-level``).
 
-``show_progress``
+``show-progress``
     Show the progress bar, the elapsed time and the total duration of the file
     on the OSD.
 
-``discnav "<command>"``
-    Send a menu control command to the DVD/BD menu implementation. The following
-    commands are defined: ``up``, ``down``, ``left``, ``right``,
-    ``menu`` (request to enter menu), ``prev`` (previous screen),
-    ``select`` (activate current button), ``mouse`` (the mouse was clicked),
-    ``mouse_move`` (the mouse cursor changed position).
-
-    ``mouse_move`` will use the current mouse position.
-
-    Note that while the menu is active, the input section ``discnav-menu`` will
-    be enabled, so different key bindings can be mapped for menu mode.
-
-``write_watch_later_config``
-    Write the resume config file that the ``quit_watch_later`` command writes,
+``write-watch-later-config``
+    Write the resume config file that the ``quit-watch-later`` command writes,
     but continue playback normally.
 
 ``stop``
@@ -389,30 +383,47 @@ List of Input Commands
     <double>
         The mouse event represents double-click.
 
-``audio_add "<file>" [<flags> [<title> [<lang>]]]``
-    Load the given audio file. See ``sub_add`` command.
+``keypress <key_name>``
+    Send a key event through mpv's input handler, triggering whatever
+    behavior is configured to that key. ``key_name`` uses the ``input.conf``
+    naming scheme for keys and modifiers. Useful for the client API: key events
+    can be sent to libmpv to handle internally.
 
-``audio_remove [<id>]``
-    Remove the given audio track. See ``sub_remove`` command.
+``keydown <key_name>``
+    Similar to ``keypress``, but sets the ``KEYDOWN`` flag so that if the key is
+    bound to a repeatable command, it will be run repeatedly with mpv's key
+    repeat timing until the ``keyup`` command is called.
 
-``audio_reload [<id>]``
-    Reload the given audio tracks. See ``sub_reload`` command.
+``keyup [<key_name>]``
+    Set the ``KEYUP`` flag, stopping any repeated behavior that had been
+    triggered. ``key_name`` is optional. If ``key_name`` is not given or is an
+    empty string, ``KEYUP`` will be set on all keys. Otherwise, ``KEYUP`` will
+    only be set on the key specified by ``key_name``.
 
-``rescan_external_files [<mode>]``
+``audio-add "<file>" [<flags> [<title> [<lang>]]]``
+    Load the given audio file. See ``sub-add`` command.
+
+``audio-remove [<id>]``
+    Remove the given audio track. See ``sub-remove`` command.
+
+``audio-reload [<id>]``
+    Reload the given audio tracks. See ``sub-reload`` command.
+
+``rescan-external-files [<mode>]``
     Rescan external files according to the current ``--sub-auto`` and
     ``--audio-file-auto`` settings. This can be used to auto-load external
     files *after* the file was loaded.
 
     The ``mode`` argument is one of the following:
 
-    <keep-selection> (default)
-        Do not change current track selections.
-
-    <reselect>
-        Select the default audio and video streams, which typically selects
+    <reselect> (default)
+        Select the default audio and subtitle streams, which typically selects
         external files with highest preference. (The implementation is not
         perfect, and could be improved on request.)
 
+    <keep-selection>
+        Do not change current track selections.
+
 
 Input Commands that are Possibly Subject to Change
 --------------------------------------------------
@@ -455,7 +466,7 @@ Input Commands that are Possibly Subject to Change
 
     The ``vf`` command shows the list of requested filters on the OSD after
     changing the filter chain. This is roughly equivalent to
-    ``show_text ${vf}``. Note that auto-inserted filters for format conversion
+    ``show-text ${vf}``. Note that auto-inserted filters for format conversion
     are not shown on the list, only what was requested by the user.
 
     Normally, the commands will check whether the video chain is recreated
@@ -470,7 +481,7 @@ Input Commands that are Possibly Subject to Change
         - ``b vf set ""`` remove all video filters on ``b``
         - ``c vf toggle lavfi=gradfun`` toggle debanding on ``c``
 
-``cycle_values ["!reverse"] <property> "<value1>" "<value2>" ...``
+``cycle-values ["!reverse"] <property> "<value1>" "<value2>" ...``
     Cycle through a list of values. Each invocation of the command will set the
     given property to the next value in the list. The command maintains an
     internal counter which value to pick next, and which is initially 0. It is
@@ -488,7 +499,7 @@ Input Commands that are Possibly Subject to Change
     Note that there is a static limit of (as of this writing) 10 arguments
     (this limit could be raised on demand).
 
-``enable_section "<section>" [default|exclusive]``
+``enable-section "<section>" [flags]``
     Enable all key bindings in the named input section.
 
     The enabled input sections form a stack. Bindings in sections on the top of
@@ -497,14 +508,45 @@ Input Commands that are Possibly Subject to Change
     implicitly removed beforehand. (A section cannot be on the stack more than
     once.)
 
-    If ``exclusive`` is specified as second argument, all sections below the
-    newly enabled section are disabled. They will be re-enabled as soon as
-    all exclusive sections above them are removed.
+    The ``flags`` parameter can be a combination (separated by ``+``) of the
+    following flags:
 
-``disable_section "<section>"``
-    Disable the named input section. Undoes ``enable_section``.
+    <exclusive>
+        All sections enabled before the newly enabled section are disabled.
+        They will be re-enabled as soon as all exclusive sections above them
+        are removed. In other words, the new section shadows all previous
+        sections.
+    <allow-hide-cursor>
+        This feature can't be used through the public API.
+    <allow-vo-dragging>
+        Same.
 
-``overlay_add <id> <x> <y> "<file>" <offset> "<fmt>" <w> <h> <stride>``
+``disable-section "<section>"``
+    Disable the named input section. Undoes ``enable-section``.
+
+``define-section "<section>" "<contents>" [default|forced]``
+    Create a named input section, or replace the contents of an already existing
+    input section. The ``contents`` parameter uses the same syntax as the
+    ``input.conf`` file (except that using the section syntax in it is not
+    allowed), including the need to separate bindings with a newline character.
+
+    If the ``contents`` parameter is an empty string, the section is removed.
+
+    The section with the name ``default`` is the normal input section.
+
+    In general, input sections have to be enabled with the ``enable-section``
+    command, or they are ignored.
+
+    The last parameter has the following meaning:
+
+    <default> (also used if parameter omitted)
+        Use a key binding defined by this section only if the user hasn't
+        already bound this key to a command.
+    <forced>
+        Always bind a key. (The input section that was made active most recently
+        wins if there are ambiguities.)
+
+``overlay-add <id> <x> <y> "<file>" <offset> "<fmt>" <w> <h> <stride>``
     Add an OSD overlay sourced from raw data. This might be useful for scripts
     and applications controlling mpv, and which want to display things on top
     of the video window.
@@ -520,7 +562,7 @@ Input Commands that are Possibly Subject to Change
     ``id`` is an integer between 0 and 63 identifying the overlay element. The
     ID can be used to add multiple overlay parts, update a part by using this
     command with an already existing ID, or to remove a part with
-    ``overlay_remove``. Using a previously unused ID will add a new overlay,
+    ``overlay-remove``. Using a previously unused ID will add a new overlay,
     while reusing an ID will update it. (Future directions: there should be
     something to ensure different programs wanting to create overlays don't
     conflict with each others, should that ever be needed.)
@@ -534,11 +576,11 @@ Input Commands that are Possibly Subject to Change
     vdpau), so no actual copying is involved. Truncating the source file while
     the overlay is active will crash the player. You shouldn't change the data
     while the overlay is active, because the data is essentially accessed at
-    random points. Instead, call ``overlay_add`` again (preferably with a
+    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.
+    by passing a memory address as integer prefixed with an ``&`` character.
     Passing the wrong thing here will crash the player. This mode might be
     useful for use with libmpv. The ``offset`` parameter is simply added to the
     memory address (since mpv 0.8.0, ignored before).
@@ -578,22 +620,22 @@ Input Commands that are Possibly Subject to Change
         an overlay's memory at random times whenever it feels the need to do
         so, for example when redrawing the screen.
 
-``overlay_remove <id>``
-    Remove an overlay added with ``overlay_add`` and the same ID. Does nothing
+``overlay-remove <id>``
+    Remove an overlay added with ``overlay-add`` and the same ID. Does nothing
     if no overlay with this ID exists.
 
-``script_message "<arg1>" "<arg2>" ...``
+``script-message "<arg1>" "<arg2>" ...``
     Send a message to all clients, and pass it the following list of arguments.
     What this message means, how many arguments it takes, and what the arguments
     mean is fully up to the receiver and the sender. Every client receives the
     message, so be careful about name clashes (or use ``script_message_to``).
 
-``script_message_to "<target>" "<arg1>" "<arg2>" ...``
+``script-message-to "<target>" "<arg1>" "<arg2>" ...``
     Same as ``script_message``, but send it only to the client named
     ``<target>``. Each client (scripts etc.) has a unique name. For example,
     Lua scripts can get their name via ``mp.get_script_name()``.
 
-``script_binding "<name>"``
+``script-binding "<name>"``
     Invoke a script-provided key binding. This can be used to remap key
     bindings provided by external Lua scripts.
 
@@ -614,24 +656,24 @@ Input Commands that are Possibly Subject to Change
     tracked). The second letter whether the event originates from the mouse,
     either ``m`` (mouse button) or ``-`` (something else).
 
-``ab_loop``
+``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>"``
+``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-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.
 
-``screenshot_raw [subtitles|video|window]``
+``screenshot-raw [subtitles|video|window]``
     Return a screenshot in memory. This can be used only through the client
     API. The MPV_FORMAT_NODE_MAP returned by this command has the ``w``, ``h``,
     ``stride`` fields set to obvious contents. A ``format`` field is set to
@@ -640,14 +682,14 @@ Input Commands that are Possibly Subject to Change
     field is of type MPV_FORMAT_BYTE_ARRAY with the actual image data. The image
     is freed as soon as the result node is freed.
 
-Undocumented commands: ``tv_last_channel`` (TV/DVB only),
-``get_property`` (deprecated), ``ao_reload`` (experimental/internal).
+Undocumented commands: ``tv-last-channel`` (TV/DVB only),
+``ao-reload`` (experimental/internal).
 
 Hooks
 ~~~~~
 
 Hooks are synchronous events between player core and a script or similar. This
-applies to the Lua scripting interface and the client API and only. Normally,
+applies to client API (including the Lua scripting interface). Normally,
 events are supposed to be asynchronous, and the hook API provides an awkward
 and obscure way to handle events that require stricter coordination. There are
 no API stability guarantees made. Not following the protocol exactly can make
@@ -656,7 +698,7 @@ the player freeze randomly. Basically, nobody should use this API.
 There are two special commands involved. Also, the client must listen for
 client messages (``MPV_EVENT_CLIENT_MESSAGE`` in the C API).
 
-``hook_add <hook-name> <id> <priority>``
+``hook-add <hook-name> <id> <priority>``
     Subscribe to the hook identified by the first argument (basically, the
     name of event). The ``id`` argument is an arbitrary integer chosen by the
     user. ``priority`` is used to sort all hook handlers globally across all
@@ -681,9 +723,9 @@ client messages (``MPV_EVENT_CLIENT_MESSAGE`` in the C API).
     typically be stopped.
 
     When the client is done, it must continue the core's hook execution by
-    running the ``hook_ack`` command.
+    running the ``hook-ack`` command.
 
-``hook_ack <string>``
+``hook-ack <string>``
     Run the next hook in the global chain of hooks. The argument is the 3rd
     argument of the client message that starts hook execution for the
     current client.
@@ -756,7 +798,7 @@ Properties
 
 Properties are used to set mpv options during runtime, or to query arbitrary
 information. They can be manipulated with the ``set``/``add``/``cycle``
-commands, and retrieved with ``show_text``, or anything else that uses property
+commands, and retrieved with ``show-text``, or anything else that uses property
 expansion. (See `Property Expansion`_.)
 
 The property name is annotated with RW to indicate whether the property is
@@ -784,6 +826,16 @@ Property list
 ``speed`` (RW)
     See ``--speed``.
 
+``audio-speed-correction``, ``video-speed-correction``
+    Factor multiplied with ``speed`` at which the player attempts to play the
+    file. Usually it's exactly 1. (Display sync mode will make this useful.)
+
+    OSD formatting will display it in the form of ``+1.23456%``, with the number
+    being ``(raw - 1) * 100`` for the given raw property value.
+
+``display-sync-active``
+    Return whether ``--video-sync=display`` is actually active.
+
 ``filename``
     Currently played file, with path stripped. If this is an URL, try to undo
     percent encoding as well. (The result is not necessarily correct, but
@@ -836,11 +888,14 @@ Property list
 ``stream-end``
     Raw end position in bytes in source stream.
 
-``length``
-    Length of the current file in seconds. If the length is unknown, the
+``duration``
+    Duration of the current file in seconds. If the duration is unknown, the
     property is unavailable. Note that the file duration is not always exactly
     known, so this is an estimate.
 
+    This replaces the ``length`` property, which was deprecated after the
+    mpv 0.9 release. (The semantics are the same.)
+
 ``avsync``
     Last A/V synchronization difference. Unavailable if audio or video is
     disabled.
@@ -876,10 +931,12 @@ Property list
     always exactly known, so this is an estimate.
 
 ``playtime-remaining``
-    ``time-remaining`` scaled by the the current ``speed``.
+    ``time-remaining`` scaled by the current ``speed``.
 
-``playback-time``
-    Return the playback time, which is the time difference between start PTS and current PTS.
+``playback-time`` (RW)
+    The playback time, which is the time relative to playback start. (This can
+    be different from the ``time-pos`` property if the file does not start at
+    position ``0``, in which case ``time-pos`` is the source timestamp.)
 
 ``chapter`` (RW)
     Current chapter number. The number of the first chapter is 0.
@@ -922,16 +979,6 @@ Property list
     Current BD/DVD title number. Writing works only for ``dvdnav://`` and
     ``bd://`` (and aliases for these).
 
-``disc-menu-active``
-    Return ``yes`` if the BD/DVD menu is active, or ``no`` on normal video
-    playback. The property is unavailable when playing something that is not
-    a BD or DVD. Use the ``discnav menu`` command to actually enter or leave
-    menu mode.
-
-``disc-mouse-on-button``
-    Return ``yes`` when the mouse cursor is located on a button, or ``no``
-    when cursor is outside of any button for disc navigation.
-
 ``chapters``
     Number of chapters.
 
@@ -1092,7 +1139,7 @@ Property list
 
 ``demuxer-cache-time``
     Approximate time of video buffered in the demuxer, in seconds. Same as
-    ``demuxer-cache-duration`` but returns the last timestamp of bufferred
+    ``demuxer-cache-duration`` but returns the last timestamp of buffered
     data in demuxer.
 
 ``demuxer-cache-idle``
@@ -1126,7 +1173,7 @@ Property list
     See ``--hr-seek``.
 
 ``volume`` (RW)
-    Current volume (0-100).
+    Current volume (see ``--volume`` for details).
 
 ``mute`` (RW)
     Current mute status (``yes``/``no``).
@@ -1134,18 +1181,52 @@ Property list
 ``audio-delay`` (RW)
     See ``--audio-delay``.
 
-``audio-format``
-    Audio format as string.
-
 ``audio-codec``
     Audio codec selected for decoding.
 
-``audio-samplerate``
-    Audio samplerate.
+``audio-codec-name``
+    Audio codec.
 
-``audio-channels``
-    Number of audio channels. The OSD value of this property is actually the
-    channel layout, while the raw value returns the number of channels only.
+``audio-params``
+    Audio format as output by the audio decoder.
+    This has a number of sub-properties:
+
+    ``audio-params/format``
+        The sample format as string. This uses the same names as used in other
+        places of mpv.
+
+    ``audio-params/samplerate``
+        Samplerate.
+
+    ``audio-params/channels``
+        The channel layout as a string. This is similar to what the
+        ``--audio-channels`` accepts.
+
+    ``audio-params/hr-channels``
+        As ``channels``, but instead of the possibly cryptic actual layout
+        sent to the audio device, return a hopefully more human readable form.
+        (Usually only ``audio-out-params/hr-channels`` makes sense.)
+
+    ``audio-params/channel-count``
+        Number of audio channels. This is redundant to the ``channels`` field
+        described above.
+
+    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_MAP
+            "format"            MPV_FORMAT_STRING
+            "samplerate"        MPV_FORMAT_INT64
+            "channels"          MPV_FORMAT_STRING
+            "channel-count"     MPV_FORMAT_INT64
+            "hr-channels"       MPV_FORMAT_STRING
+
+``audio-out-params``
+    Same as ``audio-params``, but the format of the data written to the audio
+    API.
 
 ``aid`` (RW)
     Current audio track (similar to ``--aid``).
@@ -1156,7 +1237,7 @@ Property list
 ``balance`` (RW)
     Audio channel balance. (The implementation of this feature is rather odd.
     It doesn't change the volumes of each channel, but instead sets up a pan
-    matrix to mix the the left and right channels.)
+    matrix to mix the left and right channels.)
 
 ``fullscreen`` (RW)
     See ``--fullscreen``.
@@ -1167,17 +1248,18 @@ Property list
 ``field-dominance`` (RW)
     See ``--field-dominance``
 
-``colormatrix`` (RW)
-    See ``--colormatrix``.
+``colormatrix`` (R)
+    Redirects to ``video-params/colormatrix``. This parameter (as well as
+    similar ones) can be overridden with the ``format`` video filter.
 
-``colormatrix-input-range`` (RW)
-    See ``--colormatrix-input-range``.
+``colormatrix-input-range`` (R)
+    See ``colormatrix``.
 
-``colormatrix-output-range`` (RW)
-    See ``--colormatrix-output-range``.
+``colormatrix-output-range`` (R)
+    See ``colormatrix``.
 
-``colormatrix-primaries`` (RW)
-    See ``--colormatrix-primaries``.
+``colormatrix-primaries`` (R)
+    See ``colormatrix``.
 
 ``ontop`` (RW)
     See ``--ontop``.
@@ -1207,25 +1289,29 @@ Property list
     See ``--hue``.
 
 ``hwdec`` (RW)
-    Return the current hardware decoder that is used. This uses the same values
-    as the ``--hwdec`` option. If software decoding is active, this returns
-    ``no``. You can write this property. Then the ``--hwdec`` option is set to
-    the new value, and video decoding will be reinitialized (internally, the
-    player will perform a seek to refresh the video properly).
+    Reflects the ``--hwdec`` option.
 
-    Note that you don't know the success of the operation immediately after
-    writing this property. It happens with a delay as video is reinitialized.
+    Writing to it may change the currently used hardware decoder, if possible.
+    (Internally, the player may reinitialize the decoder, and will perform a
+    seek to refresh the video properly.) You can watch the other hwdec
+    properties to see whether this was successful.
 
-``detected-hwdec``
-    Return the current hardware decoder that was detected and opened. Returns
-    the same values as ``hwdec``.
+    Unlike in mpv 0.9.x and before, this does not return the currently active
+    hardware decoder.
 
-    This is known only once the VO has opened (and possibly later). With some
-    VOs (like ``opengl``), this is never known in advance, but only when the
-    decoder attempted to create the hw decoder successfully. Also, hw decoders
-    with ``-copy`` suffix are returned only while hw decoding is active (and
-    unset afterwards). All this reflects how detecting hw decoders are
-    detected and used internally in mpv.
+``hwdec-active``
+    Return ``yes`` or ``no``, depending on whether any type of hardware decoding
+    is actually in use.
+
+``hwdec-detected``
+    If software decoding is active, this returns the hardware decoder in use.
+    Otherwise, it returns either ``no``, or if applicable, the currently loaded
+    hardware decoding API. This is known only once the VO has opened (and
+    possibly later). With some VOs (like ``opengl``), this is never known in
+    advance, but only when the decoder attempted to create the hw decoder
+    successfully. Also, hw decoders with ``-copy`` suffix will return ``no``
+    while no video is being decoded. All this reflects how detecting hw decoders
+    are detected and used internally in mpv.
 
 ``panscan`` (RW)
     See ``--panscan``.
@@ -1289,26 +1375,29 @@ Property list
     ``video-params/rotate``
         Intended display rotation in degrees (clockwise).
 
+    ``video-params/stereo-in``
+        Source file stereo 3D mode. (See ``--video-stereo-mode`` option.)
+
     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 track)
-                "pixelformat"       MPV_FORMAT_STRING
-                "w"                 MPV_FORMAT_INT64
-                "h"                 MPV_FORMAT_INT64
-                "dw"                MPV_FORMAT_INT64
-                "dh"                MPV_FORMAT_INT64
-                "aspect"            MPV_FORMAT_DOUBLE
-                "par"               MPV_FORMAT_DOUBLE
-                "colormatrix"       MPV_FORMAT_STRING
-                "colorlevels"       MPV_FORMAT_STRING
-                "primaries"         MPV_FORMAT_STRING
-                "chroma-location"   MPV_FORMAT_STRING
-                "rotate"            MPV_FORMAT_INT64
+        MPV_FORMAT_NODE_MAP
+            "pixelformat"       MPV_FORMAT_STRING
+            "w"                 MPV_FORMAT_INT64
+            "h"                 MPV_FORMAT_INT64
+            "dw"                MPV_FORMAT_INT64
+            "dh"                MPV_FORMAT_INT64
+            "aspect"            MPV_FORMAT_DOUBLE
+            "par"               MPV_FORMAT_DOUBLE
+            "colormatrix"       MPV_FORMAT_STRING
+            "colorlevels"       MPV_FORMAT_STRING
+            "primaries"         MPV_FORMAT_STRING
+            "chroma-location"   MPV_FORMAT_STRING
+            "rotate"            MPV_FORMAT_INT64
+            "stereo-in"         MPV_FORMAT_STRING
 
 ``dwidth``, ``dheight``
     Video display size. This is the video size after filters and aspect scaling
@@ -1462,6 +1551,11 @@ Property list
         been unloaded yet; in this case, ``current`` refers to the new
         selection. (Since mpv 0.7.0.)
 
+    ``playlist/N/title``
+        Name of the Nth entry. Only available if the playlist file contains
+        such fields, and only if mpv's parser supports it for the given
+        playlist format.
+
     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:
@@ -1473,6 +1567,7 @@ Property list
                 "filename"  MPV_FORMAT_STRING
                 "current"   MPV_FORMAT_FLAG (might be missing; since mpv 0.7.0)
                 "playing"   MPV_FORMAT_FLAG (same)
+                "title"     MPV_FORMAT_STRING (optional)
 
 ``track-list``
     List of audio/video/sub tracks, current entry marked. Currently, the raw
@@ -1500,6 +1595,10 @@ Property list
     ``track-list/N/lang``
         Track language as identified by the file. Not always available.
 
+    ``track-list/N/audio-channels``
+        For audio tracks, the number of audio channels in the audio stream.
+        Not always accurate (depends on container hints). Not always available.
+
     ``track-list/N/albumart``
         ``yes`` if this is a video track that consists of a single picture,
         ``no`` or unavailable otherwise. This is used for video tracks that are
@@ -1509,6 +1608,10 @@ Property list
         ``yes`` if the track has the default flag set in the file, ``no``
         otherwise.
 
+    ``track-list/N/forced``
+        ``yes`` if the track has the forced flag set in the file, ``no``
+        otherwise.
+
     ``track-list/N/codec``
         The codec name used by this track, for example ``h264``. Unavailable
         in some rare cases.
@@ -1544,8 +1647,10 @@ Property list
                 "src-id"            MPV_FORMAT_INT64
                 "title"             MPV_FORMAT_STRING
                 "lang"              MPV_FORMAT_STRING
+                "audio-channels"    MPV_FORMAT_INT64
                 "albumart"          MPV_FORMAT_FLAG
                 "default"           MPV_FORMAT_FLAG
+                "forced"            MPV_FORMAT_FLAG
                 "external"          MPV_FORMAT_FLAG
                 "external-filename" MPV_FORMAT_STRING
                 "codec"             MPV_FORMAT_STRING
@@ -1624,7 +1729,7 @@ Property list
 
 ``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.
+    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``
@@ -1632,13 +1737,13 @@ Property list
     ``${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
+    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'``
-        - ``show_text "This is ${osd-ass-cc/0}{\b1}bold text"``
+        - ``show-text "This is ${osd-ass-cc/0}{\b1}bold text"``
 
     Any ASS override tags as understood by libass can be used.
 
@@ -1739,6 +1844,12 @@ Property list
     Return the working directory of the mpv process. Can be useful for JSON IPC
     users, because the command line player usually works with relative paths.
 
+``protocol-list``
+    List of protocol prefixes potentially recognized by the player. They are
+    returned without trailing ``://`` suffix (which is still always required).
+    In some cases, the protocol will not actually be supported (consider
+    ``https`` if ffmpeg is not compiled with TLS support).
+
 ``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.
@@ -1778,11 +1889,16 @@ Property list
         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>/set-locally``
+        Return ``yes`` if the option was set per-file. This is the case with
+        automatically loaded profiles, file-dir configs, and other cases. It
+        means the option value will be restored to the value before playback
+        start when playback ends.
+
     ``option-info/<name>/default-value``
         The default value of the option. May not always be available.
 
@@ -1805,11 +1921,14 @@ Property Expansion
 ------------------
 
 All string arguments to input commands as well as certain options (like
-``--term-playing-msg``) are subject to property expansion.
+``--term-playing-msg``) are subject to property expansion. Note that property
+expansion does not work in places where e.g. numeric parameters are expected.
+(For example, the ``add`` command does not do property expansion. The ``set``
+command is an exception and not a general rule.)
 
 .. admonition:: Example for input.conf
 
-    ``i show_text "Filename: ${filename}"``
+    ``i show-text "Filename: ${filename}"``
         shows the filename of the current file when pressing the ``i`` key
 
 Within ``input.conf``, property expansion can be inhibited by putting the

DOCS/man/ipc.rst

@@ -79,6 +79,24 @@ mpv will also send events to clients with JSON messages of the following form:
 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.
 
+Because events can occur at any time, it may be difficult at times to determine
+which response goes with which command. Commands may optionally include a
+``request_id`` which, if provided in the command request, will be copied
+verbatim into the response. mpv does not intrepret the ``request_id`` in any
+way; it is solely for the use of the requester.
+
+For example, this request:
+
+::
+
+    { "command": ["get_property", "time-pos"], "request_id": 100 }
+
+Would generate this response:
+
+::
+
+    { "error": "success", "data": 1.468135, "request_id": 100 }
+
 All commands, replies, and events are separated from each other with a line
 break character (``\n``).
 

DOCS/man/lua.rst

@@ -73,6 +73,10 @@ The ``mp`` module is preloaded, although it can be loaded manually with
     Run the given command. This is similar to the commands used in input.conf.
     See `List of Input Commands`_.
 
+    By default, this will show something on the OSD (depending on the command),
+    as if it was used in ``input.conf``. See `Input Command Prefixes`_ how
+    to influence OSD usage per command.
+
     Returns ``true`` on success, or ``nil, error`` on error.
 
 ``mp.commandv(arg1, arg2, ...)``
@@ -94,6 +98,9 @@ The ``mp`` module is preloaded, although it can be loaded manually with
     the ``expand-properties`` prefix, or the ``mp.get_property`` family of
     functions.
 
+    Unlike ``mp.command``, this will not use OSD by default either (except
+    for some OSD-specific commands).
+
 ``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
@@ -512,7 +519,7 @@ Example implementation::
         optionC = true,
     }
     read_options(options, "myscript")
-    print(option.optionA)
+    print(options.optionA)
 
 
 The config file will be stored in ``lua-settings/identifier.conf`` in mpv's user
@@ -623,6 +630,10 @@ strictly part of the guaranteed API.
             On Windows, ``killed`` is only returned when the process has been
             killed by mpv as a result of ``cancellable`` being set to ``true``.
 
+        ``killed_by_us``
+            Set to ``true`` if 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])``
@@ -683,6 +694,33 @@ List of events
     Happens after a file was unloaded. Typically, the player will load the
     next file right away, or quit if this was the last file.
 
+    The event has the ``reason`` field, which takes one of these values:
+
+    ``eof``
+        The file has ended. This can (but doesn't have to) include
+        incomplete files or broken network connections under
+        circumstances.
+
+    ``stop``
+        Playback was ended by a command.
+
+    ``quit``
+        Playback was ended by sending the quit command.
+
+    ``error``
+        An error happened. In this case, an ``error`` field is present with
+        the error string.
+
+    ``redirect``
+        Happens with playlists and similar. Details see
+        ``MPV_END_FILE_REASON_REDIRECT`` in the C API.
+
+    ``unknown``
+        Unknown. Normally doesn't happen, unless the Lua API is out of sync
+        with the C API. (Likewise, it could happen that your script gets
+        reason strings that did not exist yet at the time your script was
+        written.)
+
 ``file-loaded``
     Happens after a file was loaded and begins playback.
 

DOCS/man/mpv.rst

@@ -116,11 +116,7 @@ o (also P)
     Show progression bar, elapsed time and total duration on the OSD.
 
 O
-    Toggle OSD states: none / seek / seek + timer / seek + timer + total time.
-
-d
-    Toggle frame dropping states: none / skip display / skip decoding (see
-    ``--framedrop``).
+    Toggle OSD states between normal and playback time/duration.
 
 v
     Toggle subtitle visibility.
@@ -134,6 +130,9 @@ x and z
 l
     Set/clear A-B loop points. See ``ab_loop`` command for details.
 
+L
+    Toggle infinite looping.
+
 Ctrl + and Ctrl -
     Adjust audio delay by +/- 0.1 seconds.
 
@@ -156,6 +155,10 @@ S
     Take a screenshot, without subtitles. (Whether this works depends on VO
     driver support.)
 
+Ctrl s
+    Take a screenshot, as the window shows it (with subtitles, OSD, and scaled
+    video).
+
 I
     Show filename on the OSD.
 
@@ -168,7 +171,7 @@ Shift+PGUP and Shift+PGDWN
     Seek backward or forward by 10 minutes. (This used to be mapped to
     PGUP/PGDWN without Shift.)
 
-D
+d
     Activate/deactivate deinterlacer.
 
 A
@@ -273,7 +276,7 @@ spaces or characters like ``,`` or ``:``, you need to quote them:
     ``mpv '--vo=opengl:icc-profile="file with spaces.icc",xv'``
 
 Shells may actually strip some quotes from the string passed to the commandline,
-so the example quotes the string twice, ensuring that mpv recieves the ``"``
+so the example quotes the string twice, ensuring that mpv receives the ``"``
 quotes.
 
 The ``[...]`` form of quotes wraps everything between ``[`` and ``]``. It's
@@ -302,7 +305,7 @@ Suboptions passed to the client API are also subject to escaping. Using
 command line (but without shell processing of the string). Some options
 support passing values in a more structured way instead of flat strings, and
 can avoid the suboption parsing mess. For example, ``--vf`` supports
-``MPV_FORMAT_NODE``, which let's you pass suboptions as a nested data structure
+``MPV_FORMAT_NODE``, which lets you pass suboptions as a nested data structure
 of maps and arrays. (``--vo`` supports this in the same way, although this
 fact is undocumented.)
 
@@ -327,7 +330,17 @@ additionally wrapped in the fixed-length syntax, e.g. ``%n%string_of_length_n``
 Some mpv options interpret paths starting with ``~``. Currently, the prefix
 ``~~/`` expands to the mpv configuration directory (usually ``~/.config/mpv/``).
 ``~/`` expands to the user's home directory. (The trailing ``/`` is always
-required.)
+required.) There are the following paths as well:
+
+================ ===============================================================
+Name             Meaning
+================ ===============================================================
+``~~home/``      same as ``~~/``
+``~~global/``    the global config path, if available (not on win32)
+``~~osxbundle/`` the OSX bundle resource path (OSX only)
+``~~desktop/``   the path to the desktop (win32, OSX)
+================ ===============================================================
+
 
 Per-File Options
 ----------------
@@ -378,7 +391,8 @@ Location and Syntax
 You can put all of the options in configuration files which will be read every
 time mpv is run. The system-wide configuration file 'mpv.conf' is in your
 configuration directory (e.g. ``/etc/mpv`` or ``/usr/local/etc/mpv``), the
-user-specific one is ``~/.config/mpv/mpv.conf``.
+user-specific one is ``~/.config/mpv/mpv.conf``. For details and platform
+specifics see the `FILES`_ section.
 User-specific options override system-wide options and options given on the
 command line override either. The syntax of the configuration files is
 ``option=<value>``; everything after a *#* is considered a comment. Options
@@ -402,7 +416,7 @@ but option values still need to be quoted as a whole if it contains certain
 characters like spaces. A config entry can be quoted with ``"`` and ``'``,
 as well as with the fixed-length syntax (``%n%``) mentioned before. This is like
 passing the exact contents of the quoted string as command line option. C-style
-escapes are currently _not_ interpreted on this level, although some options to
+escapes are currently _not_ interpreted on this level, although some options do
 this manually. (This is a mess and should probably be changed at some point.)
 
 Putting Command Line Options into the Configuration File
@@ -512,13 +526,21 @@ listed.
   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.
+- Display sync state. If display sync is active (``display-sync-active``
+  property), this shows ``DS: +0.02598%``, where the number is the speed change
+  factor applied to audio to achieve sync to display, expressed in percent
+  deviation from 1.0 (``audio-speed-correction`` property). In sync modes which
+  don't resample, this will always be ``+0.00000%``.
+- Missed frames, e.g. ``Missed: 4``. (``vo-missed-frame-count`` property.) Shows
+  up in display sync mode only. This is incremented each time a frame took
+  longer to display than intended.
 - 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 happens only if
-  decoder-framedropping is enabled with the ``--framedrop`` options.
+  decoder frame dropping is enabled with the ``--framedrop`` options.
   (``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,
@@ -549,13 +571,9 @@ PROTOCOLS
     you must mount the ISO file as filesystem, and point ``--bluray-device``
     to the mounted directory directly.
 
-``bdnav://[title][/device]``
-    Play a Blu-Ray disc, with navigation features enabled. This feature is
-    permanently experimental.
-
 ``dvd://[title|[starttitle]-endtitle][/device]`` ``--dvd-device=PATH``
-    Play a DVD. If you want dvdnav menus, use ``dvd://menu``. If no title
-    is given, the longest title is auto-selected.
+    Play a DVD. DVD menus are not supported. If no title is given, the longest
+    title is auto-selected.
 
     ``dvdnav://`` is an old alias for ``dvd://`` and does exactly the same
     thing.
@@ -600,6 +618,11 @@ PROTOCOLS
     ``PATH`` itself should start with a third ``/`` to make the path an
     absolute path.
 
+``fd://123``
+    Read data from the given UNIX FD (for example 123). This is similar to
+    piping data to stdin via ``-``, but can use an arbitrary file descriptor.
+    Will not work correctly on MS Windows.
+
 ``edl://[edl specification as in edl-mpv.rst]``
     Stitch together parts of multiple files and play them.
 
@@ -634,6 +657,7 @@ the ``pseudo-gui`` profile being predefined with the following contents:
     terminal=no
     force-window=yes
     idle=once
+    screenshot-directory=~~desktop/
 
 This follows the mpv config file format. To customize pseudo-GUI mode, you can
 put your own ``pseudo-gui`` profile into your ``mpv.conf``. This profile will
@@ -788,6 +812,8 @@ If errors happen, the following exit codes can be returned:
         immediately after initialization.
     :3: There were some files that could be played, and some files which
         couldn't (using the definition of success from above).
+    :4: Quit due to a signal, Ctrl+c in a VO window (by default), or from the
+        default quit key bindings in encoding mode.
 
 Note that quitting the player manually will always lead to exit code 0,
 overriding the exit code that would be returned normally. Also, the ``quit``
@@ -796,6 +822,8 @@ input command can take an exit code: in this case, that exit code is returned.
 FILES
 =====
 
+For Windows-specifics, see `FILES ON WINDOWS`_ section.
+
 ``/usr/local/etc/mpv/mpv.conf``
     mpv system-wide settings (depends on ``--prefix`` passed to configure - mpv
     in default configuration will use ``/usr/local/etc/mpv/`` as config
@@ -857,6 +885,14 @@ Other config files (such as ``input.conf``) are in the same directory. See the
 The environment variable ``$MPV_HOME`` completely overrides these, like on
 UNIX.
 
+If a directory named ``portable_config`` next to the mpv.exe exists, all
+config will be loaded from this directory only. Watch later config files are
+written to this directory as well. (This exists on Windows only and is redundant
+with ``$MPV_HOME``. However, since Windows is very scripting unfriendly, a
+wrapper script just setting ``$MPV_HOME``, like you could do it on other
+systems, won't work. ``portable_config`` is provided for convenience to get
+around this restriction.)
+
 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

DOCS/man/options.rst

@@ -129,6 +129,17 @@ Playback Control
     Specify which chapter to start playing at. Optionally specify which
     chapter to end playing at. Also see ``--start``.
 
+``--playlist-pos=<no|index>``
+    Set which file on the internal playlist to start playback with. The index
+    is an integer, with 0 meaning the first file. The value ``no`` means that
+    the selection of the entry to play is left to the playback resume mechanism
+    (default). If an entry with the given index doesn't exist, the behavior is
+    unspecified and might change in future mpv versions. The same applies if
+    the playlist contains further playlists (don't expect any reasonable
+    behavior). Passing a playlist file to mpv should work with this option,
+    though. E.g. ``mpv playlist.m3u --playlist-pos=123`` will work as expected,
+    as long as ``playlist.m3u`` does not link to further playlists.
+
 ``--playlist=<filename>``
     Play files according to a playlist file (Supports some common formats. If
     no format is detected, it will be treated as list of files, separated by
@@ -348,7 +359,7 @@ Program Behavior
 ``--idle=<no|yes|once>``
     Makes mpv wait idly instead of quitting when there is no file to play.
     Mostly useful in slave mode, where mpv can be controlled through input
-    commands (see also ``--slave-broken``).
+    commands.
 
     ``once`` will only idle at start and let the player close once the
     first playlist has finished playing back.
@@ -457,14 +468,14 @@ Program Behavior
     (Default: ``best``)
 
 ``--ytdl-raw-options=<key>=<value>[,<key>=<value>[,...]]``
-    Pass arbitraty options to youtube-dl. Parameter and argument should be
+    Pass arbitrary options to youtube-dl. Parameter and argument should be
     passed as a key-value pair. Options without argument must include ``=``.
-    
+
     There is no sanity checking so it's possible to break things (i.e.
     passing invalid parameters to youtube-dl).
 
     .. admonition:: Example
- 
+
         ``--ytdl-raw-options=username=user,password=pass``
         ``--ytdl-raw-options=force-ipv6=``
 
@@ -498,6 +509,10 @@ Video
     Do not play video. With some demuxers this may not work. In those cases
     you can try ``--vo=null`` instead.
 
+    mpv will try to download the audio only if media is streamed with
+    youtube-dl, because it saves bandwidth. This is done by setting the ytdl_format
+    to "bestaudio/best" in the ytdl_hook.lua script.
+
 ``--untimed``
     Do not sleep when outputting video frames. Useful for benchmarks when used
     with ``--no-audio.``
@@ -524,7 +539,7 @@ Video
         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
+        but can produce unwatchable choppy output, or even freeze the display
         completely. Not recommended.
         The ``--vd-lavc-framedrop`` option controls what frames to drop.
     <decoder+vo>
@@ -555,6 +570,7 @@ Video
     :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`` (OS X only)
+    :videotoolbox: requires ``--vo=opengl`` (newer OS X only)
     :dxva2-copy: copies video back to system RAM (Windows only)
     :rpi:      requires ``--vo=rpi`` (Raspberry Pi only - default if available)
 
@@ -575,6 +591,22 @@ Video
         codecs. See ``--hwdec-codecs`` to enable hardware decoding for more
         codecs.
 
+``--hwdec-preload=<api>``
+    This is useful for the ``opengl`` and ``opengl-cb`` VOs for creating the
+    hardware decoding OpenGL interop context, but without actually enabling
+    hardware decoding itself (like ``--hwdec`` does).
+
+    If set to ``no`` (default), the ``--hwdec`` option is used.
+
+    For ``opengl``, if set, do not create the interop context on demand, but
+    when the VO is created.
+
+    For ``opengl-cb``, if set, load the interop context as soon as the OpenGL
+    context is created. Since ``opengl-cb`` has no on-demand loading, this
+    allows enabling hardware decoding at runtime at all, without having to
+    to temporarily set the ``hwdec`` option just during OpenGL context
+    initialization with ``mpv_opengl_cb_init_gl()``.
+
 ``--panscan=<0.0-1.0>``
     Enables pan-and-scan functionality (cropping the sides of e.g. a 16:9
     video to make it fit a 4:3 display without black bands). The range
@@ -601,6 +633,23 @@ Video
     Ignore aspect ratio information from video file and assume the video has
     square pixels. See also ``--video-aspect``.
 
+``--video-aspect-method=<hybrid|bitstream|container>``
+    This sets the default video aspect determination method (if the aspect is
+    _not_ overridden by the user with ``--video-aspect`` or others).
+
+    :hybrid:    Prefer the container aspect ratio. If the bitstream aspect
+                switches mid-stream, switch to preferring the bitstream aspect.
+                This is the default behavior in mpv and mplayer2.
+    :container: Strictly prefer the container aspect ratio. This is apparently
+                the default behavior with VLC, at least with Matroska.
+    :bitstream: Strictly prefer the bitstream aspect ratio, unless the bitstream
+                aspect ratio is not set. This is apparently the default behavior
+                with XBMC/kodi, at least with Matroska.
+
+    Normally you should not set this. Try the ``container`` and ``bitstream``
+    choices if you encounter video that has the wrong aspect ratio in mpv,
+    but seems to be correct in other players.
+
 ``--video-unscaled``
     Disable scaling of the video. If the window is larger than the video,
     black bars are added. Otherwise, the video is cropped. The video still
@@ -635,17 +684,17 @@ Video
     which means the value ``0`` would rotate the video according to the
     rotation metadata.)
 
-``--video-stereo-mode=<mode>``
+``--video-stereo-mode=<no|mode>``
     Set the stereo 3D output mode (default: ``mono``). This is done by inserting
     the ``stereo3d`` conversion filter.
 
+    The pseudo-mode ``no`` disables automatic conversion completely.
+
     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
@@ -729,7 +778,7 @@ Video
     You can get the list of allowed codecs with ``mpv --vd=help``. Remove the
     prefix, e.g. instead of ``lavc:h264`` use ``h264``.
 
-    By default this is set to ``h264,vc1,wmv3``. Note that the hardware
+    By default this is set to ``h264,vc1,wmv3,hevc``. Note that the hardware
     acceleration special codecs like ``h264_vdpau`` are not relevant anymore,
     and in fact have been removed from Libav in this form.
 
@@ -804,10 +853,11 @@ Video
     Set framedropping mode used with ``--framedrop`` (see skiploopfilter for
     available skip values).
 
-``--vd-lavc-threads=<0-16>``
+``--vd-lavc-threads=<N>``
     Number of threads to use for decoding. Whether threading is actually
-    supported depends on codec. 0 means autodetect number of cores on the
-    machine and use that, up to the maximum of 16 (default: 0).
+    supported depends on codec (default: 0). 0 means autodetect number of cores
+    on the machine and use that, up to the maximum of 16. You can set more than
+    16 threads manually.
 
 
 
@@ -834,7 +884,7 @@ Audio
     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).
+    selection of ``--audio-device`` (but not the device selection).
 
     Currently not implemented for most AOs.
 
@@ -851,6 +901,25 @@ Audio
     ``--af-clr`` exist to modify a previously specified list, but you
     should not need these for typical use.
 
+``--audio-spdif=<codecs>``
+    List of codecs for which compressed audio passthrough should be used. This
+    works for both classic S/PDIF and HDMI.
+
+    Possible codecs are ``ac3``, ``dts``, ``dts-hd``. Multiple codecs can be
+    specified by separating them with ``,``. ``dts`` refers to low bitrate DTS
+    core, while ``dts-hd`` refers to DTS MA (receiver and OS support varies).
+    You should only use either ``dts`` or ``dts-hd`` (if both are specified,
+    and ``dts`` comes first, only ``dts`` will be used).
+
+    In general, all codecs in the ``spdif`` family listed with ``--ad=help``
+    are supported in theory.
+
+    .. admonition:: Warning
+
+        There is not much reason to use this. HDMI supports uncompressed
+        multichannel PCM, and mpv supports lossless DTS-HD decoding via
+        FFmpeg's libdcadec wrapper.
+
 ``--ad=<[+|-]family1:(*|decoder1),[+|-]family2:(*|decoder2),...[-]>``
     Specify a priority list of audio decoders to be used, according to their
     family and decoder name. Entries like ``family:*`` prioritize all decoders
@@ -878,9 +947,21 @@ Audio
         ``--ad=help``
             List all available decoders.
 
-``--volume=<-1-100>``
-    Set the startup volume. A value of -1 (the default) will not change the
-    volume. See also ``--softvol``.
+    .. admonition:: Warning
+
+        Enabling compressed audio passthrough (AC3 and DTS via SPDIF/HDMI) with
+        this option is deprecated. Use ``--audio-spdif`` instead.
+
+``--volume=<value>``
+    Set the startup volume. 0 means silence, 100 means no volume reduction or
+    amplification. A value of -1 (the default) will not change the volume. See
+    also ``--softvol``.
+
+    .. note::
+
+        This was changed after the mpv 0.9 release. Before that, 100 actually
+        meant maximum volume. At the same time, the volume scale was made cubic,
+        so the old values won't match up with the new ones anyway.
 
 ``--audio-delay=<sec>``
     Audio delay in seconds (positive or negative float value). Positive values
@@ -942,17 +1023,12 @@ Audio
     welcome. A full list of AVOptions can be found in the FFmpeg manual.
 
 ``--ad-spdif-dtshd=<yes|no>``, ``--dtshd``, ``--no-dtshd``
-    When using DTS pass-through, output any DTS-HD track as-is.
-    With ``ad-spdif-dtshd=no`` (the default), only the DTS Core parts will be
-    output.
+    If DTS is passed through, use DTS-HD.
 
-    DTS-HD tracks can be sent over HDMI but not over the original
-    coax/TOSLINK S/PDIF system.
+    .. admonition:: Warning
 
-    Some receivers don't accept DTS core-only when ``--ad-spdif-dtshd=yes`` is
-    used, even though they accept DTS-HD.
-
-    ``--dtshd`` and ``--no-dtshd`` are deprecated aliases.
+        This and enabling passthrough via ``--ad`` are deprecated in favor of
+        using ``--audio-spdif=dts-hd``.
 
 ``--audio-channels=<number|layout>``
     Request a channel layout for audio output (default: auto). This  will ask
@@ -981,6 +1057,14 @@ Audio
     If the channel layout of the media file (i.e. the decoder) and the AO's
     channel layout don't match, mpv will attempt to insert a conversion filter.
 
+    .. admonition:: Warning
+
+        Using ``auto`` can cause issues when using audio over HDMI. The OS will
+        typically report all channel layouts that _can_ go over HDMI, even if
+        the receiver does not support them. If a receiver gets an unsupported
+        channel layout, random things can happen, such as dropping the
+        additional channels, or adding noise.
+
 ``--audio-display=<no|attachment>``
     Setting this option to ``attachment`` (default) will display image
     attachments (e.g. album cover art) when playing audio files. It will
@@ -1049,19 +1133,9 @@ Audio
     their start timestamps differ, and then video timing is gradually adjusted
     if necessary to reach correct synchronization later.
 
-``--softvol-max=<10.0-10000.0>``
-    Set the maximum amplification level in percent (default: 200). A value of
-    200 will allow you to adjust the volume up to a maximum of double the
-    current level. With values below 100 the initial volume (which is 100%)
-    will be above the maximum, which e.g. the OSD cannot display correctly.
-
-    .. admonition:: Note
-
-        The maximum value of ``--volume`` as well as the ``volume`` property
-        is always 100. Likewise, the volume OSD bar always goes from 0 to 100.
-        This means that with ``--softvol-max=200``, ``--volume=100`` sets
-        maximum amplification, i.e. amplify by 200%. The default volume (no
-        change in volume) will be ``50`` in this case.
+``--softvol-max=<100.0-1000.0>``
+    Set the maximum amplification level in percent (default: 130). A value of
+    130 will allow you to adjust the volume up to about double the normal level.
 
 ``--audio-file-auto=<no|exact|fuzzy|all>``, ``--no-audio-file-auto``
     Load additional audio files matching the video filename. The parameter
@@ -1136,7 +1210,7 @@ Subtitles
     the top of the screen) alongside the normal subtitle, and provides a way
     to render two subtitles at once.
 
-    there are some caveats associated with this feature. For example, bitmap
+    There are some caveats associated with this feature. For example, bitmap
     subtitles will always be rendered in their usual position, so selecting a
     bitmap subtitle as secondary subtitle will result in overlapping subtitles.
     Secondary subtitles are never shown on the terminal if video is disabled.
@@ -1352,6 +1426,18 @@ Subtitles
 
     Disabled by default.
 
+``--stretch-image-subs-to-screen=<yes|no>``
+    Stretch DVD and other image subtitles to the screen, ignoring the video
+    margins. This has a similar effect as ``--sub-use-margins`` for text
+    subtitles, except that the text itself will be stretched, not only just
+    repositioned. (At least in general it is unavoidable, as an image bitmap
+    can in theory consist of a single bitmap covering the whole screen, and
+    the player won't know where exactly the text parts are located.)
+
+    This option does not display subtitles correctly. Use with care.
+
+    Disabled by default.
+
 ``--sub-ass``, ``--no-sub-ass``
     Render ASS subtitles natively (enabled by default).
 
@@ -1381,10 +1467,11 @@ Subtitles
 
 ``--sub-codepage=<codepage>``
     If your system supports ``iconv(3)``, you can use this option to specify
-    the subtitle codepage. By default, ENCA will be used to guess the charset.
-    If mpv is not compiled with ENCA, ``UTF-8:UTF-8-BROKEN`` is the default,
-    which means it will try to use UTF-8, otherwise the ``UTF-8-BROKEN``
-    pseudo codepage (see below).
+    the subtitle codepage. By default, uchardet will be used to guess the
+    charset. If mpv is not compiled with uchardet, enca will be used.
+    If mpv is compiled with neither uchardet nor enca, ``UTF-8:UTF-8-BROKEN``
+    is the default, which means it will try to use UTF-8, otherwise the
+    ``UTF-8-BROKEN`` pseudo codepage (see below).
 
     The default value for this option is ``auto``, whose actual effect depends
     on whether ENCA is compiled.
@@ -1434,6 +1521,12 @@ Subtitles
     mode. Use ``--sub-codepage=guess:help`` to get a list of
     languages subject to the same caveat as with ENCA above.
 
+    If the player was compiled with uchardet support you can use it with:
+
+    ``--sub-codepage=uchardet``
+
+    This mode doesn't take language or fallback codepage.
+
 ``--sub-fix-timing``, ``--no-sub-fix-timing``
     By default, external text subtitles are preprocessed to remove minor gaps
     or overlaps between subtitles (if the difference is smaller than 200 ms,
@@ -1586,7 +1679,7 @@ Window
     file.mkv normally, then fail to open ``/dev/null``, then exit). (In
     mpv 0.8.0, ``always`` was introduced, which restores the old behavior.)
 
-``--force-window``
+``--force-window=<yes|no|immediate>``
     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
     always has the size 640x480, and is subject to ``--geometry``,
@@ -1598,7 +1691,9 @@ Window
         window placement still works if the video size is different from the
         ``--force-window`` default window size). This can be a problem if
         initialization doesn't work perfectly, such as when opening URLs with
-        bad network connection, or opening broken video files.
+        bad network connection, or opening broken video files. The ``immediate``
+        mode can be used to create the window always on program start, but this
+        may cause other issues.
 
 ``--ontop``
     Makes the player window stay on top of other windows.
@@ -1849,7 +1944,7 @@ Window
     ``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
+    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.
@@ -1943,7 +2038,7 @@ Disc Devices
     (Never) accept imperfect data reconstruction.
 
 ``--cdda-cdtext=<yes|no>``
-    Print CD text. This is disabled by default, because it ruins perfomance
+    Print CD text. This is disabled by default, because it ruins performance
     with CD-ROM drives for unknown reasons.
 
 ``--dvd-speed=<speed>``
@@ -2074,7 +2169,7 @@ Demuxer
     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
+    how much 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
@@ -2099,15 +2194,20 @@ Demuxer
 ``--demuxer-mkv-subtitle-preroll-secs=<value>``
     See ``--demuxer-mkv-subtitle-preroll``.
 
-``--demuxer-mkv-probe-video-duration``
+``--demuxer-mkv-probe-video-duration=<yes|no|full>``
     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.
 
+    The ``yes`` mode merely uses the index and reads a small number of blocks
+    from the end of the file. The ``full`` mode actually traverses the entire
+    file and can make a reliable estimate even without an index present (such
+    as partial files).
+
 ``--demuxer-mkv-fix-timestamps=<yes|no>``
-    Fix rounded Matroska timestamps (enabled by default). Matroska usually
+    Fix rounded Matroska timestamps (disabled by default). Matroska usually
     stores timestamps rounded to milliseconds. This means timestamps jitter
     by some amount around the intended timestamp. mpv can correct the timestamps
     based on the framerate value stored in the file: the timestamp is rounded
@@ -2140,7 +2240,7 @@ Demuxer
         Play a raw YUV sample::
 
             mpv sample-720x576.yuv --demuxer=rawvideo \
-            --demuxer-rawvideo=w=720:h=576
+            --demuxer-rawvideo-w=720 --demuxer-rawvideo-h=576
 
 ``--demuxer-rawvideo-format=<value>``
     Color space (fourcc) in hex or string for ``--demuxer=rawvideo``
@@ -2158,6 +2258,20 @@ Demuxer
 ``--demuxer-rawvideo-size=<value>``
     Frame size in bytes when using ``--demuxer=rawvideo``.
 
+``--demuxer-max-packets=<packets>``, ``--demuxer-max-bytes=<bytes>``
+    This controls how much the demuxer is allowed to buffer ahead. The demuxer
+    will normally try to read ahead as much as necessary, or as much is
+    requested with ``--demuxer-readahead-secs``. The ``--demuxer-max-...``
+    options can be used to restrict the maximum readahead. This limits excessive
+    readahead in case of broken files or desynced playback. The demuxer will
+    stop reading additional packets as soon as one of the limits is reached.
+    (The limits still can be slightly overstepped due to technical reasons.)
+
+    Set these limits highher if you get a packet queue overflow warning, and
+    you think normal playback would be possible with a larger packet queue.
+
+    See ``--list-options`` for defaults and value range.
+
 ``--demuxer-thread=<yes|no>``
     Run the demuxer in a separate thread, and let it prefetch a certain amount
     of packets (default: yes). Having this enabled may lead to smoother
@@ -2176,21 +2290,11 @@ Demuxer
     (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=<bytes>``
-    See ``--demuxer-readahead-packets``.
-
+``--force-seekable=<yes|no>``
+    If the player thinks that the media is not seekable (e.g. playing from a
+    pipe, or it's a http stream with a server that doesn't support range
+    requests), seeking will be disabled. This option can forcibly enable it.
+    For seeks within the cache, there's a good chance of success.
 
 Input
 -----
@@ -2247,8 +2351,6 @@ Input
     This can also specify a direct file descriptor with ``fd://N`` (UNIX only).
     In this case, JSON replies will be written if the FD is writable.
 
-    See also ``--slave-broken``.
-
     .. note::
 
         When the given file is a FIFO mpv opens both ends, so you can do several
@@ -2562,14 +2664,19 @@ Screenshot
 
     Note that not all formats are supported.
 
-    Default: ``yes``.
+    Default: ``no``.
+
+``--screenshot-high-bit-depth=<yes|no>``
+    If possible, write screenshots with a bit depth similar to the source
+    video (default: yes). This is interesting in particular for PNG, as this
+    sometimes triggers writing 16 bit PNGs with huge file sizes.
 
 ``--screenshot-template=<template>``
     Specify the filename template used to save screenshots. The template
     specifies the filename without file extension, and can contain format
     specifiers, which will be substituted when taking a screenshot.
-    By default the template is ``shot%n``, which results in filenames like
-    ``shot0012.png`` for example.
+    By default the template is ``mpv-shot%n``, which results in filenames like
+    ``mpv-shot0012.png`` for example.
 
     The template can start with a relative or absolute path, in order to
     specify a directory location where screenshots should be saved.
@@ -2612,7 +2719,7 @@ Screenshot
 
         .. note::
 
-            This is a simple way for getting unique per-frame timestamps. Frame
+            This is a simple way for getting unique per-frame timestamps. (Frame
             numbers would be more intuitive, but are not easily implementable
             because container formats usually use time stamps for identifying
             frames.)
@@ -2651,9 +2758,26 @@ Screenshot
     ``%%``
         Replaced with the ``%`` character itself.
 
+``--screenshot-directory=<path>``
+    Store screenshots in this directory. This path is joined with the filename
+    generated by ``--screenshot-template``. If the template filename is already
+    absolute, the directory is ignored.
+
+    If the directory does not exist, it is created on the first screenshot. If
+    it is not a directory, an error is generated when trying to write a
+    screenshot.
+
+    This option is not set by default, and thus will write screenshots to the
+    directory from which mpv was started. In pseudo-gui mode
+    (see `PSEUDO GUI MODE`_), this is set to the desktop.
+
 ``--screenshot-jpeg-quality=<0-100>``
     Set the JPEG quality level. Higher means better quality. The default is 90.
 
+``--screenshot-jpeg-source-chroma=<yes|no>``
+    Write JPEG files with the same chroma subsampling as the video
+    (default: yes). If disabled, the libjpeg default is used.
+
 ``--screenshot-png-compression=<0-9>``
     Set the PNG compression level. Higher means better compression. This will
     affect the file size of the written screenshot file and the time it takes
@@ -2984,13 +3108,13 @@ Cache
     seeking, such as MP4.
 
     Note that half the cache size will be used to allow fast seeking back. This
-    is also the reason why a full cache is usually reported as 50% full. The
-    cache fill display does not include the part of the cache reserved for
-    seeking back. Likewise, when starting a file the cache will be at 100%,
-    because no space is reserved for seeking back yet.
+    is also the reason why a full cache is usually not reported as 100% full.
+    The cache fill display does not include the part of the cache reserved for
+    seeking back. The actual maximum percentage will usually be the ratio
+    between readahead and backbuffer sizes.
 
 ``--cache-default=<kBytes|no>``
-    Set the size of the cache in kilobytes (default: 150000 KB). Using ``no``
+    Set the size of the cache in kilobytes (default: 75000 KB). Using ``no``
     will not automatically enable the cache e.g. when playing from a network
     stream. Note that using ``--cache`` will always override this option.
 
@@ -3010,6 +3134,12 @@ Cache
     on the situation, either of these might be slower than the other method.
     This option allows control over this.
 
+``--cache-backbuffer=<kBytes>``
+    Size of the cache back buffer (default: 75000 KB). This will add to the total
+    cache size, and reserved the amount for seeking back. The reserved amount
+    will not be used for readahead, and instead preserves already read data to
+    enable fast seeking back.
+
 ``--cache-file=<TMP|path>``
     Create a cache file on the filesystem.
 
@@ -3124,7 +3254,7 @@ Network
     network transport when playing ``rtsp://...`` URLs. The value ``lavf``
     leaves the decision to libavformat.
 
-``--hls-bitrate=<no|min|max>``
+``--hls-bitrate=<no|min|max|<rate>>``
     If HLS streams are played, this option controls what streams are selected
     by default. The option allows the following parameters:
 
@@ -3133,6 +3263,9 @@ Network
     :min:       Pick the streams with the lowest bitrate.
     :max:       Same, but highest bitrate. (Default.)
 
+    Additionally, if the option is a number, the stream with the highest rate
+    equal or below the option value is selected.
+
     The bitrate as used is sent by the server, and there's no guarantee it's
     actually meaningful.
 
@@ -3258,6 +3391,72 @@ Miscellaneous
     out. This delay in reaction time to sudden A/V offsets should be the only
     side-effect of turning this option on, for all sound drivers.
 
+``--video-sync=<audio|...>``
+    How the player synchronizes audio and video.
+
+    The modes starting with ``display-`` try to output video frames completely
+    synchronously to the display, using the detected display vertical refresh
+    rate as a hint how fast frames will be displayed on average. These modes
+    change video speed slightly to match the display. See ``--video-sync-...``
+    options for fine tuning. The robustness of this mode is further reduced by
+    making a some idealized assumptions, which may not always apply in reality.
+    Behavior can depend on the VO and the system's video and audio drivers.
+    Media files must use constant framerate. Section-wise VFR might work as well
+    with some container formats (but not e.g. mkv). If the sync code detects
+    severe A/V desync, or the framerate cannot be detected, the player
+    automatically reverts to ``audio`` mode for some time or permanently.
+
+    The modes with ``desync`` in their names do not attempt to keep audio/video
+    in sync. They will slowly (or quickly) desync, until e.g. the next seek
+    happens. These modes are meant for testing, not serious use.
+
+    :audio:             Time video frames to audio. This is the most robust
+                        mode, because the player doesn't have to assume anything
+                        about how the display behaves. The disadvantage is that
+                        it can lead to occasional frame drops or repeats. If
+                        audio is disabled, this uses the system clock. This is
+                        the default mode.
+    :display-resample:  Resample audio to match the video. This mode will also
+                        try to adjust audio speed to compensate for other drift.
+                        (This means it will play the audio at a different speed
+                        every once in a while to reduce the A/V difference.)
+    :display-resample-vdrop:  Resample audio to match the video. Drop video
+                        frames to compensate for drift.
+    :display-resample-desync: Like the previous mode, but no A/V compensation.
+    :display-vdrop:     Drop or repeat video frames to compensate desyncing
+                        video. (Although it should have the same effects as
+                        ``audio``, the implementation is very different.)
+    :display-desync:    Sync video to display, and let audio play on its own.
+    :desync:            Sync video according to system clock, and let audio play
+                        on its own.
+
+``--video-sync-max-video-change=<value>``
+    Maximum speed difference in percent that is applied to video with
+    ``--video-sync=display-...`` (default: 1). Display sync mode will be
+    disabled if the monitor and video refresh way do not match within the
+    given range. It tries multiples as well: playing 30 fps video on a 60 Hz
+    screen will duplicate every second frame. Playing 24 fps video on a 60 Hz
+    screen will play video in a 2-3-2-3-... pattern.
+
+    The default settings are not loose enough to speed up 23.976 fps video to
+    25 fps. We consider the pitch change too extreme to allow this behavior
+    by default. Set this option to a value of ``5`` to enable it.
+
+    Note that in the ``--video-sync=display-resample`` mode, audio speed will
+    additionally be changed by a small amount if necessary for A/V sync. See
+    ``--video-sync-max-audio-change``.
+
+``--video-sync-max-audio-change=<value>``
+    Maximum *additional* speed difference in percent that is applied to audio
+    with ``--video-sync=display-...`` (default: 0.125). Normally, the player
+    play the audio at the speed of the video. But if the difference between
+    audio and video position is too high, e.g. due to drift or other timing
+    errors, it will attempt to speed up or slow down audio by this additional
+    factor. Too low values could lead to video frame dropping or repeating if
+    the A/V desync cannot be compensated, too high values could lead to chaotic
+    frame dropping due to the audio "overshooting" and skipping multiple video
+    frames before the sync logic can react.
+
 ``--mf-fps=<value>``
     Framerate used when decoding from multiple PNG or JPEG files with ``mf://``
     (default: 1).
@@ -3305,7 +3504,7 @@ Miscellaneous
     you should not need to change this option.
 
     :decoder: Use decoder reordering functionality. Unlike in classic MPlayer
-              and mplayer2, this includes a dTS fallback. (Default.)
+              and mplayer2, this includes a DTS fallback. (Default.)
     :sort:    Maintain a buffer of unused pts values and use the lowest value
               for the frame.
     :auto:    Try to pick a working mode from the ones above automatically.
@@ -3313,25 +3512,7 @@ Miscellaneous
     You can also try to use ``--no-correct-pts`` for files with completely
     broken timestamps.
 
-``--media-title=<string>``
+``--force-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.
-
-    .. attention::
-        Changes incompatible to slave mode applications have been made. In
-        particular, the status line output was changed, which is used by some
-        applications to determine the current playback position. This switch
-        has been renamed to prevent these applications from working with this
-        version of mpv, because it would lead to buggy and confusing behavior
-        only. Moreover, the slave mode protocol is so horribly bad that it
-        should not be used for new programs, nor should existing programs
-        attempt to adapt to the changed output and use the ``--slave-broken``
-        switch. Instead, a new, saner protocol should be developed (and will be,
-        if there is enough interest).
-
-        This affects most third-party GUI frontends.

DOCS/man/osc.rst

@@ -5,7 +5,7 @@ The On Screen Controller (short: OSC) is a minimal GUI integrated with mpv to
 offer basic mouse-controllability. It is intended to make interaction easier
 for new users and to enable precise and direct seeking.
 
-The OSC is enabled by default if mpv was compiled with lua support. It can be
+The OSC is enabled by default if mpv was compiled with Lua support. It can be
 disabled entirely using the ``--osc=no`` option.
 
 Using the OSC
@@ -169,34 +169,35 @@ Configurable Options
 
 ``showwindowed``
     | Default: yes
-    | Show OSC when windowed?
+    | Enable the OSC when windowed
 
 ``showfullscreen``
     | Default: yes
-    | Show OSC when fullscreen?
+    | Enable the OSC when fullscreen
 
 ``scalewindowed``
-    | Default: 1
-    | Scaling of the controller when windowed
+    | Default: 1.0
+    | Scale factor of the OSC when windowed
 
 ``scalefullscreen``
-    | Default: 1
-    | Scaling of the controller when fullscreen
+    | Default: 1.0
+    | Scale factor of the OSC when fullscreen
 
 ``scaleforcedwindow``
-    | Default: 2
-    | Scaling of the controller when rendered on a forced (dummy) window
+    | Default: 2.0
+    | Scale factor of the OSC when rendered on a forced (dummy) window
 
 ``vidscale``
     | Default: yes
-    | Scale the controller with the video?
+    | Scale the OSC with the video
+    | ``no`` tries to keep the OSC size constant as much as the window size allows
 
 ``valign``
     | Default: 0.8
     | Vertical alignment, -1 (top) to 1 (bottom)
 
 ``halign``
-    | Default: 0
+    | Default: 0.0
     | Horizontal alignment, -1 (left) to 1 (right)
 
 ``boxalpha``
@@ -234,6 +235,14 @@ Configurable Options
     | Default: slider
     | Sets the style of the seekbar, slider (diamond marker) or bar (fill)
 
+``timetotal``
+    | Default: no
+    | Show total time instead of time remaining
+
+``timems``
+    | Default: no
+    | Display timecodes with milliseconds
+
 Script Commands
 ~~~~~~~~~~~~~~~
 

DOCS/man/vf.rst

@@ -97,7 +97,7 @@ Available filters are:
 
     All parameters are optional.
 
-    ``<w>,<h>``
+    ``<w>:<h>``
         scaled width/height (default: original width/height)
 
         :0:      scaled d_width/d_height
@@ -152,7 +152,7 @@ Available filters are:
     number (1.33). Alternatively, you may specify the exact display width and
     height desired. Note that this filter does *not* do any scaling itself; it
     just affects what later scalers (software or hardware) will do when
-    auto-scaling to correct aspect.
+    auto-scaling to the correct aspect.
 
     ``<w>,<h>``
         New display width and height.
@@ -198,7 +198,7 @@ Available filters are:
     ``<aspect>``
         Force an aspect ratio.
 
-``format[=fmt[:outfmt]]``
+``format=fmt=<value>:colormatrix=<value>:...``
     Restricts the color space for the next filter without doing any conversion.
     Use together with the scale filter for a real conversion.
 
@@ -518,7 +518,7 @@ Available filters are:
         video. The main purpose of setting ``mp`` to a chroma plane is to reduce
         CPU load and make pullup usable in realtime on slow machines.
 
-``yadif=[mode[:enabled=yes|no]]``
+``yadif=[mode:interlaced-only]``
     Yet another deinterlacing filter
 
     ``<mode>``
@@ -527,10 +527,10 @@ Available filters are:
         :frame-nospatial: Like ``frame`` but skips spatial interlacing check.
         :field-nospatial: Like ``field`` but skips spatial interlacing check.
 
-    ``<enabled>``
-        :yes: Filter is active (default).
-        :no:  Filter is not active, but can be activated with the ``D`` key
-              (or any other key that toggles the ``deinterlace`` property).
+    ``<interlaced-only>``
+        :no:  Deinterlace all frames (default).
+        :yes: Only deinterlace frames marked as interlaced (default if this
+              filter is inserted via ``deinterlace`` property).
 
     This filter, is automatically inserted when using the ``D`` key (or any
     other key that toggles the ``deinterlace`` property or when using the
@@ -669,6 +669,8 @@ Available filters are:
     Loads an external library to filter the image. The library interface
     is the ``vf_dlopen`` interface specified using ``libmpcodecs/vf_dlopen.h``.
 
+    .. warning:: This filter is deprecated.
+
     ``dll=<library>``
         Specify the library to load. This may require a full file system path
         in some cases. This argument is required.
@@ -778,7 +780,7 @@ Available filters are:
 ``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
+    is completely different, and absolutely no convenience features are
     provided. There's no type checking either, and you can trigger crashes.
 
     .. admonition:: Example:
@@ -820,6 +822,10 @@ Available filters are:
             depends on the GPU hardware, the GPU drivers, driver bugs, and
             mpv bugs.
 
+    ``<interlaced-only>``
+        :no:  Deinterlace all frames.
+        :yes: Only deinterlace frames marked as interlaced (default).
+
 ``vdpaupp``
     VDPAU video post processing. Works with ``--vo=vdpau`` and ``--vo=opengl``
     only. This filter is automatically inserted if deinterlacing is requested
@@ -864,12 +870,21 @@ Available filters are:
     ``pullup``
         Try to apply inverse telecine, needs motion adaptive temporal
         deinterlacing.
+    ``interlaced-only=<yes|no>``
+        If ``yes`` (default), only deinterlace frames marked as interlaced.
     ``hqscaling=<0-9>``
         0
             Use default VDPAU scaling (default).
         1-9
             Apply high quality VDPAU scaling (needs capable hardware).
 
+``vdpaurb``
+    VDPAU video read back. Works with ``--vo=vdpau`` and ``--vo=opengl`` only.
+    This filter will read back frames decoded by VDPAU so that other filters,
+    which are not normally compatible with VDPAU, can be used like normal.
+    This filter must be specified before ``vdpaupp`` in the filter chain if
+    ``vdpaupp`` is used.
+
 ``buffer=<num>``
     Buffer ``<num>`` frames in the filter chain. This filter is probably pretty
     useless, except for debugging. (Note that this won't help smoothing out

DOCS/man/vo.rst

@@ -69,11 +69,11 @@ Available video output drivers are:
     ``no-colorkey``
         Disables color-keying.
 
-``x11`` (X11 only)
-    Shared memory video output driver without hardware acceleration that works
-    whenever X11 is present.
-
-    .. note:: This is a fallback only, and should not be normally used.
+    ``buffers=<number>``
+        Number of image buffers to use for the internal ringbuffer (default: 2).
+        Increasing this will use more memory, but might help with the X server
+        not responding quickly enough if video FPS is close to or higher than
+        the display refresh rate.
 
 ``vdpau`` (X11 only)
     Uses the VDPAU interface to display and optionally also decode video.
@@ -336,6 +336,9 @@ Available video output drivers are:
             exchange for adding some blur. This filter is good at temporal
             interpolation, and also known as "smoothmotion" (see ``tscale``).
 
+        ``custom``
+            A user-defined custom shader (see ``scale-shader``).
+
         There are some more filters, but most are not as useful. For a complete
         list, pass ``help`` as value, e.g.::
 
@@ -452,6 +455,11 @@ Available video output drivers are:
         Unfortunately, this can lead to flicker on LCD displays, since these
         have a high reaction time.
 
+    ``temporal-dither-period=<1-128>``
+        Determines how often the dithering pattern is updated when
+        ``temporal-dither`` is in use. 1 (the default) will update on every
+        video frame, 2 on every other frame, etc.
+
     ``debug``
         Check for OpenGL errors, i.e. call ``glGetError()``. Also request a
         debug OpenGL context (which does nothing with current graphics drivers
@@ -494,10 +502,15 @@ Available video output drivers are:
         for ``tscale`` are separable convolution filters (use ``tscale=help``
         to get a list). The default is ``oversample``.
 
-        Note that the maximum supported filter radius is currently 3, and that
-        using filters with larger radius may introduce issues when pausing or
-        framestepping, proportional to the radius used. It is recommended to
-        stick to a radius of 1 or 2.
+        Note that the maximum supported filter radius is currently 3, due to
+        limitations in the number of video textures that can be loaded
+        simultaneously.
+
+    ``tscale-clamp``
+        Clamp the ``tscale`` filter kernel's value range to [0-1]. This reduces
+        excessive ringing artifacts in the temporal domain (which typically
+        manifest themselves as short flashes or fringes of black, mostly
+        around moving edges) in exchange for potentially adding more blur.
 
     ``dscale-radius``, ``cscale-radius``, ``tscale-radius``, etc.
         Set filter parameters for ``dscale``, ``cscale`` and ``tscale``,
@@ -506,10 +519,8 @@ Available video output drivers are:
         See the corresponding options for ``scale``.
 
     ``linear-scaling``
-        Scale in linear light. This is automatically enabled if
-        ``target-prim``, ``target-trc``, ``icc-profile`` or
-        ``sigmoid-upscaling`` is set. It should only be used with a
-        ``fbo-format`` that has at least 16 bit precision.
+        Scale in linear light. It should only be used with a ``fbo-format``
+        that has at least 16 bit precision.
 
     ``fancy-downscaling``
         When using convolution based filters, extend the filter size
@@ -519,9 +530,73 @@ Available video output drivers are:
         feature doesn't work correctly with different scale factors in
         different directions.
 
+    ``source-shader=<file>``, ``scale-shader=<file>``, ``pre-shaders=<files>``, ``post-shaders=<files>``
+        Custom GLSL fragment shaders.
+
+        source-shader
+            This gets applied directly onto the source planes, before
+            any sort of upscaling or conversion whatsoever. For YCbCr content,
+            this means it gets applied on the luma and chroma planes
+            separately. In general, this shader shouldn't be making any
+            assumptions about the colorspace. It could be RGB, YCbCr, XYZ or
+            something else entirely. It's used purely for fixing numerical
+            quirks of the input, eg. debanding or deblocking.
+        pre-shaders (list)
+            These get applied after conversion to RGB and before linearization
+            and upscaling. Operates on non-linear RGB (same as input). This is
+            the best place to put things like sharpen filters.
+        scale-shader
+            This gets used instead of scale/cscale when those options are set
+            to ``custom``. The colorspace it operates on depends on the values
+            of ``linear-scaling`` and ``sigmoid-upscaling``, so no assumptions
+            should be made here.
+        post-shaders (list)
+            These get applied after upscaling and subtitle blending (when
+            ``blend-subtitles`` is enabled), but before color management.
+            Operates on linear RGB if ``linear-scaling`` is in effect,
+            otherwise non-linear RGB. This is the best place for colorspace
+            transformations (eg. saturation mapping).
+
+        These files must define a function with the following signature::
+
+            vec4 sample(sampler2D tex, vec2 pos, vec2 tex_size)
+
+        The meanings of the parameters are as follows:
+
+        sampler2D tex
+            The source texture for the shader.
+        vec2 pos
+            The position to be sampled, in coordinate space [0-1].
+        vec2 tex_size
+            The size of the texture, in pixels. This may differ from image_size,
+            eg. for subsampled content or for post-shaders.
+
+        In addition to these parameters, the following uniforms are also
+        globally available:
+
+        float random
+            A random number in the range [0-1], different per frame.
+        int frame
+            A simple count of frames rendered, increases by one per frame and
+            never resets (regardless of seeks).
+        vec2 image_size
+            The size in pixels of the input image.
+        float cmul (source-shader only)
+            The multiplier needed to pull colors up to the right bit depth. The
+            source-shader must multiply any sampled colors by this, in order
+            to normalize them to the full scale.
+
+        For example, a shader that inverts the colors could look like this::
+
+            vec4 sample(sampler2D tex, vec2 pos, vec2 tex_size)
+            {
+                vec4 color = texture(tex, pos);
+                return vec4(1.0 - color.rgb, color.a);
+            }
+
     ``sigmoid-upscaling``
         When upscaling, use a sigmoidal color transform to avoid emphasizing
-        ringing artifacts. This also enables ``linear-scaling``.
+        ringing artifacts. This also implies ``linear-scaling``.
 
     ``sigmoid-center``
         The center of the sigmoid curve used for ``sigmoid-upscaling``, must
@@ -531,10 +606,6 @@ Available video output drivers are:
         The slope of the sigmoid curve used for ``sigmoid-upscaling``, must
         be a float between 1.0 and 20.0. Defaults to 6.5 if not specified.
 
-    ``no-npot``
-        Force use of power-of-2 texture sizes. For debugging only.
-        Borders will be distorted due to filtering.
-
     ``glfinish``
         Call ``glFinish()`` before and after swapping buffers (default: disabled).
         Slower, but might help getting better results when doing framedropping.
@@ -555,7 +626,8 @@ Available video output drivers are:
         full screen).
         This may help getting more consistent frame intervals, especially with
         high-fps clips - which might also reduce dropped frames. Typically a
-        value of 1 should be enough since full screen may bypass the DWM.
+        value of ``windowed`` should be enough since full screen may bypass the
+        DWM.
 
         Windows only.
 
@@ -573,12 +645,15 @@ Available video output drivers are:
             Cocoa/OS X
         win
             Win32/WGL
-        x11, x11es
-            X11/GLX (the ``es`` variant forces GLES)
+        x11
+            X11/GLX
         wayland
             Wayland/EGL
-        x11egl, x11egles
-            X11/EGL (the ``es`` variant forces GLES)
+        x11egl
+            X11/EGL
+
+    ``es``
+        Force or prefer GLES2/3 over desktop OpenGL, if supported.
 
     ``fbo-format=<fmt>``
         Selects the internal format of textures used for FBOs. The format can
@@ -586,7 +661,7 @@ Available video output drivers are:
         always used, and typically only when using extended scalers.)
         ``fmt`` can be one of: rgb, rgba, rgb8, rgb10, rgb10_a2, rgb16, rgb16f,
         rgb32f, rgba12, rgba16, rgba16f, rgba32f.
-        Default: rgba.
+        Default: rgba16.
 
     ``gamma=<0.1..2.0>``
         Set a gamma value (default: 1.0). If gamma is adjusted in other ways
@@ -660,8 +735,7 @@ Available video output drivers are:
     ``icc-profile=<file>``
         Load an ICC profile and use it to transform linear RGB to screen output.
         Needs LittleCMS 2 support compiled in. This option overrides the
-        ``target-prim`` and ``target-trc`` options. It also enables
-        ``linear-scaling``.
+        ``target-prim``, ``target-trc`` and ``icc-profile-auto`` options.
 
     ``icc-profile-auto``
         Automatically select the ICC display profile currently specified by
@@ -669,11 +743,14 @@ Available video output drivers are:
 
         NOTE: Only implemented on OS X and X11
 
-    ``icc-cache=<file>``
-        Store and load the 3D LUT created from the ICC profile in this file.
+    ``icc-cache-dir=<dirname>``
+        Store and load the 3D LUTs created from the ICC profile in this directory.
         This can be used to speed up loading, since LittleCMS 2 can take a while
-        to create the 3D LUT. Note that this file contains an uncompressed LUT.
-        Its size depends on the ``3dlut-size``, and can be very big.
+        to create a 3D LUT. Note that these files contain uncompressed LUTs.
+        Their size depends on the ``3dlut-size``, and can be very big.
+
+        NOTE: This is not cleaned automatically, so old, unused cache files
+        may stick around indefinitely.
 
     ``icc-intent=<value>``
         Specifies the ICC intent used for the color transformation (when using
@@ -697,8 +774,8 @@ Available video output drivers are:
         Blend subtitles directly onto upscaled video frames, before
         interpolation and/or color management (default: no). Enabling this
         causes subtitles to be affected by ``icc-profile``, ``target-prim``,
-        ``target-trc``, ``interpolation``, ``gamma`` and ``linear-scaling``.
-        It also increases subtitle performance when using ``interpolation``.
+        ``target-trc``, ``interpolation``, ``gamma`` and ``post-shader``. It
+        also increases subtitle performance when using ``interpolation``.
 
         The downside of enabling this is that it restricts subtitles to the
         visible portion of the video, so you can't have subtitles exist in the
@@ -744,7 +821,7 @@ Available video output drivers are:
 
     This is equivalent to::
 
-        --vo=opengl:scale=spline36:cscale=spline36:dscale=mitchell:dither-depth=auto:fbo-format=rgba16:fancy-downscaling:sigmoid-upscaling
+        --vo=opengl:scale=spline36:cscale=spline36:dscale=mitchell:dither-depth=auto:fancy-downscaling:sigmoid-upscaling:pbo
 
     Note that some cheaper LCDs do dithering that gravely interferes with
     ``opengl``'s dithering. Disabling dithering with ``dither-depth=no`` helps.
@@ -816,6 +893,12 @@ Available video output drivers are:
 ``null``
     Produces no video output. Useful for benchmarking.
 
+    Usually, it's better to disable video with ``--no-video`` instead.
+
+    ``fps=<value>``
+        Simulate display FPS. This artificially limits how many frames the
+        VO accepts per second.
+
 ``caca``
     Color ASCII art video output driver that works on a text console.
 
@@ -895,13 +978,15 @@ Available video output drivers are:
     ``frame-queue-size=<1..100>``
         The maximum count of frames which the frame queue can hold (default: 1)
 
-    ``frame-drop-mode=<pop|clear>``
+    ``frame-drop-mode=<pop|clear|block>``
         Select the behavior when the frame queue is full.
 
         pop
-            Drop the oldest frame in the frame queue. (default)
+            Drop the oldest frame in the frame queue.
         clear
             Drop all frames in the frame queue.
+        block
+            Wait for a short time, behave like ``clear`` on timeout. (default)
 
     This also supports many of the suboptions the ``opengl`` VO has. Run
     ``mpv --vo=opengl-cb:help`` for a list.
@@ -921,6 +1006,11 @@ Available video output drivers are:
         selected layer, to handle the window background and OSD. Actual video
         rendering will happen on the layer above the selected layer.
 
+    ``background=<yes|no>``
+        Whether to render a black background behind the video (default: no).
+        Normally it's better to kill the console framebuffer instead, which
+        gives better performance.
+
 ``drm`` (Direct Rendering Manager)
     Video output driver using Kernel Mode Setting / Direct Rendering Manager.
     Does not support hardware acceleration. Should be used when one doesn't
@@ -933,3 +1023,7 @@ Available video output drivers are:
     ``devpath=<filename>``
         Path to graphic card device.
         (default: /dev/dri/card0)
+
+    ``mode=<number>``
+        Mode ID to use (resolution, bit depth and frame rate).
+        (default: 0)

DOCS/mplayer-changes.rst

@@ -49,8 +49,10 @@ Input
 * Classic LIRC support was removed. Install remotes as input devices instead.
   This way they will send X11 key events to the mpv window, which can be bound
   using the normal ``input.conf``.
+  Also see: http://github.com/mpv-player/mpv/wiki/IR-remotes
 * Joystick support was removed. It was considered useless and was the cause
   of some problems (e.g. a laptop's accelerator being recognized as joystick).
+* Support for relative seeking by percentage.
 
 Audio
 ~~~~~
@@ -92,6 +94,8 @@ Video
 
 * Image subtitles (DVDs etc.) are rendered in color and use more correct
   positioning (color for image subs can be disabled with ``--sub-gray``).
+* Support for the X11 video output is removed, since it was considered
+  deprecated. SDL video output can still be used as a fallback.
 
 OSD and terminal
 ~~~~~~~~~~~~~~~~
@@ -275,7 +279,7 @@ Command Line Switches
     ``-xineramascreen``         ``--screen`` (different values)
     ``-xy W``                   ``--autofit=W``
     ``-zoom``                   Inverse available as ``--video-unscaled``
-    ``dvdnav://``               ``dvdnav://menu``
+    ``dvdnav://``               Removed.
     ``dvd://1``                 ``dvd://0`` (0-based offset)
     =========================== ========================================
 
@@ -395,7 +399,8 @@ Slave mode
 
   Assuming the system supports ``/dev/stdin``.
 
-  (The option was readded in 0.5.1 and sets exactly these options.)
+  (The option was added back in 0.5.1 and sets exactly these options. It was
+  removed in 0.10.x again.)
 
 * A JSON RPC protocol giving access to the client API is also supported. See
   `JSON IPC`_ for more information.

DOCS/tech-overview.txt

@@ -75,8 +75,7 @@ talloc.h & talloc.c:
     allocation call will never return NULL.
 
     One very useful feature of talloc is fast tracking of memory leaks. ("Fast"
-    as in it doesn't require valgrind.) You can enable it by passing the option
-    --leak-report as first parameter, or better, setting the
+    as in it doesn't require valgrind.) You can enable it by setting the
     MPV_LEAK_REPORT environment variable to "1":
         export MPV_LEAK_REPORT=1
     This will list all unfree'd allocations on exit.

DOCS/waf-buildsystem.rst

@@ -1,7 +1,7 @@
 waf build system overview
 =========================
 
-mpv's new build system is based on waf and it should completly replace the
+mpv's new build system is based on waf and it should completely replace the
 custom ./configure + Makefile based system inherited from MPlayer.
 
 Goals and the choice of waf
@@ -65,7 +65,7 @@ This defines a feature called ``vdpau`` which can be enabled or disabled by
 the users with configure flags (that's the meaning of ``--``). This feature
 depends on another feature whose name is ``x11``, and the autodetection check
 consists of running ``pkg-config`` and looking for ``vdpau`` with version
-``>= 0.2``. If the check succeds a ``#define HAVE_VDPAU 1`` will be added to
+``>= 0.2``. If the check succeeds a ``#define HAVE_VDPAU 1`` will be added to
 ``config.h``, if not ``#define HAVE_VDPAU 0`` will be added.
 
 The defines names are automatically prepended with ``HAVE_``, capitalized and

README.md

@@ -57,9 +57,6 @@ To build the software you can use `./waf build`: the result of the compilation
 will be located in `build/mpv`. You can use `./waf install` to install mpv
 to the *prefix* after it is compiled.
 
-NOTE: Using the old build system (with `./old-configure`) should still work,
-but will be removed in a future version of mpv.
-
 Essential dependencies (incomplete list):
 
 - gcc or clang
@@ -174,11 +171,9 @@ list of changes is located [here][mplayer-changes].
 Most activity happens on the IRC channel and the github issue tracker. The
 mailing lists are mostly unused.
 
- - **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])
+ - **GitHub issue tracker**: [issue tracker][issue-tracker] (report bugs here)
+ - **User IRC Channel**: `#mpv` on `irc.freenode.net`
+ - **Developer IRC Channel**: `#mpv-devel` on `irc.freenode.net`
 
 To contact the `mpv` team in private write to `mpv-team@googlegroups.com`. Use
 only if discretion is required.
@@ -187,8 +182,6 @@ only if discretion is required.
 [mpv-build]: https://github.com/mpv-player/mpv-build
 [homebrew-mpv]: https://github.com/mpv-player/homebrew-mpv
 [issue-tracker]:  https://github.com/mpv-player/mpv/issues
-[mpv-users]: https://groups.google.com/forum/?hl=en#!forum/mpv-users
-[mpv-devel]: https://groups.google.com/forum/?hl=en#!forum/mpv-devel
 [ffmpeg_vs_libav]: https://github.com/mpv-player/mpv/wiki/FFmpeg-versus-Libav
 [release-policy]: https://github.com/mpv-player/mpv/blob/master/DOCS/release-policy.md
 [windows_compilation]: https://github.com/mpv-player/mpv/blob/master/DOCS/compile-windows.md

RELEASE_NOTES

@@ -0,0 +1,979 @@
+Release 0.10.0
+==============
+
+
+Features
+--------
+
+New
+~~~
+
+- uchardet support
+- Matroska: reading cue sheets embedded in tags
+- Support for VideoToolbox hardware decoding
+- Display sync mode (--display-sync)
+- --force-window=immediate mode
+- fd:// protocol
+- libarchive wrapper for reading compressed archives
+- TOOLS/lua: zones.lua
+- Support for the "new" libavcodec VDPAU API
+- vf_vdpaurb, Add a new filter for reading back VDPAU decoded frames
+- DXVA2: HEVC support
+- Enabled HEVC profiles with VA API
+- HEVC added to whitelist of hwdec codecs
+- vo_null: framerate emulation
+- vo_opengl: support for custom shaders
+- vo_opengl: temporal-dither-period option
+- vo_opengl: tscale-clamp option
+- vo_opengl: option to attach target-prim/target-csp to window screenshots
+- vo_opengl_cb: "block" framedrop mode (now default)
+- vo_opengl_cb: support for interpolation
+- vo_vdpau: rotation support
+- ytdl_hook: support for 'multi_video' results
+
+
+Removed
+~~~~~~~
+
+- af_convert24 (af_lavrresample does this now)
+- af_dummy
+- audio: S8, U16, U24, U32 formats
+- DVD and BD menu support
+- TOOLS: youtube-starttime.lua
+- VA API: compatibility crap (< 0.34.0) and vo_vaapi deinterlacer
+- vo_x11
+
+
+Deprecated
+~~~~~~~~~~
+
+- vf_dlopen
+
+
+Behavior
+--------
+
+- audio: softvol scale is now cubic
+- cache: readahead size is limited to half the cache size at the beginning
+- charset_conv: "auto" encoding detection now prefers uchardet
+- demux_playlist: skip hidden directories
+- input.conf: O toggles between 2 states only
+- input.conf: Ctrl+s key binding for window screenshots
+- input.conf: mouse volume control (horizontal scrolling) inverted
+- input.conf: L to toggle infinite looping
+- input.conf: remap d/D keys (dropped framedrop cycle, replaced with deinterlacing toggle)
+- player: disabled seeking on unseekable streams even if the cache is enabled
+- player: parses and exposes m3u playlist titles
+- player: --term-playing-msg is now in a separate log category
+- player: removed automatic DVB channel advancement on no data
+- player: now restores video-aspect on playback resume
+- player: now uses exit code 0 by default for quit, 4 for signals, etc.
+- player: warns against using HLS URLs with --playlist
+- screenshots: changed the default directory in pseudo-gui mode to desktop
+- screenshots: screenshot directory is now created automatically
+- screenshots: default template is now prefixed with "mpv-"
+- TOOLS/lua/autoload: adds all files on start
+- vo: vo_wayland moved up in autoprobe list
+- vo_opengl: enabled pbo by default with opengl-hq
+- vo_opengl: cache dir for ICC profiles is now created automatically
+- w32: shift drag and drop appends
+- x11: shift drag and drop appends
+
+
+Options and Commands
+--------------------
+
+Added
+~~~~~
+
+- af_volume: replaygain-fallback option
+- ao_coreaudio: change-physical-format option
+- ao_coreaudio: exclusive option
+- ao_null: channel-layouts option for testing channel layout selection
+- audio: --audio-spdif as new method for enabling passthrough
+- cache: --cache-backbuffer to configure cache backbuffer size
+- command: define-section command for defining input bindings
+- command: audio-params and audio-out-params properties
+- command: keypress, keydown, and keyup commands
+- command: playlist_shuffle command
+- command: option-info/N/set-locally property indicating per-file options
+- command: protocol-list property
+- command: track-list/N/audio-channels property
+- demux: --demuxer-max-packets and --demuxer-max-bytes options to control maximum queue size
+- input: relative percentage seek
+- osc: time display configuration options
+- player: --playlist-pos option
+- screenshots: --screenshot-jpeg-source-chroma option to disable JPEG 4:4:4 output
+- screenshots: --screenshot-high-bit-depth option to allow or disallow 16 bit output
+- screenshots: --screenshot-directory option
+- sub: --stretch-image-subs-to-screen option for stretching image subtitles to screen
+- TOOLS/stats-conv: allow passing regex via command line
+- video: --video-aspect-method option to configure container vs. bitstream aspect ratio
+- vo_drm: mode suboption to set the mode ID to use
+- vo_opengl_cb, vo_opengl: --hwdec-preload option for preloading hwdec context
+- vo_rpi: background disabled by default
+- vo_xv: buffers suboption to configure number of buffers
+- win32: portable config mode
+
+
+Changed
+~~~~~~~
+
+- audio: changed the range of the volume option/property (0 is still silence, and 100 now always means unchanged volume)
+- command: allow changing deinterlace property any time
+- command: allow changing track properties while no file is loaded
+- command: always make video-aspect property accessible
+- command: better choice when to allow playback-related commands
+- command: change OSD symbol for absolute perc. seek
+- command: change the default action for rescan_external_files
+- command: change the hwdec properties
+- command: define-section with empty contents removes a section
+- command: export stereo 3D tags
+- command: make auto-deinterlacing output at field rate
+- command: make deinterlace property use interlaced-only yadif mode
+- command: make property event mask matching more restrictive
+- command: make the playback-time property writable
+- input: allow - as separator between commands, instead of _
+- options: --loop without argument means looping forever
+- options: make keyvalue list parsing less strict
+- player: extend --hls-bitrate option
+- vf_yadif: expose interlaced frame mode
+- video: --video-stereo-mode=no to disable automatic stereo conversion
+- vo_opengl_cb, vo_opengl: --hwdec-preload for preloading hwdec context
+- vo_opengl: replace icc-cache with icc-cache-dir
+- vo_opengl: icc-profile overrides icc-profile-auto
+
+
+Renamed
+~~~~~~~
+
+- command: rename audio-format property to audio-codec-name
+- options: rename --media-title option to --force-media-title
+- vo_opengl: rename use_full_range to use_normalized_range
+
+
+Deprecated
+~~~~~~~~~~
+
+- --ad-spdif-dtshd (use --audio-spdif=dts-hd)
+- audio-samplerate property
+- length property
+
+
+Removed
+~~~~~~~
+
+- get_property command
+- --demuxer-readahead-packets and --demuxer-readahead-bytes
+- image_writer: don't use jpeg baseline, and remove useless jpeg-optimize and jpeg-baseline options
+- --leak-report
+- --slave-broken
+- vo_opengl: npot suboption
+
+
+Fixes and Minor Enhancements
+----------------------------
+
+- ad_spdif: use DTS-HD passthrough only if the audio is really DTS-HD
+- af: fix behavior with filter chains that require a large number of auto-inserted conversion filters
+- af_lavcac3enc: fix A/V sync
+- ao_alsa: accept 7.1 over HDMI
+- ao_alsa: refuse to use spdif if AES flags can't be set
+- ao_wasapi: fix crash on hotplug init error
+- audio: avoid wasting CPU due to continuous wakeup
+- audio: do not exit when loading small files in paused mode
+- audio: fix channel map fallback selection
+- audio: fix crash on uninit
+- audio: fix --end handling
+- audio: fix EOF state with --keep-open
+- audio: fix restoring volume
+- charset_conv: fix switched parameters
+- charset_conv: use our own UTF-8 check with ENCA only
+- cocoa: don't load hardcoded icon if running from bundle
+- cocoa: hide cursor using a blank image instead of a system-wide API
+- command: do not exit playback if the B point of A-B loop is past EOF
+- command: fix audio-out-detected-device property
+- command: fix track property when no file is loaded
+- command: fix video-aspect property update notification
+- command: let track properties return option value in idle mode
+- demux: don't get stuck on some cases of timestamp resets
+- demux: handle Matroska-style replaygain tags as well
+- demux_lavf: do a better job at guessing the vobsub .sub filename
+- demux_mkv: disable ordered chapters if ChapterTimeEnd is missing
+- demux_mkv: discard broken index
+- demux_mkv: fix mpeg2 mapping
+- demux_mkv: ignore deprecated FrameRate, do not assume PAL
+- demux_mkv: improve video duration detection heuristic
+- demux_mkv: parse FLAC channel layouts
+- demux_playlist: make mime type comparison case-insensitive
+- dxva2: fix handling of cropped video
+- idet.sh: Support larger files
+- mp_image: fix vf_vdpaupp references
+- options: fix conversion of flags to strings
+- options: move program name to end of window title
+- options: remove the period at the end of "No file."
+- osc: completely disable if no VO window exists
+- osc: exit tick immediately if disabled
+- osc: reinit on playlist changes
+- osx: add NULL check for input context in a missing case
+- player: fix crashes when adding external tracks before loading main file
+- player: increase tick event update frequency
+- player: make decoding cover art more robust
+- player: raise maximum idle time
+- player: return better guess for playback time during seeks
+- player: show larger cache sizes in MB on status line
+- player: slim down A/V desync warning
+- sd_ass: assume negative durations are unknown durations, and handle them
+- terminal: disable terminal foreground state polling
+- terminal-unix: set terminal mode on init
+- timer: fix a corner case on clock changes
+- TOOLS: make autodeint detect telecine in parallel
+- TOOLS/zsh.pl: die loudly if mpv fails to run
+- vaapi: prefer direct display over copy-back
+- vaapi: fix some videos only showing up green
+- vaapi: treat cropped decoder output slightly more correctly
+- vda: add support for nv12 image formats
+- vd_lavc: fix a hw decoding fallback case
+- vf_stereo3d: drop internal implementation
+- vf_vavpp: don't attempt to deinterlace progressive frames
+- vf_vavpp: fix bob deinterlacing for bottom field first videos
+- vf_vdpaupp: Don't crash when evaluating interlacing of NULL mpi
+- video: always re-probe auto deint filter on filter reconfig
+- video: better heuristic for timestamp resets
+- video: fix panscan in vertical case
+- video: fix VideoToolbox/VDA autodetection
+- video: unbreak EOF with video-only files that have timestamp resets
+- vo_direct3d: fix broken pseudo GUI drag and drop hint
+- vo_drm: fix centering with regard to stride
+- vo_drm: fix crashes with --profile=pseudo-gui
+- vo_drm: fix resolution not restored after exiting
+- vo_drm: fix stride problem for certain devices
+- vo_drm: make VT switching non mandatory
+- vo: free frames before killing VO
+- vo_opengl: avoid broken shader if hwdec fails to provide textures
+- vo_opengl_cb: drop frames eagerly if frames are not rendered
+- vo_opengl: CMS no longer implies linear scaling
+- vo_opengl: fix alpha video in one case
+- vo_opengl: fix dangling pointers with vo_cmdline
+- vo_opengl: fix framestepping/pausing + interpolation
+- vo_opengl: fix "freezes" after seeking with interpolation on
+- vo_opengl: fix scale=oversample's threshold calculations
+- vo_opengl: framebuffers work under GLES 2
+- vo_opengl: improve robustness against PBO failure
+- vo_opengl: reimplement tscale=oversample
+- vo_opengl: reject future images in different formats
+- vo_opengl: X11: don't leak when GL init fails
+- vo: restore frame-drop logic for high-fps clips
+- vo_rpi: fix blackscreen before the first subtitle/OSD is rendered
+- vo_rpi, vo_opengl: do not globally terminate EGL on VO uninit
+- vo_sdl: fix glaring memory leak
+- vo_vdpau: check maximum video size
+- vo_vdpau: limit output surfaces to allowed maximum dimensions
+- win32: fix window resize logic
+- win32: fix crashes when changing system time
+- x11: Handle external fullscreen toggles
+- ytdl: catch bogus extractor info
+- ytdl: do not use deprecated media-title option
+- ytdl: don't print failure warning when youtube-dl was killed by us
+- ytdl: get start_time
+
+This listing is not complete. Check DOCS/client-api-changes.rst for a history
+of changes to the client API. A complete changelog can be seen by running
+`git log v0.9.2..v0.10.0` in the git repository or by visiting
+https://github.com/mpv-player/mpv/compare/v0.9.2...v0.10.0.
+
+
+Release 0.9.2
+=============
+
+Changes
+-------
+
+- The Lua check now also checks for `lua52.pc`, as used by Arch Linux testing.
+- (X11) `vo_opengl`'s `icc-profile-auto` now queries the current ICC profile
+  relative to the center of the window.
+- `ao_coreaudio`, `ao_alsa` now support adding dummy padding channels for
+  better compatibility with hardware decoders that only support specific
+  channel counts (e.g. 5.1 now should work on a decoder that only accepts 7.1).
+- Channel fallback (in case the audio device doesn't natively support a given
+  channel layout) has been improved.
+- `vf_vapoursynth` now rejects unaligned video instead of outputting corrupted
+  video.
+- mpv now tries to autoload `.sup` subtitles as well.
+
+Bug fixes
+---------
+
+- `vo_opengl`'s default for `fbo-format` is now `rgba16`, to avoid rounding
+  errors when using non-default `cscale` (issue #1918).
+- Improved framedrop behavior when playing video that has higher framerate
+  than the display (issue #1897).
+- Trying to play a directory will no longer spam `Connection lost!` to the
+  console log.
+- (Linux) Several `vo_rpi` bugfixes.
+- (Linux) Several `vo_drm` bugfixes. Pan&Scan is now supported.
+- (X11) Fix fullscreen behavior on certain window managers (issues #1937,
+  #1920).
+- (OSX) The OSD no longer always shows up on startup.
+- (OSX) Several `ao_coreaudio` and `ao_coreaudio_exclusive` bugfixes.
+- (OSX) Fixed potential crash on exit when using Cocoa.
+- (ClientAPI) `vo_opengl_cb` now actually applies options changed at runtime.
+- (OSX, ClientAPI) Cocoa now works when both the `cplayer` (`mpv`) and `libmpv`
+  are built at the same time; however, `libmpv` now always creates an
+  application singleton. Cocoa has to be disabled completely to prevent
+  `libmpv` from creating the singleton.
+
+This listing is not complete. A complete changelog can be seen by running
+`git log v0.9.1..v0.9.2` in the git repository or by visiting
+https://github.com/mpv-player/mpv/compare/v0.9.1...v0.9.2.
+
+Release 0.9.1
+=============
+
+Changes
+-------
+
+- mpv's IRC channel moved from #mpv-player to #mpv on chat.freenode.net.
+- Documentation updates.
+- The default value for the `--ytdl-format` option is now `best`, in order to
+  work around `youtube-dl`'s 2015.04.26 release enabling DASH by default, as
+  FFmpeg / Libav do not yet properly support DASH.
+- When seeking, the current timestamp will show the predicted seek timestamp
+  instead, until the final timestamp is resolved. Improves UI responsiveness
+  on slow streams and/or large seeks.
+
+Bug fixes
+---------
+
+- Corrected the release marker on DOCS/client-api-changes. The release manager
+  forgot to fix it before release...
+- Fix `vo_vdpau` rendering garbage lines on H.264 video with non-mod16 size
+  (issue #1863).
+- Fix a crash on exit if the "sub_reload" command had run successfully.
+- Fixed seeking with the mouse when `osc-seekbarstyle=bar` is set (issue
+  #1876).
+- (IPC) Fixed encoding of UTF-8 data in JSON (issue #1874).
+
+New features
+------------
+
+This listing is not complete. A complete changelog can be seen by running
+`git log v0.9.0..v0.9.1` in the git repository or by visiting
+https://github.com/mpv-player/mpv/compare/v0.9.0...v0.9.1.
+
+Release 0.9.0
+=============
+
+Changes
+-------
+
+Changes that may break users' config files have been annotated with a `(!)`.
+
+- Note: mpv is not compatible with Lua 5.3. Lua 5.1 or 5.2 is required.
+- The minimum required libass version is now 0.12.1 or newer.
+- The minimum required FFmpeg version is now 2.4.0 (equiv. Libav 11) or newer.
+- The internal libmpg123 support was removed. This was already not used by
+  default in the previous release.
+- `(!)` The LIRC support was removed. Configure LIRC remotes as input devices
+  instead.
+- `(!)` The Linux Joystick support was removed.
+- `(!)` `vf_screenshot` was removed, as they are now handled at a VO level and
+  is compatible with all VOs.
+- `(!)` `--ass-use-margins` has been renamed to `--sub-use-margins` and applies
+  only to plain-text (non-ASS) subtitles (enabled by default). The new
+  `--ass-force-margins` option applies only to ASS subtitles (disabled by
+  default). To get the old behaviour back, enable both at the same time.
+- `(!)` The `--sub-scale-with-window` option now only applies to plain text
+  (non-ASS) subtitles (enabled by default). The new `--ass-scale-with-window`
+  option does the same but only with ASS subtitles (disabled by default).
+- `(!)` The range for the `param1` for the `gaussian` `vo_opengl` scaler has
+  been redefined. Instead of being an arbitrary 1-100 range, have a default
+  value of 1.0, and anything higher is blurrier.
+- `(!)` The `seek`, `playlist_next, `playlist_prev`, `loadfile` and `loadlist`
+  parameters no longer accept numerical parameters where symbolic parameter
+  names exist.
+- `(!)` `vo_opengl` changes:
+  - `(!)` The `smoothmotion` suboption has been renamed to `interpolation`.
+    The old name is still supported for now.
+  - `(!)` The `bilinear_slow` scaler has been renamed to `triangle`.
+  - `(!)` `scale-down` has been renamed to `dscale` and now has its own set of
+    config options (e.g. `dscale-radius`).
+  - `(!)` Scaler radius no longer defaults to `3` but to a preferred value
+    that may be different for each filter.
+  - The `scale-radius` option may now go down as low as `0.5`, which is the
+    value used by the `nearest` filter.
+  - `spline36` is the new `cscale` default for `opengl-hq`. This might break
+    setups that use `fbo-format=rgb8`. To work around it, leave `fbo-format`
+    as its default, or set to something higher than 8, or set `cscale=bilinear`,
+    the previous default.
+- `(!)` `vf_format` no longer converts video to YUYV if there is no parameter.
+  Video is now passed unchanged unless a format is specifically requested.
+- `(!)` The `--colormatrix`, `--colormatrix-input-range`,
+  `--colormatrix-output-range` and `--colormatrix-primaries` options have been
+  converted into `vf_format` suboptions. See commit 27715b7 and the manual for
+  details.
+- `vf_mirror`'s implementation was replaced with calling into `libavfilter`'s
+  `vf_hflip` filter, thus depending on `libavfilter` to function.
+- The `device` subption to `ao_wasapi` has been deprecated in favor of
+  `--audio-device`.
+- `--video-rotate` now allows 360 as an argument instead of stopping at 359.
+- Several improvements to `af_scaletempo`.
+- Options that have multiple options and also include a "yes" option now
+  default to that if specified with no arguments.
+- The default value of `--cache-default` is now 150000 (153.6 MB, ~146 MiB).
+- JPEG screenshots now use the same subsampling as the source video. The images
+  are still RGB regardless of source format though.
+
+Bug fixes
+---------
+
+- mpv no longer saves position on files that can't be resumed (issue #1701).
+- (X11) Fix the player thinking the mouse has left the window in some WMs /
+  embeddings (issue #1672).
+- mpv no longer freezes on wayland when the compositor stops asking it to draw
+  itself (e.g. when minimized) (issue #249).
+- `.ac3` files are no longer rejected by `--audio-file-auto` (issue #1759).
+- `ao_wasapi` now automatically enables `exclusive` when passthrough is
+  attempted (issue #1742).
+- Attempt to fix flickering on Intel VAAPI drivers with `--hwdec=vaapi` and
+  `--vo=opengl` (issue #1765).
+- `youtube-dl` will no longer download video streams when video playback
+  is disabled with `--no-video`.
+- (Windows) mpv now prevents system sleep when playing a video-only file.
+  Previously, only files with an opened audio track would prevent sleep.
+
+New features
+------------
+
+- `vo_opengl` features:
+  - Added `ewa_ginseng`, `ewa_hanning`, `robidoux`, `robidouxsharp`,
+    `oversample` and `haasnsoft` scalers.
+  - There are now `ewa_lanczossoft` and `ewa_lanczossharp` aliases to
+    `ewa_lanczos` that are tuned to be blurrier and sharper, respectively.
+  - Added `gamma-auto` option that uses ambient light sensors to automatically
+    adjust the video gamma. See commit c028d78 for details.
+  - Added `blend-subtitles` option to draw subtitles directly into the video
+    instead of rendered afterwards. Potentially necessary for correct rendering
+    with files that use ASS subtitles for typesetting in combination with an
+    `icc-profile`. There is a default option for drawing after upscaling,
+    and a `video` option for drawing before upscaling. See details and warning
+    on the manual.
+  - There is now a `tscale` option, used to choose the temporal scaler used
+    in the `interpolation` mode (previously `smoothmotion`).
+  - There is a new `scale-blur` parameter to adjust the amount of blur that
+    most of the filters produce. Deviating from the default may introduce
+    artifacts in EWA filters.
+  - (Windows) There is now a `dwmflush` option that might help improve
+    rendering of high-fps video. Disabled by default. See manual for options.
+- New Linux-only `vo_drm` video output driver. Uses the direct rendering /
+  kernel modesetting drivers to draw directly to the framebuffer, but with
+  no hardware acceleration. See manual for details.
+- New `pseudo-gui` builtin profile, automatically used when launched from
+  `mpv.desktop` by opening `mpv.exe` on windows (`mpv.com` still works as
+  usual), or by opening the `mpv.app` bundle. The `pseudo-gui` tries to make
+  the player window behave closer to what a desktop player would do, by not
+  immediately closing and allowing the user to drag&drop files for playback.
+  See manual for details.
+- mpv can now play directories by automatically playing their contents instead.
+  Works everywhere but on Windows, due to issues with Windows' C runtime.
+- Add support to pitch correct stretched audio with librubberband.
+- Add support for the Raspberry Pi 2's hardware decoder when FFmpeg (or Libav)
+  is built with `--enable-mmal`. See commit 8fff125 for details.
+- The `--cache` option now accepts a "yes" option, that always enables a
+  `--cache-default`-sized cache on all cases a cache can be used, unless
+  `--cache-default` disables caching.
+- `ao_pulse`, `ao_coreaudio` and `ao_wasapi` now support device hotplugging.
+- New `--osd-align-x` and `--osd-align-y` options can be used to align the OSD
+  independently from subtitles.
+- New `--osd-bold` and `--sub-text-bold` options can be enabled to bold all
+  OSD or plain-text subtitle text, respectively.
+- Added a default keybind to the `u` key that enables/disables ASS style
+  overriding. When enabled, is equivalent to `--ass-style-override=force`.
+- There is now a `MOUSE_ENTER` keybind, that is called when the mouse cursor
+  enters the VO from outside.
+- The new `--ytdl-params` option can be set to arguments that are always given
+  to `youtube-dl` invocations. There is no sanity checking, so invalid options
+  can prevent `mpv` from working with http URLs.
+- There's a new `--demuxer-mkv-fix-timestamps` option, enabled by default, that
+  tries to guess more accurate video timestamps by using FPS information, if
+  available. See manual for details.
+- The new `--window-scale` option can be used to scale the video window by the
+  specified multiplier, before other options such as `--autofit` are applied.
+- `vo_direct3d_shaders` now supports NV12 colorspace without using stretchrect,
+  but it seems no drivers actually support that.
+- (Client API) Added a `rescan_external_files` command, as requested on issue
+  #1586.
+- (Client API) If enabled, initializing the Client API will now also load the
+  user's `mpv.conf`.
+- (Client API) There's a new `mpv_opengl_cb_report_flip()` call for API users
+  to call to inform mpv of when exactly a frame was displayed.
+- (Client API) The `mpv_opengl_cb_render()` function was deprecated in favor
+  of the new, simpler `mpv_opengl_cb_draw()` function.
+- (Client API) There is a new `screenshot_raw` command. See `input.rst` for
+  details.
+- The `--input-file` argument may now also be a file descriptor in the form
+  "fd://N", where N is the FD number.
+
+This listing is not complete. A complete changelog can be seen by running
+`git log v0.8.0..v0.9.0` in the git repository or by visiting
+https://github.com/mpv-player/mpv/compare/v0.8.0...v0.9.0 (only 250 commits).
+
+Release 0.8.3
+=============
+
+Changes
+-------
+
+- Documentation fixes and updates.
+- (Client API) Clarify `mpv_opengl_cb_render`'s viewport parameter behavior.
+
+Bug fixes
+---------
+
+- (X11) Fix crash on `vo_xv:no-colorkey` with the Overlay adapter. (bug #1629)
+- (X11) `--stop-screensaver` is now implemented by calling into
+  `xdg-screensaver`, fixing some compatibility issues that prevented it from
+  actually stopping screensavers.
+- Make the video equalizer work correctly on some VAAPI drivers. (bug #1647)
+- Prevent OSD from disappearing when clicking on mozplugger. (bug #1672)
+- The new DVB-S2 support code, using S2API, now builds on FreeBSD.
+- Fix decoding of seekable matroska from unseekable network stream. (bug #1656)
+- (OSX) Fix crashing when closing a VO's window (usually at exit). (bug #1657)
+- (OSX) Unhide the mouse cursor when over the Dock or Launchpad. (bug #513)
+- (OSX) Fix mouse cursor autohiding when the player is fullscreen.
+- `vf_vavpp` (VAAPI postprocessing) now deinterlaces video correctly.
+- `smb://` streams now use the stream cache (used to be inefficient).
+- (Windows) Do not hide the mouse cursor when it is hovering the window menu.
+- Fix 8-channel output on `ao_jack`. (bug #1688)
+- Fix `--mf-fps` parameter on JPEG files. (bug #1689)
+- Fix anamorphic scaling being ignored if it was very minor.
+
+This listing is not complete. A complete changelog can be seen by running
+`git log v0.8.2..v0.8.3` in the git repository or by visiting
+https://github.com/mpv-player/mpv/compare/v0.8.2...v0.8.3.
+
+Release 0.8.2
+=============
+
+Bug fixes
+---------
+
+- Fix OSD placement accidentally messed up in 0.8.1 (commit 22863d6).
+
+Release 0.8.1
+=============
+
+Changes
+-------
+
+- DOCS/client-api-changes has been correctly updated for the v0.8.0 release
+  series.
+- Quality and major perfomance improvements to smoothmotion.
+- Silence the "[ytdl_hook] WARNING: video doesn't have subtitles" warning.
+- Other documentation updates.
+
+Bug fixes
+---------
+
+- Attempt to fix OpenGL shader compilation on Intel windows drivers. (bug
+  #1536)
+- Attempt to improve the OpenGL shader's compatibility with GLES2.
+- (Windows) Fix noise when seeking while using wasapi:exclusive. (bug #1529)
+- Fix the waf-based build system when used with newer versions of waf with
+  python3.
+- Documentation fixes. (includes bug #1608)
+
+New features
+------------
+
+- mpv now tries to load '.vtt' subtitles.
+
+This listing is not complete. A complete changelog can be seen by running
+`git log v0.8.0..v0.8.1` in the git repository or by visiting
+https://github.com/mpv-player/mpv/compare/v0.8.0...v0.8.1.
+
+Release 0.8.0
+=============
+
+Changes
+-------
+
+Changes that may break users' config files have been annotated with a `(!)`.
+
+- `(!)` vo_opengl_old has been removed. OpenGL rendering now requires hardware
+  that can do at least OpenGL 2.1, the oldest version supported by vo_opengl.
+- `(!)` vf_pp has been removed. A version of it can still be accessed through
+  lavfilter (e.g. --vf=lavfi=[pp...]), but it receives no QP information.
+  Also, vf_dlopen no longer passes QP information to its loaded plugin either
+  (it's always 0).
+- `(!)` vf_softpulldown, vf_swapuv, vf_phase, vf_divtc and vf_ilpack have all
+  been removed. Ports or alternative versions of these filters are available
+  through libavfilter (vf_lavfi).
+- vf_pullup and vf_noise now are simply wrappers to their libavfilter
+  equivalents. Both are only currently available on FFmpeg (not Libav).
+- `(!)` ao_portaudio has been removed. There is good support for the native
+  audio drivers of all major desktop platforms, as well as ao_sdl and
+  ao_openal as fallbacks.
+- `(!)` vo_opengl's `lscale` suboption has been renamed to `scale`; `lradius`
+  is now `scale-radius`, `lparam1` and `lparam2` are now respectively
+  `scale-param1` and `scale-param2`; `lscale-down` is now `scale-down`.
+  The `cscale` suboption remains as is.
+- `(!)` Several of the `vo_opengl` radius-preset aliases supported by `scale`
+  have been removed; use `scale-radius` to set if needed. For example, use
+  `--vo=opengl:scale=lanczos:scale-radius=2` instead of `scale=lanczos2`.
+  The default radius is recommended for most filters.
+- `(!)` vo_opengl no longer supports the `stereo` suboption. The anaglyph
+  effect can be reproduced with the stereo3d filter. The quadbuffer support,
+  which requires expensive specialized hardware to begin with, is removed.
+- `(!)` The `approx-gamma` suboption to vo_opengl has been removed. The
+  BT.1886 specification says that it's actually how it's supposed to be done
+  so it is now the default when `srgb` or `icc-profile` are enabled. This does
+  not include BT.1886's gamma drop.
+- `(!)` The `no-scale-sep` and `indirect` suboptions to vo_opengl have been
+  removed. These are now autodetected and enabled whenever they would have
+  benefit.
+- `(!)` The `--lua` and `--lua-opts` parameters / options are now called
+  `--script` and `--script-opts`, respectively. The `lua` subdir of the mpv
+  config dir is also now expected to be called `scripts` instead.
+- `(!)` The `--fixed-vo` option has been removed. It has been the default for
+  a long time anyway, and disabling fixed-vo is not useful.
+- vo_opengl_hq has been updated to take into account new features.
+  `fancy-downscaling` and `sigmoid-upscaling` are enabled,
+  "mitchell" is now the default for `scale-down`.
+- The `sub-visibility` OSD message is now clearer about whether subtitles
+  are hidden or just not available / selected.
+- The device IDs given to `--audio-device` for use with ao_coreaudio now use
+  UIDs, so they don't change when devices are added/removed or after reboots.
+- `--msg-level` now also accepts ',' as separator.
+- (Client API) The client API now refuses to initialize if the LC_NUMERIC
+  locale is not "C".
+- (Client API) The native type of the `msg-level` parameter is now a key-value
+  list. Setting or reading it as a string still works.
+- (Slave API) The `get_property` command is now deprecated.
+- Documentation updates.
+- PDF documentation is now disabled by default due to rst2pdf being fickle,
+  causing weird build errors.
+- Print desync messages with negative A/V sync as well. While rare, it could
+  happen with some uses of `--autosync`.
+
+Bug fixes
+---------
+
+- (Windows) Avoid resizing the video window when player is minimized, might
+  address bug #1547.
+- (NetBSD) Fix build with v4l2.
+- (Linux) Attempt to address conflicts with the pulseaudio mixer. (bug #1578)
+- Multiple cdda:// fixes. (bugs #1555, #1560)
+- VP9 timestamps no longer cause "missing PTS" warnings with new enough FFmpeg
+  builds.
+- Fix a crash when using H.264 hardware decoding on new enough libavcodecs.
+  (bug #1587)
+
+New features
+------------
+
+- vo_opengl now supports frame blending to eliminate stuttering when the video
+  framerate does not nicely match the display framerate through the
+  `smoothmotion` suboption. This is not meant to artificially increase the
+  video's FPS, so there is no "soap opera effect" or difficulties with some
+  video types (e.g. anime).
+- vo_opengl now supports sigmoidal upscaling (e.g. for fullscreen), which
+  reduces ringing induced by upscaling, enabled through the
+  `sigmoid-upscaling` suboption.
+- vo_opengl now supports ewa_lanczos (Jinc) scaling, which provides higher
+  quality with less aliasing. It supports an experimental `scale-antiringing`
+  parameter, which tries to further reduce video ringing.
+- vo_opengl now has a `linear-scaling` suboption, that makes the scalers work
+  in linear light. Implied by the `srgb`, `icc-profile` or the new
+  `sigmoid-upscaling` suboption.
+- vo_opengl has improved downscaling for higher downscale ratios when
+  `fancy-downscaling` is enabled.
+- Add `--keep-open=always` to make `--keep-open` apply to all files instead of
+  only the last.
+- mpv now filters the tags that are printed on the console to try to hide
+  useless metadata. The filter works as a whitelist, and can be configured
+  with the `--display-tags` option.
+- Add a `--sub-scale-by-window` option that, when set to 'no', prevents
+  subtitles from changing their pixel size when the window changes sizes.
+- vo_opengl now technically has GLES 2 and 3 support (but not GLES 1). GLES 3
+  support is only tested on Mesa's software emulation. GLES 2 has been tested
+  on nvidia drivers. Both GLES modes are feature deficient and inefficient due
+  to GLES limitations, so they should not be preferred over the standard
+  OpenGL mode.
+- (X11) vo_opengl now supports the `icc-profile-auto` option on X11 hosts.
+- The 'A' key now by default cycles through a list of preset aspect ratios.
+  Meant to be used to work around broken files.
+- There is now a "force" mode for `--loop`. Works like "inf", but tries to
+  open files/streams even if there was an error on the previous attempt.
+- There is now a `--log-file` option to write mpv's log messages to a specifed
+  file.
+- There is now a `--audio-file-auto` option to automatically load an external
+  audio file with certain constraints. Will only load external audio if the
+  playing file has a video stream. Set to "exact" by default. (bug #967)
+- There is now a `--network-timeout` option to specify a timeout in seconds
+  for network access. If 0 (default), uses the libavformat default. If a
+  protocol that doesn't support timeouts is used, this option is ignored.
+- (X11) The XF86 special keys are now all mappable in `input.conf`. All keys
+  documented in `XF86keysym.h` are available, but only as their numeric codes.
+- (Linux) The DVB implementation now supports DVB-S2 through S2API.
+- (Linux) The VDR format for `channels.conf` is now supported and preferred.
+  See http://linuxtv.org/vdrwiki/index.php/Syntax_of_channels.conf for the
+  syntax. Tuning to DVB-S2 channels requires the VDR-format `channels.conf`.
+- (Linux) There is now a `--dvbin-full-transponder` option for handling
+  special broadcast cases where PIDs switch, or just to work around incomplete
+  PID data.
+- (Client API) There is now an opengl_cb VO, which receives the OpenGL context
+  from the libmpv client. This allows a client to render libmpv video directly
+  to a provided OpenGL context without having to use --wid embedding.
+- (Client API) The aforementioned --wid parameter can now be set at any time,
+  but has no effect if a file is already loaded.
+- (Client API) The list of DVD titles can now be queried.
+- (Client API) There is now a `filtered-metadata` property, containing only
+  the tags allowed by `--display-tags`.
+- (Client API) There is now support for XEmbed `--wid` targets, such as a
+  `GtkSocket`'s `gtk_socket_get_id()`.
+- (Client API) There is now a `file-format` property with a symbolic name
+  for the file format. In some cases this might be a comma separated list
+  of various different extensions due to libavformat idiosyncrasies.
+- (Client API) There is now a `mouse` command for generating mouse events
+  over the video window.
+- (Client API) There is now a `partially-seekable` property that indicates
+  whether a file is only considered `seekable` because of the stream cache,
+  but would not be seekable otherwise (e.g. realtime stream, or HTTP with
+  no resume support). Small relative seeks may be fine, but larger seeks
+  will likely fail.
+- (Client API) There is now an MPV_EVENT_QUEUE_OVERFLOW event that is posted
+  whenever events have to be dropped because the queue is full. No further
+  events are posted until the MPV_EVENT_QUEUE_OVERFLOW is consumed to avoid
+  duplicated posting.
+- (Client API) There is now a helper `mpv_wait_async_requests()` function
+  that blocks until all known async requests have completed. (bug #1542)
+- (Client API) There is now a `detected-hwdec` property that returns the
+  detected hardware decoder when one is successfully opened. This probably
+  only returns a valid value after playback has started.
+- (Client API) There are now `audio_add`, `audio_remove` and `audio_reload`
+  counterparts to the equivalent `sub_` commands for handling external audio
+  files.
+
+This listing is not complete. A complete changelog can be seen by running
+`git log v0.7.0..v0.8.0` in the git repository or by visiting
+https://github.com/mpv-player/mpv/compare/v0.7.0...v0.8.0 (only 250 commits).
+
+Release 0.7.3
+=============
+
+Changes
+-------
+
+- Several documentation updates and typo fixes.
+- Don't show "0%" position when the stream is infinite / has unknown length.
+- Fixes the config file loading order so that lower priority `mpv.conf` files
+  don't override higher priority `config` files.
+- (OSX) Remove coreaudio_exclusive from the "auto" codec list.
+- Failing to create a GL3 context is now a warning instead of error, to reduce
+  user confusion.
+- The subtitle decoder now gets reset when cycling subtitles. This makes the
+  currently shown subtitle event disappear even if cycling back to the current
+  subtitle track.
+- `--shuffle` and `--merge-files` now affects the contents of playlist files,
+  instead of just the list of files given as arguments to mpv.
+- `./waf install` avoids installing a few additional data/config files if only
+  libmpv was built.
+- Improved channel mapping when the file's channel map doesn't match the AO's
+  available channel maps.
+- (OSX) VDA now gracefully refuses to run on non-OpenGL3-capable systems instead
+  of trying and failing to build its shaders.
+- Add workaround for vf_vapoursynth filters that expect an FPS.
+- The default value for `--screenshot-template` now shows up in
+  `--list-options`.
+- `ao_pulse`'s `latency-hacks` suboption is now off by default, as it causes
+  issues with newer pulseaudio releases. If A/V Sync issues happen, either
+  add the `latency-hacks=on` pulse suboption in mpv.conf, or update the
+  pulseaudio daemon (bug #1430).
+- (Linux) `ao_alsa` now deals better with audio device disconnection.
+- Client API: timestamp properties that have no value return "no" instead of
+  magic number.
+
+Bug fixes
+---------
+
+- (Windows, OSX) Many `ao_wasapi` and `ao_coreaudio` fixes.
+- (OSX) Make the window title update correctly on OS X Yosemite.
+- Fixes for handling mono audio on various AOs.
+- (Linux) Fixes resuming from suspend on ao_alsa.
+- Fixes for playlist file parsing.
+- Overly long options in the --list-options output now break the column layout
+  instead of getting truncated; fixes shell completion scripts.
+- Fix rendering resolution of certain DVB subtitles (bug #1425).
+- Fix EDL or --merge-files breaking timestamps with .avi files.
+- Workaround for libavcodec bug with the VP9 codec parser (bug #1448),
+  fixes possible crash.
+- Improve robustness of the matroska parser with broken files (bugs #1457,
+  #1461).
+- Improve 10bit video compatibility with older GPUs (specially Intel) (bug
+  #1383).
+- Fixes flashing the VO window when playing a list of files that includes
+  non-media files (bug #1459).
+- Workaround for window embedding in OpenBox (bug #1235)
+- Fixes for several crashes and lockups (bugs #838, #1389, #1408, #1463, #1473,
+  #1474).
+- Fixes for the zsh completion script (bugs #997)
+
+New features
+------------
+
+- mpv now prints the contents of its config.h when running in verbose mode, to
+  aid with debugging.
+- The `include=` option now accepts `~` to refer to the HOME dir (bug #1406).
+- `af_volume` now prints ReplayGain values in verbose mode.
+- m3u playlists that don't have the `#EXTM3U` header are now accepted if they
+  "look like" ASCII or UTF-8 text and have the .m3u extension.
+- Chapter marks in the seek bar now update when switching files.
+- Supports embedded cover art in MKV files (bug #1374).
+- (Windows) Video window can now be resized even with --no-border.
+- (Windows) Client API: the "run" command now works on Windows too.
+- (Linux) vo_wayland now supports key modifiers (Meta, Alt, Control, Shift)
+- Client API: "display-names" property is now observable.
+
+This listing is not complete. A complete changelog can be seen by running
+`git log v0.7.2..v0.7.3` in the git repository or by visiting
+https://github.com/mpv-player/mpv/compare/v0.7.2...v0.7.3.
+
+Release 0.7.2
+=============
+
+Changes
+-------
+
+- Give precedence to the DVD menu navigation keyboard bindings so that user
+  defined LEFT/RIGHT/... bindings don't break DVD menu navigation.
+- Try to fallback to the "default" device if the selected device is busy in the
+  alsa AO.
+- Don't create Dock icon for audio only files on OS X.
+- Save screenshots to desktop when using the app bundle on OS X.
+- Restore ab-loop settings with playback resume.
+- Bump required youtube-dl version to 2014.11.26 and enable the ytdl_hook Lua
+  script by default (now playing videos from YouTube and the like will work
+  out of the box without any configuration change needed).
+
+Bug fixes
+---------
+
+- Don't signal an error if --stream-dump is used.
+- Fix removing key bindings from Lua scripts.
+- Reject channel descriptions with too many channels in the coreaudio AO.
+- Don't async redraw when waiting for VO redraw on OS X (this fixes the very
+  annoying glitch where the black bars disappear for a single frame when going
+  fullscreen).
+- Fix mono playback with the also AO.
+- Don't crash if framebuffers are not available in the opengl VO.
+
+New features
+------------
+
+- Try to handle multi-arc videos in the ytdl_hook Lua script.
+
+This listing is not complete. There are many more bug fixes and changes. The
+complete change log can be viewed by running ``git log 47ec404..c7d6b21`` in
+the git repository or by visiting
+https://github.com/mpv-player/mpv/compare/47ec404...c7d6b21.
+
+Release 0.7.1
+=============
+
+Changes
+-------
+
+- Don't show the volume neutral marker on the OSD if softvol is disabled.
+- Don't select a subtitle track when executing the sub_add input command in
+  "auto" mode.
+
+Bug fixes
+---------
+
+- Fix busy loop when seeking while paused (this fixes a problem with pulseaudio
+  that caused mpv and the pulseaudio daemon to use 100% CPU).
+- Fix Lua function utils.subprocess() in Windows versions older than Vista.
+- Avoid creating a window bigger than the screen on Windows.
+- Don't ignore the last line in m3u playlists.
+- Don't crash if a codec could not be opened.
+- Dynamically allocate audio channel map entries (this should fix a crash in
+  the alsa and coreaudio AOs with audio devices that support more than 20
+  channel maps).
+- Ignore the "srgb" option in the opengl VO if hardware decoding is enabled.
+- Linearize non-RGB sRGB files correctly (eg. JPEG).
+- Fix opening reference URLs (.file/id=) on OS X.
+
+This listing is not complete. There are many more bug fixes and changes. The
+complete change log can be viewed by running ``git log 8d8b36d..6583ad6`` in
+the git repository or by visiting
+https://github.com/mpv-player/mpv/compare/8d8b36d...6583ad6.
+
+Release 0.7.0
+=============
+
+Changes
+-------
+
+- 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
+------------
+
+- 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 7759c18..9479daa`` in
+the git repository or by visiting
+https://github.com/mpv-player/mpv/compare/7759c18...9479daa (only 250 commits).

TOOLS/idet.sh

@@ -47,7 +47,7 @@ judge()
         progressive=$((progressive + progressive1))
         undetermined=$((undetermined + undetermined1))
     done <<EOF
-$(testfun "$@")
+$(testfun "$@" | sed 's/:/: /g')
 EOF
 
     interlaced=$((bff + tff))

TOOLS/lib/Parse/Matroska/Definitions.pm

@@ -84,7 +84,7 @@ sub elem_by_hexid {
 # used by elem when setting the 'valname' key
 use constant TYPE_MAP => {
     uint    => 'uint64_t',
-    str     => 'struct bstr',
+    str     => 'char *',
     binary  => 'struct bstr',
     ebml_id => 'uint32_t',
     float   => 'double',

TOOLS/lua/autodeint.lua

@@ -2,21 +2,16 @@
 -- 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.
+-- It registers the key-binding ctrl+d, which when pressed, inserts the filters
+-- ``vf=lavfi=idet,pullup,vf=lavfi=idet``. After 4 seconds, it removes these
+-- filters and decides whether the content is progressive, interlaced, or
+-- telecined and the interlacing field dominance.
 --
--- 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.
+-- Based on this information, it may set mpv's ``deinterlace`` property (which
+-- usually inserts the yadif filter), or insert the ``pullup`` filter if the
+-- content is telecined.  It also sets mpv's ``field-dominance`` property.
 --
+-- OPTIONS:
 -- The default detection time may be overridden by adding
 --
 -- --script-opts=autodeint.detect_seconds=<number of seconds>
@@ -37,6 +32,7 @@ require "mp.msg"
 script_name = mp.get_script_name()
 detect_label = string.format("%s-detect", script_name)
 pullup_label = string.format("%s", script_name)
+ivtc_detect_label = string.format("%s-ivtc-detect", script_name)
 
 -- number of seconds to gather cropdetect data
 detect_seconds = tonumber(mp.get_opt(string.format("%s.detect_seconds", script_name)))
@@ -69,30 +65,29 @@ function start_detect()
     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)
+    local cmd = string.format('vf add @%s:lavfi=graph="idet",@%s:pullup,@%s:lavfi=graph="idet"',
+                              detect_label, pullup_label, ivtc_detect_label)
     if not mp.command(cmd) then
-        mp.msg.error("failed to insert detection filter")
+        mp.msg.error("failed to insert detection filters")
         return
     end
 
     -- wait to gather data
-    timer = mp.add_timeout(detect_seconds, judge_field_dominance)
+    timer = mp.add_timeout(detect_seconds, select_filter)
 end
 
 function stop_detect()
     del_filter_if_present(detect_label)
+    del_filter_if_present(ivtc_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()
+function judge(label)
     -- get the metadata
-    local result = mp.get_property_native(string.format("vf-metadata/%s", detect_label))
+    local result = mp.get_property_native(string.format("vf-metadata/%s", 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"])
@@ -100,10 +95,10 @@ function judge()
     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)
+    mp.msg.verbose(label.." progressive    = "..num_progressive)
+    mp.msg.verbose(label.." interlaced-tff = "..num_tff)
+    mp.msg.verbose(label.." interlaced-bff = "..num_bff)
+    mp.msg.verbose(label.." undetermined   = "..num_undetermined)
 
     if num_determined < num_undetermined then
         mp.msg.warn("majority undetermined frames")
@@ -119,11 +114,12 @@ function judge()
     end
 end
 
-function judge_field_dominance()
-    verdict = judge()
+function select_filter()
+    -- handle the first detection filter results
+    verdict = judge(detect_label)
     if verdict == progressive then
-        stop_detect()
         mp.msg.info("progressive: doing nothing")
+        stop_detect()
         return
     elseif verdict == interlaced_tff then
         mp.set_property("field-dominance", "top")
@@ -133,21 +129,8 @@ function judge_field_dominance()
         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()
-
+    -- handle the ivtc detection filter results
+    verdict = judge(ivtc_detect_label)
     if verdict == progressive then
         mp.msg.info(string.format("telecinied with %s field dominance: using pullup", mp.get_property("field-dominance")))
         stop_detect()
@@ -157,8 +140,6 @@ function judge_ivtc()
         mp.set_property("deinterlace","yes")
         stop_detect()
     end
-
-    return
 end
 
 mp.add_key_binding("ctrl+d", script_name, start_detect)

TOOLS/lua/autoload.lua

@@ -4,8 +4,8 @@
 -- alphabetically, and adds entries before and after the current file to
 -- the internal playlist. (It stops if the it would add an already existing
 -- playlist entry at the same position - this makes it "stable".)
--- Add at most 5 * 2 files when starting a file (before + after).
-MAXENTRIES = 5
+-- Add at most 5000 * 2 files when starting a file (before + after).
+MAXENTRIES = 5000
 
 function Set (t)
     local set = {}
@@ -47,6 +47,13 @@ function find_and_add_entries()
     if #dir == 0 then
         return
     end
+    local pl_count = mp.get_property_number("playlist-count", 1)
+    if (pl_count > 1 and autoload == nil) or
+       (pl_count == 1 and EXTENSIONS[string.lower(get_extension(filename))] == nil) then
+        return
+    else
+        autoload = true
+    end
 
     local files = mputils.readdir(dir, "files")
     if files == nil then

TOOLS/lua/drc-control.lua

@@ -4,7 +4,7 @@
 -- parameters. See the entry for "drc" under the "AUDIO FILTERS"
 -- section of the man page for a complete description of the filter.
 --
--- This script registers the key-binding "\" to toggle the filter between 
+-- This script registers the key-binding "\" to toggle the filter between
 --
 -- * off
 -- * method=1 (single-sample smoothing)
@@ -92,7 +92,7 @@ function drc_quieter_handler()
     drc_scale_target(0.5)
 end
 
--- toggle between off, method 1 and method 2 
+-- toggle between off, method 1 and method 2
 mp.add_key_binding("\\", "drc_toggle_method", drc_toggle_method_handler)
 -- increase or decrease target volume
 mp.add_key_binding("ctrl+9", "drc_quieter", drc_quieter_handler)

TOOLS/lua/status-line.lua

@@ -1,5 +1,7 @@
 -- Rebuild the terminal status line as a lua script
 -- Be aware that this will require more cpu power!
+-- Also, this is based on a rather old version of the
+-- builtin mpv status line.
 
 -- Add a string to the status line
 function atsl(s)
@@ -28,7 +30,7 @@ function update_status_line()
     atsl(mp.get_property_osd("time-pos"))
 
     atsl(" / ");
-    atsl(mp.get_property_osd("length"));
+    atsl(mp.get_property_osd("duration"));
 
     atsl(" (")
     atsl(mp.get_property_osd("percent-pos", -1))

TOOLS/lua/youtube-starttime.lua

@@ -1,34 +0,0 @@
---sets the startime of a youtube video as specified in the "t=HHhMMmSSs" part of the url
---NOTE: This might become obsolete once youtube-dl adds the functionality
-
-local msg = require 'mp.msg'
-
-function youtube_starttime()
-  url = mp.get_property("path", "")
-  start = 0
-
-  if string.find(url, "youtu%.?be") and
-    ((url:find("http://") == 1) or (url:find("https://") == 1)) then
-      time = string.match(url, "[#&%?]t=%d*h?%d*m?%d+s?m?h?")
-      --the time-string can start with #, & or ? followed by t= and the timing parameters
-      --at least one number needs to be present after t=, followed by h, m, s or nothing (>implies s)
-
-      if time then
-        for pos in string.gmatch(time,"%d+%a?") do
-          if string.match(pos,"%d+h") then            --find out multiplier for
-            multiplier = 60*60                        --hours
-          elseif string.match(pos,"%d+m") then
-            multiplier = 60                           --minutes
-          else multiplier = 1 end                     --seconds
-
-          start = start + (string.match(pos,"%d+") * multiplier)
-        end
-
-        msg.info("parsed '" .. time .. "' into '" .. start .. "' seconds")
-      end
-
-      mp.set_property("file-local-options/start",start)
-  end
-end
-
-mp.add_hook("on_load", 50, youtube_starttime)

TOOLS/lua/zones.lua

@@ -0,0 +1,74 @@
+--  zones.lua: mpv script for handling commands depending on where the mouse pointer is at,
+--             mostly for mouse wheel handling, by configuring it via input.conf, e.g.:
+--
+--  Ported from avih's ( https://github.com/avih ) zones.js
+--
+--  Vertical positions can be top, middle, bottom or "*" to represent the whole column.
+--  Horizontal positions can be left, middle, bottom or "*" to represent the whole row.
+--  "default" will be the fallback command to be used if no command is assigned to that area.
+--
+--  input.conf example of use:
+-- # wheel up/down with mouse
+-- MOUSE_BTN3 script_message_to zones commands "middle-right: add brightness  1" "*-left: add volume  5" "default: seek  10"
+-- MOUSE_BTN4 script_message_to zones commands "middle-right: add brightness -1" "*-left: add volume -5"  "default: seek -10"
+
+local ZONE_THRESH_PERCENTAGE = 20;
+-- sides get 20% each, mid gets 60%, same vertically
+
+local msg = mp.msg
+
+function getMouseZone()
+    -- returns the mouse zone as two strings [top/middle/bottom], [left/middle/right], e.g. "middle", "right"
+
+    local screenW, screenH = mp.get_osd_resolution()
+    local mouseX, mouseY   = mp.get_mouse_pos()
+
+    local threshY = screenH * ZONE_THRESH_PERCENTAGE / 100
+    local threshX = screenW * ZONE_THRESH_PERCENTAGE / 100
+
+    local yZone = (mouseY < threshY) and "top"  or (mouseY < (screenH - threshY)) and "middle" or "bottom"
+    local xZone = (mouseX < threshX) and "left" or (mouseX < (screenW - threshX)) and "middle" or "right"
+
+    return yZone, xZone
+end
+
+function main (...)
+    local arg={...}
+    msg.debug('commands: \n\t'..table.concat(arg,'\n\t'))
+
+    local keyY, keyX = getMouseZone()
+    msg.debug("mouse at: " .. keyY .. '-' .. keyX)
+
+    local fallback = nil
+
+    for i, v in ipairs(arg) do
+        cmdY = v:match("^([%w%*]+)%-?[%w%*]*:")
+        cmdX = v:match("^[%w%*]*%-([%w%*]+)%s*:")
+        cmd  = v:match("^[%S]-%s*:%s*(.+)")
+        msg.debug('cmdY: '..tostring(cmdY))
+        msg.debug('cmdX: '..tostring(cmdX))
+        msg.debug('cmd : '..tostring(cmd))
+
+        if (cmdY == keyY and cmdX == keyX) then
+            msg.verbose("running cmd: "..cmd)
+            mp.command(cmd)
+            return
+        elseif  (cmdY == "*"  and cmdX == keyX) or
+                (cmdY == keyY and cmdX == "*") then
+            msg.verbose("running cmd: "..cmd)
+            mp.command(cmd)
+            return
+        elseif cmdY == "default" then
+            fallback = cmd
+        end
+    end
+    if fallback ~= nil then
+        msg.verbose("running cmd: "..fallback)
+        mp.command(fallback)
+        return
+    else
+        msg.debug("no command assigned for "..keyY .. '-' .. keyX)
+        return
+    end
+end
+mp.register_script_message("commands", main)

TOOLS/old-configure

@@ -4,6 +4,13 @@
 # Cleanups all over the place (c) 2001 pl
 # Rewritten for mpv in 2014.
 #
+#
+# Warning: this is not officially supported. Use on your own risk. Might require
+#          modifications when used on systems other than Linux. Contributors
+#          are not expected to update these files when making changes to the
+#          normal/preferred waf build system.
+#
+#
 # This configure script is *not* autoconf-based and has different semantics.
 # It attempts to autodetect all settings and options where possible. It is
 # possible to override autodetection with the --enable-option/--disable-option
@@ -176,6 +183,7 @@ options_state_machine() {
     opt_yes_no _dvdread     "libdvdread"
     opt_yes_no _dvdnav      "libdvdnav"
     opt_yes_no _enca        "ENCA charset oracle library"
+    opt_yes_no _uchardet    "uchardet charset detection library"
     opt_yes_no _libass      "subtitle rendering with libass"
     opt_yes_no _libavdevice "libavdevice demuxers"
     opt_yes_no _libavfilter "libavfilter"
@@ -208,6 +216,7 @@ options_state_machine() {
     opt_yes_no _lua         "Lua scripting"
     opt_yes_no _vapoursynth "VapourSynth filter bridge (Python)"
     opt_yes_no _vapoursynth_lazy "VapourSynth filter bridge (Lua)"
+    opt_yes_no _libarchive  "libarchive"
     opt_yes_no _encoding    "encoding functionality" yes
     opt_yes_no _build_man   "building manpage"
 }
@@ -376,7 +385,7 @@ addcflags() { cflag_check "$@" && OURCFLAGS="$OURCFLAGS $@" ; }
 OURCFLAGS="-std=c99 -Wall $_opt"
 
 addcflags -g -g3 -ggdb
-addcflags -Wundef -Wmissing-prototypes -Wshadow -Wno-switch -Wparentheses -Wpointer-arith -Wredundant-decls -Wno-pointer-sign -Werror=implicit-function-declaration -Wno-error=deprecated-declarations -Wno-error=unused-function
+addcflags -Wundef -Wmissing-prototypes -Wshadow -Wno-switch -Wparentheses -Wpointer-arith -Wno-redundant-decls -Wno-pointer-sign -Werror=implicit-function-declaration -Wno-error=deprecated-declarations -Wno-error=unused-function
 # clang
 addcflags -Wno-logical-op-parentheses -fcolor-diagnostics -Wno-tautological-compare -Wno-tautological-constant-out-of-range-compare
 # extra
@@ -506,9 +515,12 @@ check_statement_libs "compiler support for __sync built-ins" $_sync SYNC_BUILTIN
     stdint.h 'int64_t test = 0; test = __sync_add_and_fetch(&test, 1)'
 _sync=$(defretval)
 
+_any_atomic=yes
 if test "$_atomic" = no && test "$_sync" = no && test "$_stdatomic" = no ; then
   echo "your compiler must support either stdatomic.h, or __atomic, or __sync built-ins."
+  _any_atomic=no
 fi
+define_yes_no $_any_atomic HAVE_ATOMICS
 
 check_compile "iconv" $_iconv ICONV waftools/fragments/iconv.c " " "-liconv" "-liconv $_ld_dl"
 _iconv=$(defretval)
@@ -709,7 +721,7 @@ check_pkg_config "OpenAL" $_openal OPENAL 'openal >= 1.13'
 
 check_pkg_config "ALSA audio" $_alsa ALSA 'alsa >= 1.0.9'
 
-check_pkg_config "Blu-ray support" $_bluray LIBBLURAY 'libbluray >= 0.2.1'
+check_pkg_config "Blu-ray support" $_bluray LIBBLURAY 'libbluray >= 0.3.0'
 
 check_pkg_config "dvdread" $_dvdread DVDREAD 'dvdread >= 4.1.0'
 
@@ -732,6 +744,7 @@ echo "LIBASS_OSD = $_libass" >> $CONFIG_MAK
 echo "DUMMY_OSD = $_dummy_osd" >> $CONFIG_MAK
 
 check_pkg_config "ENCA" $_enca ENCA 'enca'
+check_pkg_config "uchardet" $_uchardet UCHARDET 'uchardet'
 
 check_pkg_config "zlib" auto ZLIB 'zlib'
 test $(defretval) = no && die "Unable to find development files for zlib."
@@ -786,6 +799,12 @@ api_statement_check \
     libavutil/frame.h \
     'enum AVFrameSideDataType type = AV_FRAME_DATA_SKIP_SAMPLES'
 
+api_statement_check \
+    "libavutil av_version_info()" \
+    HAVE_AV_VERSION_INFO \
+    libavutil/avutil.h \
+    'const char *x = av_version_info()'
+
 check_pkg_config "libavfilter" $_libavfilter LIBAVFILTER 'libavfilter >= 5.0.0'
 
 check_pkg_config "libavdevice" $_libavdevice LIBAVDEVICE 'libavdevice >= 55.0.0'
@@ -817,9 +836,9 @@ test_lua "lua >= 5.1.0 lua < 5.2.0"
 test_lua "lua5.1 >= 5.1.0" # debian
 test_lua "lua51 >= 5.1.0" # OpenBSD
 test_lua "luajit >= 2.0.0"
-test_lua "lua >= 5.2.0"
 test_lua "lua5.2 >= 5.2.0" # debian
 test_lua "lua52 >= 5.2.0" # OpenBSD
+test_lua "lua >= 5.2.0"
 
 test "$_lua" != yes && check_yes_no no LUA
 
@@ -841,6 +860,8 @@ if test "$_vapoursynth" = no && test "$_vapoursynth_lazy" = no ; then
 fi
 check_trivial "VapourSynth core" $_vapoursynth_core VAPOURSYNTH_CORE
 
+check_pkg_config "libarchive support" $_libarchive LIBARCHIVE 'libarchive >= 3.0.0'
+
 check_trivial "encoding" $_encoding ENCODING
 
 # needs dlopen on unix
@@ -858,7 +879,7 @@ CFLAGS="$CFLAGS -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -D_LARGEFILE64_SOURCE
 # in-tree builds.
 
 if test ! -f Makefile ; then
-    ln -s old-makefile Makefile
+    ln -s TOOLS/old-makefile Makefile
 fi
 
 cat > old_build/config.mak << EOF
@@ -905,7 +926,7 @@ cat > $TMPC << EOF
 /* we didn't bother to add actual config checks for this, or they are
    for platforms not supported by this configure script */
 #define HAVE_BSD_FSTATFS 0
-#define HAVE_LINUX_FSTATFS 0
+#define HAVE_LINUX_FSTATFS 1
 #define HAVE_VDA_AV_VDA_ALLOC_CONTEXT 0
 #define HAVE_VDA_HWACCEL 0
 #define HAVE_VDA_LIBAVCODEC_REFCOUNTING 0
@@ -936,6 +957,9 @@ cat > $TMPC << EOF
 #define HAVE_RPI_GLES 0
 #define HAVE_AV_PIX_FMT_MMAL 0
 #define HAVE_DRM 0
+#define HAVE_VIDEOTOOLBOX_HWACCEL 0
+#define HAVE_VIDEOTOOLBOX_GL 0
+#define HAVE_VIDEOTOOLBOX_VDA_GL 0
 
 #ifdef __OpenBSD__
 #define DEFAULT_CDROM_DEVICE "/dev/rcd0c"

TOOLS/old-makefile

@@ -80,7 +80,8 @@ SOURCES-$(RSOUND)               += audio/out/ao_rsound.c
 SOURCES-$(SNDIO)                += audio/out/ao_sndio.c
 SOURCES-$(VDPAU)                += video/vdpau.c video/vdpau_mixer.c \
                                    video/out/vo_vdpau.c video/decode/vdpau.c \
-                                   video/filter/vf_vdpaupp.c
+                                   video/filter/vf_vdpaupp.c \
+                                   video/filter/vf_vdpaurb.c
 SOURCES-$(VDPAU_GL_X11)         += video/out/gl_hwdec_vdpau.c
 SOURCES-$(VAAPI)                += video/out/vo_vaapi.c \
                                    video/decode/vaapi.c \
@@ -88,8 +89,7 @@ SOURCES-$(VAAPI)                += video/out/vo_vaapi.c \
 SOURCES-$(VAAPI_VPP)            += video/filter/vf_vavpp.c
 SOURCES-$(VAAPI_GLX)            += video/out/gl_hwdec_vaglx.c
 
-SOURCES-$(X11)                  += video/out/vo_x11.c video/out/x11_common.c
-SOURCES-$(XV)                   += video/out/vo_xv.c
+SOURCES-$(XV)                   += video/out/x11_common.c video/out/vo_xv.c
 SOURCES-$(WAYLAND)              += video/out/vo_wayland.c \
                                    video/out/wayland_common.c \
                                    video/out/wayland/buffer.c \
@@ -108,6 +108,8 @@ SOURCES-$(LIBAVFILTER)          += video/filter/vf_lavfi.c \
 
 SOURCES-$(LUA)                  += player/lua.c
 SOURCES-$(VAPOURSYNTH_CORE)     += video/filter/vf_vapoursynth.c
+SOURCES-$(LIBARCHIVE)           += demux/demux_libarchive.c \
+                                   stream/stream_libarchive.c
 SOURCES-$(DLOPEN)               += video/filter/vf_dlopen.c
 
 SOURCES = audio/audio.c \
@@ -123,10 +125,7 @@ SOURCES = audio/audio.c \
           audio/filter/af.c \
           audio/filter/af_center.c \
           audio/filter/af_channels.c \
-          audio/filter/af_convert24.c \
-          audio/filter/af_convertsignendian.c \
           audio/filter/af_delay.c \
-          audio/filter/af_dummy.c \
           audio/filter/af_equalizer.c \
           audio/filter/af_export.c \
           audio/filter/af_extrastereo.c \
@@ -160,6 +159,7 @@ SOURCES = audio/audio.c \
           common/tags.c \
           common/version.c \
           demux/codec_tags.c \
+          demux/cue.c \
           demux/demux.c \
           demux/demux_edl.c \
           demux/demux_cue.c \
@@ -195,6 +195,7 @@ SOURCES = audio/audio.c \
           options/parse_configfile.c \
           options/path.c \
           osdep/io.c \
+          osdep/path-unix.c \
           osdep/semaphore_osx.c \
           osdep/subprocess.c \
           osdep/subprocess-posix.c \
@@ -206,7 +207,6 @@ SOURCES = audio/audio.c \
           player/client.c \
           player/configfiles.c \
           player/command.c \
-          player/discnav.c \
           player/loadfile.c \
           player/main.c \
           player/misc.c \
@@ -334,7 +334,7 @@ all: $(ALL_TARGETS)
 %.o: %.c
 	$(CC) $(DEPFLAGS) $(CFLAGS) -c -o $@ $<
 
-mpv: $(OBJECTS) player/main-fn-unix.o
+mpv: $(OBJECTS) osdep/main-fn-unix.o
 	$(CC) -o $@ $^ $(EXTRALIBS)
 
 input/input.c: input/input_conf.h
@@ -381,8 +381,8 @@ config.mak: configure
 	@echo "####### Please run ./configure again - it's changed! #######"
 	@echo "############################################################"
 
-version.h .version: version.sh
-	./$<
+old_build/version.h .version: version.sh
+	./version.sh --versionh=old_build/version.h
 
 # Force version.sh to run to potentially regenerate version.h
 -include .version
@@ -393,7 +393,7 @@ version.h .version: version.sh
 
 ###### dependency declarations / specific CFLAGS ######
 
-common/version.c: version.h
+common/version.c: old_build/version.h
 
 DOCS/man/mpv.1: DOCS/man/af.rst \
                    DOCS/man/ao.rst \
@@ -464,7 +464,7 @@ clean:
 	-$(RM) $(call ADD_ALL_DIRS,/*.o /*.a /*.ho /*~)
 	-$(RM) mpv
 	-$(RM) DOCS/man/*/mpv.1
-	-$(RM) version.h
+	-$(RM) old_build/version.h version.h
 	-$(RM) input/input_conf.h
 	-$(RM) video/out/vdpau_template.c
 	-$(RM) demux/ebml_types.h demux/ebml_defs.c

TOOLS/osxbundle.py

@@ -40,7 +40,7 @@ def apply_plist_template(plist_file, version):
         print (line.rstrip().replace('${VERSION}', version))
 
 def main():
-    version = sh("./version.sh --print").strip()
+    version = sh("./version.sh").strip()
 
     usage = "usage: %prog [options] arg"
     parser = OptionParser(usage)

TOOLS/osxbundle/mpv.app/Contents/Resources/mpv.conf

@@ -1,2 +1 @@
-screenshot-template=~/Desktop/shot%n
 profile=pseudo-gui

TOOLS/stats-conv.py

@@ -5,7 +5,10 @@ import re
 
 filename = sys.argv[1]
 
-event_regex = re.compile(".*")
+events = ".*"
+if len(sys.argv) > 2:
+    events = sys.argv[2]
+event_regex = re.compile(events)
 
 """
 This script is meant to display stats written by mpv --dump-stats=filename.
@@ -26,8 +29,8 @@ Currently, the following event types are supported:
     '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
-    'value-timed' <ts> <float> <name>
-                                a value for an event at the given timestamp
+    'value-timed' <ts> <float> <name>       a value for an event at the given timestamp
+    'range-timed' <ts1> <ts2> <name>        like start/end, but explicit times
     <name>                      singular event (same as 'signal')
 
 """
@@ -91,6 +94,15 @@ for line in [line.split("#")[0].strip() for line in open(filename, "r")]:
         val = int(val) / SCALE - G.start
         e = get_event(name, "event-signal")
         e.vals.append((val, 1))
+    elif event.startswith("range-timed "):
+        _, ts1, ts2, name = event.split(" ", 3)
+        ts1 = int(ts1) / SCALE - G.start
+        ts2 = int(ts2) / SCALE - G.start
+        e = get_event(name, "event")
+        e.vals.append((ts1, 0))
+        e.vals.append((ts1, 1))
+        e.vals.append((ts2, 1))
+        e.vals.append((ts2, 0))
     elif event.startswith("value-timed "):
         _, tsval, val, name = event.split(" ", 3)
         tsval = int(tsval) / SCALE - G.start

TOOLS/zsh.pl

@@ -141,7 +141,7 @@ sub parse_main_opts {
     my ($cmd, $regex) = @_;
 
     my @list;
-    my @lines = split /\n/, `"$mpv" --no-config $cmd`;
+    my @lines = call_mpv($cmd);
 
     foreach my $line (@lines) {
         my ($name, $desc) = ($line =~ /^$regex/) or next;
@@ -206,7 +206,7 @@ sub parse_opts {
     my ($cmd, $regex) = @_;
 
     my @list;
-    my @lines = split /\n/, `"$mpv" --no-config $cmd`;
+    my @lines = call_mpv($cmd);
 
     foreach my $line (@lines) {
         if ($line !~ /^$regex/) {
@@ -226,3 +226,14 @@ sub parse_opts {
 
     return @list;
 }
+
+sub call_mpv {
+    my ($cmd) = @_;
+    my $output = `"$mpv" --no-config $cmd`;
+    if ($? == -1) {
+        die "Could not run mpv: $!";
+    } elsif ($? != 0) {
+        die "mpv returned " . ($? >> 8) . " with output:\n$output";
+    }
+    return split /\n/, $output;
+}

VERSION

@@ -0,0 +1 @@
+0.10.0

audio/audio.c

@@ -34,8 +34,8 @@ static void update_redundant_info(struct mp_audio *mpa)
     assert(mp_chmap_is_empty(&mpa->channels) ||
            mp_chmap_is_valid(&mpa->channels));
     mpa->nch = mpa->channels.num;
-    mpa->bps = af_fmt2bps(mpa->format);
-    if (AF_FORMAT_IS_PLANAR(mpa->format)) {
+    mpa->bps = af_fmt_to_bytes(mpa->format);
+    if (af_fmt_is_planar(mpa->format)) {
         mpa->spf = 1;
         mpa->num_planes = mpa->nch;
         mpa->sstride = mpa->bps;
@@ -93,15 +93,19 @@ bool mp_audio_config_valid(const struct mp_audio *mpa)
 
 char *mp_audio_config_to_str_buf(char *buf, size_t buf_sz, struct mp_audio *mpa)
 {
+    char ch[128];
+    mp_chmap_to_str_buf(ch, sizeof(ch), &mpa->channels);
+    char *hr_ch = mp_chmap_to_str_hr(&mpa->channels);
+    if (strcmp(hr_ch, ch) != 0)
+        mp_snprintf_cat(ch, sizeof(ch), " (%s)", hr_ch);
     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));
+             ch, mpa->channels.num, af_fmt_to_str(mpa->format));
     return buf;
 }
 
 void mp_audio_force_interleaved_format(struct mp_audio *mpa)
 {
-    if (AF_FORMAT_IS_PLANAR(mpa->format))
+    if (af_fmt_is_planar(mpa->format))
         mp_audio_set_format(mpa, af_fmt_from_planar(mpa->format));
 }
 
@@ -159,8 +163,8 @@ void mp_audio_realloc(struct mp_audio *mpa, int samples)
         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;
         }
+        mpa->planes[n] = mpa->allocated[n]->data;
     }
     for (int n = mpa->num_planes; n < MP_NUM_CHANNELS; n++) {
         av_buffer_unref(&mpa->allocated[n]);
@@ -191,8 +195,18 @@ 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->allocated[n]->size / mpa->sstride : 0;
-        size = n == 0 ? s : MPMIN(size, s);
+        for (int i = 0; i < MP_NUM_CHANNELS && mpa->allocated[i]; i++) {
+            uint8_t *start = mpa->allocated[i]->data;
+            uint8_t *end = start + mpa->allocated[i]->size;
+            uint8_t *plane = mpa->planes[n];
+            if (plane >= start && plane < end) {
+                int s = MPMIN((end - plane) / mpa->sstride, INT_MAX);
+                size = n == 0 ? s : MPMIN(size, s);
+                goto next;
+            }
+        }
+        return 0; // plane is not covered by any buffer
+    next: ;
     }
     return size;
 }

audio/audio.h

@@ -35,9 +35,12 @@ struct mp_audio {
     int nch;            // number of channels (redundant with chmap)
     int spf;            // sub-samples per sample on each plane
     int num_planes;     // number of planes
-    int bps;            // size of sub-samples (af_fmt2bps(format))
+    int bps;            // size of sub-samples (af_fmt_to_bytes(format))
 
-    // private
+    // --- private
+    // These do not necessarily map directly to planes[]. They can have
+    // different order or count. There shouldn't be more buffers than planes.
+    // If allocated[n] is NULL, allocated[n+1] must also be NULL.
     struct AVBufferRef *allocated[MP_NUM_CHANNELS];
 };
 

audio/chmap.c

@@ -18,6 +18,8 @@
 #include <stdlib.h>
 #include <assert.h>
 
+#include <libavutil/common.h>
+
 #include "common/common.h"
 #include "common/msg.h"
 #include "chmap.h"
@@ -50,6 +52,7 @@ static const char *const speaker_names[MP_SPEAKER_ID_COUNT][2] = {
     [MP_SPEAKER_ID_SDL]         = {"sdl",  "surround direct left"},
     [MP_SPEAKER_ID_SDR]         = {"sdr",  "surround direct right"},
     [MP_SPEAKER_ID_LFE2]        = {"lfe2", "low frequency 2"},
+    [MP_SPEAKER_ID_NA]          = {"na",   "not available"},
 };
 
 // Names taken from libavutil/channel_layout.c (Not accessible by API.)
@@ -67,6 +70,7 @@ static const char *const std_layout_names[][2] = {
     {"quad",            "fl-fr-bl-br"},
     {"quad(side)",      "fl-fr-sl-sr"},
     {"3.1",             "fl-fr-fc-lfe"},
+    {"3.1(back)",       "fl-fr-lfe-bc"}, // not in lavc
     {"5.0",             "fl-fr-fc-bl-br"},
     {"5.0(alsa)",       "fl-fr-bl-br-fc"}, // not in lavc
     {"5.0(side)",       "fl-fr-fc-sl-sr"},
@@ -80,9 +84,11 @@ static const char *const std_layout_names[][2] = {
     {"hexagonal",       "fl-fr-fc-bl-br-bc"},
     {"6.1",             "fl-fr-fc-lfe-bc-sl-sr"},
     {"6.1(back)",       "fl-fr-fc-lfe-bl-br-bc"}, // lavc calls this "6.1" too
+    {"6.1(top)",        "fl-fr-fc-lfe-bl-br-tc"}, // not in lavc
     {"6.1(front)",      "fl-fr-lfe-flc-frc-sl-sr"},
     {"7.0",             "fl-fr-fc-bl-br-sl-sr"},
     {"7.0(front)",      "fl-fr-fc-flc-frc-sl-sr"},
+    {"7.0(rear)",       "fl-fr-fc-bl-br-sdl-sdr"}, // not in lavc
     {"7.1",             "fl-fr-fc-lfe-bl-br-sl-sr"},
     {"7.1(alsa)",       "fl-fr-bl-br-fc-lfe-sl-sr"}, // not in lavc
     {"7.1(wide)",       "fl-fr-fc-lfe-bl-br-flc-frc"},
@@ -94,7 +100,7 @@ static const char *const std_layout_names[][2] = {
     {0}
 };
 
-static const struct mp_chmap default_layouts[MP_NUM_CHANNELS + 1] = {
+static const struct mp_chmap default_layouts[] = {
     {0},                                        // empty
     MP_CHMAP_INIT_MONO,                         // mono
     MP_CHMAP2(FL, FR),                          // stereo
@@ -127,7 +133,8 @@ bool mp_chmap_is_valid(const struct mp_chmap *src)
         int sp = src->speaker[n];
         if (sp >= MP_SPEAKER_ID_COUNT || mapped[sp])
             return false;
-        mapped[sp] = true;
+        if (sp != MP_SPEAKER_ID_NA)
+            mapped[sp] = true;
     }
     return src->num > 0;
 }
@@ -142,12 +149,10 @@ bool mp_chmap_is_empty(const struct mp_chmap *src)
 bool mp_chmap_is_unknown(const struct mp_chmap *src)
 {
     for (int n = 0; n < src->num; n++) {
-        int speaker = src->speaker[n];
-        if (speaker >= MP_SPEAKER_ID_UNKNOWN0 &&
-            speaker <= MP_SPEAKER_ID_UNKNOWN_LAST)
-            return true;
+        if (src->speaker[n] != MP_SPEAKER_ID_NA)
+            return false;
     }
-    return false;
+    return mp_chmap_is_valid(src);
 }
 
 // Note: empty channel maps compare as equal. Invalid ones can equal too.
@@ -171,15 +176,6 @@ bool mp_chmap_equals_reordered(const struct mp_chmap *a, const struct mp_chmap *
     return mp_chmap_equals(&t1, &t2);
 }
 
-bool mp_chmap_is_compatible(const struct mp_chmap *a, const struct mp_chmap *b)
-{
-    if (mp_chmap_equals(a, b))
-        return true;
-    if (a->num == b->num && (mp_chmap_is_unknown(a) || mp_chmap_is_unknown(b)))
-        return true;
-    return false;
-}
-
 bool mp_chmap_is_stereo(const struct mp_chmap *src)
 {
     static const struct mp_chmap stereo = MP_CHMAP_INIT_STEREO;
@@ -199,16 +195,36 @@ void mp_chmap_reorder_norm(struct mp_chmap *map)
     qsort(arr, map->num, 1, comp_uint8);
 }
 
+// Remove silent (NA) channels, if any.
+void mp_chmap_remove_na(struct mp_chmap *map)
+{
+    struct mp_chmap new = {0};
+    for (int n = 0; n < map->num; n++) {
+        int sp = map->speaker[n];
+        if (sp != MP_SPEAKER_ID_NA)
+            new.speaker[new.num++] = map->speaker[n];
+    }
+    *map = new;
+}
+
+// Add silent (NA) channels to map until map->num >= num.
+void mp_chmap_fill_na(struct mp_chmap *map, int num)
+{
+    assert(num <= MP_NUM_CHANNELS);
+    while (map->num < num)
+        map->speaker[map->num++] = MP_SPEAKER_ID_NA;
+}
+
 // Set *dst to a standard layout with the given number of channels.
 // If the number of channels is invalid, an invalid map is set, and
 // mp_chmap_is_valid(dst) will return false.
 void mp_chmap_from_channels(struct mp_chmap *dst, int num_channels)
 {
-    if (num_channels < 0 || num_channels > MP_NUM_CHANNELS) {
-        *dst = (struct mp_chmap) {0};
-    } else {
+    *dst = (struct mp_chmap) {0};
+    if (num_channels >= 0 && num_channels < MP_ARRAY_SIZE(default_layouts))
         *dst = default_layouts[num_channels];
-    }
+    if (!dst->num)
+        mp_chmap_set_unknown(dst, num_channels);
 }
 
 // Try to do what mplayer/mplayer2/mpv did before channel layouts were
@@ -227,6 +243,8 @@ void mp_chmap_from_channels_alsa(struct mp_chmap *dst, int num_channels)
 // Set *dst to an unknown layout for the given numbers of channels.
 // If the number of channels is invalid, an invalid map is set, and
 // mp_chmap_is_valid(dst) will return false.
+// A mp_chmap with all entries set to NA is treated specially in some
+// contexts (watch out for mp_chmap_is_unknown()).
 void mp_chmap_set_unknown(struct mp_chmap *dst, int num_channels)
 {
     if (num_channels < 0 || num_channels > MP_NUM_CHANNELS) {
@@ -234,7 +252,7 @@ void mp_chmap_set_unknown(struct mp_chmap *dst, int num_channels)
     } else {
         dst->num = num_channels;
         for (int n = 0; n < dst->num; n++)
-            dst->speaker[n] = MP_SPEAKER_ID_UNKNOWN0 + n;
+            dst->speaker[n] = MP_SPEAKER_ID_NA;
     }
 }
 
@@ -277,6 +295,7 @@ void mp_chmap_remove_useless_channels(struct mp_chmap *map,
 }
 
 // Return the ffmpeg/libav channel layout as in <libavutil/channel_layout.h>.
+// Speakers not representable by ffmpeg/libav are dropped.
 // Warning: this ignores the order of the channels, and will return a channel
 //          mask even if the order is different from libavcodec's.
 uint64_t mp_chmap_to_lavc_unchecked(const struct mp_chmap *src)
@@ -286,8 +305,10 @@ uint64_t mp_chmap_to_lavc_unchecked(const struct mp_chmap *src)
     if (mp_chmap_is_unknown(&t))
         mp_chmap_from_channels(&t, t.num);
     uint64_t mask = 0;
-    for (int n = 0; n < t.num; n++)
-        mask |= 1ULL << t.speaker[n];
+    for (int n = 0; n < t.num; n++) {
+        if (t.speaker[n] < 64) // ignore MP_SPEAKER_ID_NA etc.
+            mask |= 1ULL << t.speaker[n];
+    }
     return mask;
 }
 
@@ -348,56 +369,43 @@ void mp_chmap_reorder_to_lavc(struct mp_chmap *map)
 
 // Get reordering array for from->to reordering. from->to must have the same set
 // of speakers (i.e. same number and speaker IDs, just different order). Then,
-// for each speaker n, dst[n] will be set such that:
-//      to->speaker[dst[n]] = from->speaker[n]
-// (dst[n] gives the source channel for destination channel n)
-void mp_chmap_get_reorder(int dst[MP_NUM_CHANNELS], const struct mp_chmap *from,
+// for each speaker n, src[n] will be set such that:
+//      to->speaker[n] = from->speaker[src[n]]
+// (src[n] gives the source channel for destination channel n)
+// If *from and *to don't contain the same set of speakers, then the above
+// invariant is not guaranteed. Instead, src[n] can be set to -1 if the channel
+// at to->speaker[n] is unmapped.
+void mp_chmap_get_reorder(int src[MP_NUM_CHANNELS], const struct mp_chmap *from,
                           const struct mp_chmap *to)
 {
-    assert(from->num == to->num);
+    for (int n = 0; n < MP_NUM_CHANNELS; n++)
+        src[n] = -1;
+
     if (mp_chmap_is_unknown(from) || mp_chmap_is_unknown(to)) {
-        for (int n = 0; n < from->num; n++)
-            dst[n] = n;
+        for (int n = 0; n < to->num; n++)
+            src[n] = n < from->num ? n : -1;
         return;
     }
-    // Same set of speakers required
-    assert(mp_chmap_equals_reordered(from, to));
-    for (int n = 0; n < from->num; n++) {
-        int src = from->speaker[n];
-        dst[n] = -1;
-        for (int i = 0; i < to->num; i++) {
-            if (src == to->speaker[i]) {
-                dst[n] = i;
+
+    for (int n = 0; n < to->num; n++) {
+        for (int i = 0; i < from->num; i++) {
+            if (to->speaker[n] == from->speaker[i]) {
+                src[n] = i;
                 break;
             }
         }
-        assert(dst[n] != -1);
     }
-    for (int n = 0; n < from->num; n++)
-        assert(to->speaker[dst[n]] == from->speaker[n]);
+
+    for (int n = 0; n < to->num; n++)
+        assert(src[n] < 0 || (to->speaker[n] == from->speaker[src[n]]));
 }
 
-// Performs the difference between a and b, and store it in diff. If b has
-// channels that do not appear in a, those will not appear in the difference.
-// To get to those the argument ordering in the function call has to be
-// inverted. For the same reason, the diff with a superset will return no
-// speakers.
-void mp_chmap_diff(const struct mp_chmap *a, const struct mp_chmap *b,
-                   struct mp_chmap *diff)
+// Return the number of channels only in a.
+int mp_chmap_diffn(const struct mp_chmap *a, const struct mp_chmap *b)
 {
     uint64_t a_mask = mp_chmap_to_lavc_unchecked(a);
     uint64_t b_mask = mp_chmap_to_lavc_unchecked(b);
-    mp_chmap_from_lavc(diff, (a_mask ^ b_mask) & a_mask);
-}
-
-// Checks whether a contains all the speakers in b
-bool mp_chmap_contains(const struct mp_chmap *a, const struct mp_chmap *b)
-{
-    struct mp_chmap d1;
-    struct mp_chmap d2;
-    mp_chmap_diff(a, b, &d1);
-    mp_chmap_diff(b, a, &d2);
-    return a->num >= b->num && d1.num >= 0 && d2.num == 0;
+    return av_popcount64((a_mask ^ b_mask) & a_mask);
 }
 
 // Returns something like "fl-fr-fc". If there's a standard layout in lavc
@@ -500,6 +508,25 @@ bool mp_chmap_from_str(struct mp_chmap *dst, bstr src)
     return true;
 }
 
+// Output a human readable "canonical" channel map string. Converting this from
+// a string back to a channel map can yield a different map, but the string
+// looks nicer. E.g. "fc-fl-fr-na" becomes "3.0".
+char *mp_chmap_to_str_hr_buf(char *buf, size_t buf_size, const struct mp_chmap *src)
+{
+    struct mp_chmap map = *src;
+    mp_chmap_remove_na(&map);
+    for (int n = 0; std_layout_names[n][0]; n++) {
+        struct mp_chmap s;
+        if (mp_chmap_from_str(&s, bstr0(std_layout_names[n][0])) &&
+            mp_chmap_equals_reordered(&s, &map))
+        {
+            map = s;
+            break;
+        }
+    }
+    return mp_chmap_to_str_buf(buf, buf_size, &map);
+}
+
 void mp_chmap_print_help(struct mp_log *log)
 {
     mp_info(log, "Speakers:\n");

audio/chmap.h

@@ -47,7 +47,7 @@ enum mp_speaker_id {
     MP_SPEAKER_ID_TBL,          // TOP_BACK_LEFT
     MP_SPEAKER_ID_TBC,          // TOP_BACK_CENTER
     MP_SPEAKER_ID_TBR,          // TOP_BACK_RIGHT
-     // Inofficial/libav* extensions
+     // Unofficial/libav* extensions
     MP_SPEAKER_ID_DL = 29,      // STEREO_LEFT (stereo downmix special speakers)
     MP_SPEAKER_ID_DR,           // STEREO_RIGHT
     MP_SPEAKER_ID_WL,           // WIDE_LEFT
@@ -56,10 +56,12 @@ enum mp_speaker_id {
     MP_SPEAKER_ID_SDR,          // SURROUND_DIRECT_RIGHT
     MP_SPEAKER_ID_LFE2,         // LOW_FREQUENCY_2
 
-    // Special mpv-specific speaker entries reserved for channels which have no
-    // known meaning.
-    MP_SPEAKER_ID_UNKNOWN0 = 64,
-    MP_SPEAKER_ID_UNKNOWN_LAST = MP_SPEAKER_ID_UNKNOWN0 + MP_NUM_CHANNELS - 1,
+    // Speaker IDs >= 64 are not representable in WAVEFORMATEXTENSIBLE or libav*.
+
+    // "Silent" channels. These are sometimes used to insert padding for
+    // unused channels. Unlike other speaker types, multiple of these can
+    // occur in a single mp_chmap.
+    MP_SPEAKER_ID_NA = 64,
 
     // Including the unassigned IDs in between. This is not a valid ID anymore,
     // but is still within uint8_t.
@@ -98,10 +100,11 @@ bool mp_chmap_is_empty(const struct mp_chmap *src);
 bool mp_chmap_is_unknown(const struct mp_chmap *src);
 bool mp_chmap_equals(const struct mp_chmap *a, const struct mp_chmap *b);
 bool mp_chmap_equals_reordered(const struct mp_chmap *a, const struct mp_chmap *b);
-bool mp_chmap_is_compatible(const struct mp_chmap *a, const struct mp_chmap *b);
 bool mp_chmap_is_stereo(const struct mp_chmap *src);
 
 void mp_chmap_reorder_norm(struct mp_chmap *map);
+void mp_chmap_remove_na(struct mp_chmap *map);
+void mp_chmap_fill_na(struct mp_chmap *map, int num);
 
 void mp_chmap_from_channels(struct mp_chmap *dst, int num_channels);
 void mp_chmap_set_unknown(struct mp_chmap *dst, int num_channels);
@@ -117,17 +120,17 @@ void mp_chmap_from_lavc(struct mp_chmap *dst, uint64_t src);
 bool mp_chmap_is_lavc(const struct mp_chmap *src);
 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,
+void mp_chmap_get_reorder(int src[MP_NUM_CHANNELS], const struct mp_chmap *from,
                           const struct mp_chmap *to);
 
-void mp_chmap_diff(const struct mp_chmap *a, const struct mp_chmap *b,
-                   struct mp_chmap *diff);
-
-bool mp_chmap_contains(const struct mp_chmap *a, const struct mp_chmap *b);
+int mp_chmap_diffn(const struct mp_chmap *a, const struct mp_chmap *b);
 
 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))
 
+char *mp_chmap_to_str_hr_buf(char *buf, size_t buf_size, const struct mp_chmap *src);
+#define mp_chmap_to_str_hr(m) mp_chmap_to_str_hr_buf((char[128]){0}, 128, (m))
+
 bool mp_chmap_from_str(struct mp_chmap *dst, bstr src);
 
 struct mp_log;

audio/chmap_sel.c

@@ -58,6 +58,29 @@ static bool replace_speakers(struct mp_chmap *map, struct mp_chmap list[2])
     return false;
 }
 
+// These go strictly from the first to the second entry and always use the
+// full layout (possibly reordered and/or padding channels added).
+static const struct mp_chmap preferred_remix[][2] = {
+    // mono can be perfectly played as stereo
+    { MP_CHMAP_INIT_MONO, MP_CHMAP_INIT_STEREO },
+};
+
+// Conversion from src to dst is explicitly encouraged and should be preferred
+// over "mathematical" upmixes or downmixes (which minimize lost channels).
+static bool test_preferred_remix(const struct mp_chmap *src,
+                                 const struct mp_chmap *dst)
+{
+    struct mp_chmap src_p = *src, dst_p = *dst;
+    mp_chmap_remove_na(&src_p);
+    mp_chmap_remove_na(&dst_p);
+    for (int n = 0; n < MP_ARRAY_SIZE(preferred_remix); n++) {
+        if (mp_chmap_equals_reordered(&src_p, &preferred_remix[n][0]) &&
+            mp_chmap_equals_reordered(&dst_p, &preferred_remix[n][1]))
+            return true;
+    }
+    return false;
+}
+
 // Allow all channel layouts that can be expressed with mp_chmap.
 // (By default, all layouts are rejected.)
 void mp_chmap_sel_add_any(struct mp_chmap_sel *s)
@@ -176,6 +199,10 @@ bool mp_chmap_sel_adjust(const struct mp_chmap_sel *s, struct mp_chmap *map)
             return true;
         }
     }
+
+    if (mp_chmap_sel_fallback(s, map))
+        return true;
+
     for (int i = 0; i < MP_ARRAY_SIZE(speaker_replacements); i++) {
         struct mp_chmap  t = *map;
         struct mp_chmap *r = (struct mp_chmap *)speaker_replacements[i];
@@ -185,9 +212,6 @@ bool mp_chmap_sel_adjust(const struct mp_chmap_sel *s, struct mp_chmap *map)
         }
     }
 
-    if (mp_chmap_sel_fallback(s, map))
-        return true;
-
     // Fallback to mono/stereo as last resort
     *map = (struct mp_chmap) MP_CHMAP_INIT_STEREO;
     if (test_layout(s, map))
@@ -199,64 +223,95 @@ bool mp_chmap_sel_adjust(const struct mp_chmap_sel *s, struct mp_chmap *map)
     return false;
 }
 
-#define UPMIX_IDX 0
-#define DOWNMIX_IDX 1
-#define FALLBACK_IDX 2
-
-static bool test_fallbacks(struct mp_chmap *a, struct mp_chmap *b,
-                           int best_diffs[2], struct mp_chmap best[2])
+// Like mp_chmap_diffn(), but find the minimum difference with all possible
+// speaker replacements considered.
+static int mp_chmap_diffn_r(const struct mp_chmap *a, const struct mp_chmap *b)
 {
-    struct mp_chmap diff1, diff2;
+    int mindiff = INT_MAX;
 
-    mp_chmap_diff(a, b, &diff1);
-    if (mp_chmap_contains(a, b) && best_diffs[UPMIX_IDX] > diff1.num) {
-        best[UPMIX_IDX] = *a;
-        best_diffs[UPMIX_IDX] = diff1.num;
-        return true;
+    for (int i = -1; i < (int)MP_ARRAY_SIZE(speaker_replacements); i++) {
+        struct mp_chmap ar = *a;
+        if (i >= 0) {
+            struct mp_chmap *r = (struct mp_chmap *)speaker_replacements[i];
+            if (!replace_speakers(&ar, r))
+                continue;
+        }
+        int d = mp_chmap_diffn(&ar, b);
+        if (d < mindiff)
+            mindiff = d;
     }
 
-    mp_chmap_diff(b, a, &diff2);
-    if (mp_chmap_contains(b, a) && best_diffs[DOWNMIX_IDX] > diff2.num) {
-        best[DOWNMIX_IDX] = *a;
-        best_diffs[DOWNMIX_IDX] = diff2.num;
-        return true;
-    }
-
-    if (diff1.num > 0 && best_diffs[FALLBACK_IDX] > diff1.num) {
-        best[FALLBACK_IDX] = *a;
-        best_diffs[FALLBACK_IDX] = diff1.num;
-        return true;
-    }
-
-    if (diff2.num > 0 && best_diffs[FALLBACK_IDX] > diff2.num) {
-        best[FALLBACK_IDX] = *a;
-        best_diffs[FALLBACK_IDX] = diff2.num;
-        return true;
-    }
-
-    return false;
-}
-
-// Determine which channel map to fallback to given a source channel map. It
-// uses the following heuristic:
-// 1) If mono is required always prefer stereo to a multichannel upmix.
-// 2) Search for an upmix that is an exact superset of the required chmap.
-// 3) Search for a downmix that is the exact subset of the required chmap.
-// 4) Search for either an upmix or downmix that is the closest (minimum
-//    difference of speakers) to the required chmap.
-bool mp_chmap_sel_fallback(const struct mp_chmap_sel *s, struct mp_chmap *map)
-{
-    // special case: if possible always fallback mono to stereo (instead of
-    // looking for a multichannel upmix)
+    // Special-case: we consider stereo a replacement for mono. (This is not
+    // true in the other direction. Also, fl-fr is generally not a replacement
+    // for fc. Thus it's not part of the speaker replacement list.)
     struct mp_chmap mono   = MP_CHMAP_INIT_MONO;
     struct mp_chmap stereo = MP_CHMAP_INIT_STEREO;
-    if (mp_chmap_equals(&mono, map) && test_layout(s, &stereo)) {
-        *map = stereo;
-        return true;
-    }
+    if (mp_chmap_equals(&mono, b) && mp_chmap_equals(&stereo, a))
+        mindiff = 0;
 
-    int best_diffs[] = { INT_MAX, INT_MAX, INT_MAX };
-    struct mp_chmap best[] = { {0}, {0}, {0} };
+    return mindiff;
+}
+
+// Decide whether we should prefer old or new for the requested layout.
+// Return true if new should be used, false if old should be used.
+// If old is empty, always return new (initial case).
+static bool mp_chmap_is_better(struct mp_chmap *req, struct mp_chmap *old,
+                               struct mp_chmap *new)
+{
+    // Initial case
+    if (!old->num)
+        return true;
+
+    // Exact pick - this also ensures that the best layout is chosen if the
+    // layouts are the same, but with different order of channels.
+    if (mp_chmap_equals(req, old))
+        return false;
+    if (mp_chmap_equals(req, new))
+        return true;
+
+    // If there's no exact match, strictly do a preferred conversion.
+    bool old_pref = test_preferred_remix(req, old);
+    bool new_pref = test_preferred_remix(req, new);
+    if (old_pref && !new_pref)
+        return false;
+    if (!old_pref && new_pref)
+        return true;
+
+    int old_lost_r = mp_chmap_diffn_r(req, old); // num. channels only in req
+    int new_lost_r = mp_chmap_diffn_r(req, new);
+
+    // Imperfect upmix (no real superset) - minimize lost channels
+    if (new_lost_r != old_lost_r)
+        return new_lost_r < old_lost_r;
+
+    int old_lost = mp_chmap_diffn(req, old);
+    int new_lost = mp_chmap_diffn(req, new);
+
+    // If the situation is equal with replaced speakers, but one of them loses
+    // less if no replacements are performed, pick the better one, even if it
+    // means an upmix. This prefers exact supersets over inexact equivalents.
+    if (new_lost != old_lost)
+        return new_lost < old_lost;
+
+    struct mp_chmap old_p = *old, new_p = *new;
+    mp_chmap_remove_na(&old_p);
+    mp_chmap_remove_na(&new_p);
+
+    // Some kind of upmix. If it's perfect, prefer the smaller one. Even if not,
+    // both have equal loss, so also prefer the smaller one.
+    // Drop padding channels (NA) for the sake of this check, as the number of
+    // padding channels isn't really meaningful.
+    if (new_p.num != old_p.num)
+        return new_p.num < old_p.num;
+
+    // Again, with physical channels (minimizes number of NA channels).
+    return new->num < old->num;
+}
+
+// Determine which channel map to fallback to given a source channel map.
+bool mp_chmap_sel_fallback(const struct mp_chmap_sel *s, struct mp_chmap *map)
+{
+    struct mp_chmap best = {0};
 
     for (int n = 0; n < s->num_chmaps; n++) {
         struct mp_chmap e = s->chmaps[n];
@@ -264,25 +319,13 @@ bool mp_chmap_sel_fallback(const struct mp_chmap_sel *s, struct mp_chmap *map)
         if (mp_chmap_is_unknown(&e))
             continue;
 
-        if (test_fallbacks(&e, map, best_diffs, best))
-            continue;
-
-        // in case we didn't match any fallback retry after replacing speakers
-        for (int i = 0; i < MP_ARRAY_SIZE(speaker_replacements); i++) {
-            struct mp_chmap  t = e;
-            struct mp_chmap *r = (struct mp_chmap *)speaker_replacements[i];
-            if (replace_speakers(&t, r)) {
-                if (test_fallbacks(&t, map, best_diffs, best))
-                    continue;
-            }
-        }
+        if (mp_chmap_is_better(map, &best, &e))
+            best = e;
     }
 
-    for (int i = UPMIX_IDX; i < MP_ARRAY_SIZE(best); i++) {
-        if (best_diffs[i] < INT_MAX) {
-            *map = best[i];
-            return true;
-        }
+    if (best.num) {
+        *map = best;
+        return true;
     }
 
     return false;

audio/decode/ad_lavc.c

@@ -116,7 +116,7 @@ static int init(struct dec_audio *da, const char *decoder)
 
     mp_set_avopts(da->log, lavc_context, opts->avopts);
 
-    lavc_context->codec_tag = sh->format;
+    lavc_context->codec_tag = sh->codec_tag;
     lavc_context->sample_rate = sh_audio->samplerate;
     lavc_context->bit_rate = sh_audio->bitrate;
     lavc_context->block_align = sh_audio->block_align;
@@ -126,11 +126,7 @@ static int init(struct dec_audio *da, const char *decoder)
         lavc_context->channel_layout = mp_chmap_to_lavc(&sh_audio->channels);
 
     // demux_mkv
-    if (sh_audio->codecdata_len && sh_audio->codecdata &&
-            !lavc_context->extradata) {
-        mp_lavc_set_extradata(lavc_context, sh_audio->codecdata,
-                              sh_audio->codecdata_len);
-    }
+    mp_lavc_set_extradata(lavc_context, sh->extradata, sh->extradata_size);
 
     if (sh->lav_headers)
         mp_copy_lav_codec_headers(lavc_context, sh->lav_headers);

audio/decode/ad_spdif.c

@@ -34,10 +34,12 @@
 
 struct spdifContext {
     struct mp_log   *log;
+    enum AVCodecID   codec_id;
     AVFormatContext *lavf_ctx;
     int              out_buffer_len;
     uint8_t          out_buffer[OUTBUF_SIZE];
     bool             need_close;
+    bool             use_dts_hd;
     struct mp_audio  fmt;
 };
 
@@ -76,17 +78,78 @@ static int init(struct dec_audio *da, const char *decoder)
     struct spdifContext *spdif_ctx = talloc_zero(NULL, struct spdifContext);
     da->priv = spdif_ctx;
     spdif_ctx->log = da->log;
+    spdif_ctx->use_dts_hd = da->opts->dtshd;
+
+    if (strcmp(decoder, "dts-hd") == 0) {
+        decoder = "dts";
+        spdif_ctx->use_dts_hd = true;
+    }
+
+    spdif_ctx->codec_id = mp_codec_to_av_codec_id(decoder);
+    return spdif_ctx->codec_id != AV_CODEC_ID_NONE;
+}
+
+static int determine_codec_profile(struct dec_audio *da, AVPacket *pkt)
+{
+    struct spdifContext *spdif_ctx = da->priv;
+    int profile = FF_PROFILE_UNKNOWN;
+    AVCodecContext *ctx = NULL;
+    AVFrame *frame = NULL;
+
+    AVCodec *codec = avcodec_find_decoder(spdif_ctx->codec_id);
+    if (!codec)
+        goto done;
+
+    frame = av_frame_alloc();
+    if (!frame)
+        goto done;
+
+    ctx = avcodec_alloc_context3(codec);
+    if (!ctx)
+        goto done;
+
+    if (avcodec_open2(ctx, codec, NULL) < 0) {
+        av_free(ctx); // don't attempt to avcodec_close() an unopened ctx
+        ctx = NULL;
+        goto done;
+    }
+
+    int got_frame = 0;
+    if (avcodec_decode_audio4(ctx, frame, &got_frame, pkt) < 1 || !got_frame)
+        goto done;
+
+    profile = ctx->profile;
+
+done:
+    av_frame_free(&frame);
+    if (ctx)
+        avcodec_close(ctx);
+    avcodec_free_context(&ctx);
+
+    if (profile == FF_PROFILE_UNKNOWN)
+        MP_WARN(da, "Failed to parse codec profile.\n");
+
+    return profile;
+}
+
+static int init_filter(struct dec_audio *da, AVPacket *pkt)
+{
+    struct spdifContext *spdif_ctx = da->priv;
+
+    int profile = FF_PROFILE_UNKNOWN;
+    if (spdif_ctx->codec_id == AV_CODEC_ID_DTS)
+        profile = determine_codec_profile(da, pkt);
 
     AVFormatContext *lavf_ctx  = avformat_alloc_context();
     if (!lavf_ctx)
         goto fail;
 
+    spdif_ctx->lavf_ctx = lavf_ctx;
+
     lavf_ctx->oformat = av_guess_format("spdif", NULL, NULL);
     if (!lavf_ctx->oformat)
         goto fail;
 
-    spdif_ctx->lavf_ctx = lavf_ctx;
-
     void *buffer = av_mallocz(OUTBUF_SIZE);
     if (!buffer)
         abort();
@@ -106,14 +169,14 @@ static int init(struct dec_audio *da, const char *decoder)
     if (!stream)
         goto fail;
 
-    stream->codec->codec_id = mp_codec_to_av_codec_id(decoder);
+    stream->codec->codec_id = spdif_ctx->codec_id;
 
     AVDictionary *format_opts = NULL;
 
     int num_channels = 0;
     int sample_format = 0;
     int samplerate = 0;
-    switch (stream->codec->codec_id) {
+    switch (spdif_ctx->codec_id) {
     case AV_CODEC_ID_AAC:
         sample_format                   = AF_FORMAT_S_AAC;
         samplerate                      = 48000;
@@ -124,18 +187,21 @@ static int init(struct dec_audio *da, const char *decoder)
         samplerate                      = 48000;
         num_channels                    = 2;
         break;
-    case AV_CODEC_ID_DTS:
-        if (da->opts->dtshd) {
+    case AV_CODEC_ID_DTS: {
+        bool is_hd = profile == FF_PROFILE_DTS_HD_HRA ||
+                     profile == FF_PROFILE_DTS_HD_MA;
+        if (spdif_ctx->use_dts_hd && is_hd) {
             av_dict_set(&format_opts, "dtshd_rate", "768000", 0); // 4*192000
-            sample_format                   = AF_FORMAT_S_DTSHD;
-            samplerate                      = 192000;
-            num_channels                    = 2*4;
+            sample_format               = AF_FORMAT_S_DTSHD;
+            samplerate                  = 192000;
+            num_channels                = 2*4;
         } else {
-            sample_format                   = AF_FORMAT_S_DTS;
-            samplerate                      = 48000;
-            num_channels                    = 2;
+            sample_format               = AF_FORMAT_S_DTS;
+            samplerate                  = 48000;
+            num_channels                = 2;
         }
         break;
+    }
     case AV_CODEC_ID_EAC3:
         sample_format                   = AF_FORMAT_S_EAC3;
         samplerate                      = 192000;
@@ -167,17 +233,16 @@ static int init(struct dec_audio *da, const char *decoder)
 
     spdif_ctx->need_close = true;
 
-    return 1;
+    return 0;
 
 fail:
     uninit(da);
-    return 0;
+    return -1;
 }
 
 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;
 
     spdif_ctx->out_buffer_len  = 0;
 
@@ -195,9 +260,13 @@ static int decode_packet(struct dec_audio *da, struct mp_audio **out)
         da->pts        = mpkt->pts;
         da->pts_offset = 0;
     }
-    int ret = av_write_frame(lavf_ctx, &pkt);
+    if (!spdif_ctx->lavf_ctx) {
+        if (init_filter(da, &pkt) < 0)
+            return AD_ERR;
+    }
+    int ret = av_write_frame(spdif_ctx->lavf_ctx, &pkt);
     talloc_free(mpkt);
-    avio_flush(lavf_ctx->pb);
+    avio_flush(spdif_ctx->lavf_ctx->pb);
     if (ret < 0)
         return AD_ERR;
 
@@ -235,6 +304,8 @@ static void add_decoders(struct mp_decoder_list *list)
                            "libavformat/spdifenc audio pass-through decoder");
         }
     }
+    mp_add_decoder(list, "spdif", "dts", "dts-hd",
+                   "libavformat/spdifenc audio pass-through decoder");
 }
 
 const struct ad_functions ad_spdif = {

audio/decode/dec_audio.c

@@ -52,6 +52,7 @@ static const struct ad_functions * const ad_drivers[] = {
 
 static void uninit_decoder(struct dec_audio *d_audio)
 {
+    audio_reset_decoding(d_audio);
     if (d_audio->ad_driver) {
         MP_VERBOSE(d_audio, "Uninit audio decoder.\n");
         d_audio->ad_driver->uninit(d_audio);
@@ -59,6 +60,8 @@ static void uninit_decoder(struct dec_audio *d_audio)
     d_audio->ad_driver = NULL;
     talloc_free(d_audio->priv);
     d_audio->priv = NULL;
+    d_audio->afilter->initialized = -1;
+    d_audio->decode_format = (struct mp_audio){0};
 }
 
 static int init_audio_codec(struct dec_audio *d_audio, const char *decoder)
@@ -81,11 +84,21 @@ struct mp_decoder_list *audio_decoder_list(void)
     return list;
 }
 
-static struct mp_decoder_list *audio_select_decoders(const char *codec,
-                                                     char *selection)
+static struct mp_decoder_list *audio_select_decoders(struct dec_audio *d_audio)
 {
+    struct MPOpts *opts = d_audio->opts;
+    const char *codec = d_audio->header->codec;
+
     struct mp_decoder_list *list = audio_decoder_list();
-    struct mp_decoder_list *new = mp_select_decoders(list, codec, selection);
+    struct mp_decoder_list *new =
+        mp_select_decoders(list, codec, opts->audio_decoders);
+    if (d_audio->spdif_passthrough) {
+        struct mp_decoder_list *spdif =
+            mp_select_decoder_list(list, codec, "spdif", opts->audio_spdif);
+        mp_append_decoders(spdif, new);
+        talloc_free(new);
+        new = spdif;
+    }
     talloc_free(list);
     return new;
 }
@@ -99,14 +112,13 @@ static const struct ad_functions *find_driver(const char *name)
     return NULL;
 }
 
-int audio_init_best_codec(struct dec_audio *d_audio, char *audio_decoders)
+int audio_init_best_codec(struct dec_audio *d_audio)
 {
+    uninit_decoder(d_audio);
     assert(!d_audio->ad_driver);
-    audio_reset_decoding(d_audio);
 
     struct mp_decoder_entry *decoder = NULL;
-    struct mp_decoder_list *list =
-        audio_select_decoders(d_audio->header->codec, audio_decoders);
+    struct mp_decoder_list *list = audio_select_decoders(d_audio);
 
     mp_print_decoders(d_audio->log, MSGL_V, "Codec list:", list);
 
@@ -145,8 +157,8 @@ void audio_uninit(struct dec_audio *d_audio)
     if (!d_audio)
         return;
     MP_VERBOSE(d_audio, "Uninit audio filters...\n");
-    af_destroy(d_audio->afilter);
     uninit_decoder(d_audio);
+    af_destroy(d_audio->afilter);
     talloc_free(d_audio->waiting);
     talloc_free(d_audio);
 }

audio/decode/dec_audio.h

@@ -30,6 +30,7 @@ struct dec_audio {
     struct mp_log *log;
     struct MPOpts *opts;
     struct mpv_global *global;
+    bool spdif_passthrough;
     const struct ad_functions *ad_driver;
     struct sh_stream *header;
     struct af_stream *afilter;
@@ -57,7 +58,7 @@ enum {
 };
 
 struct mp_decoder_list *audio_decoder_list(void);
-int audio_init_best_codec(struct dec_audio *d_audio, char *audio_decoders);
+int audio_init_best_codec(struct dec_audio *d_audio);
 int audio_decode(struct dec_audio *d_audio, struct mp_audio_buffer *outbuf,
                  int minsamples);
 int initial_audio_decode(struct dec_audio *d_audio);

audio/filter/af.c

@@ -31,7 +31,6 @@
 #include "af.h"
 
 // Static list of filters
-extern const struct af_info af_info_dummy;
 extern const struct af_info af_info_delay;
 extern const struct af_info af_info_channels;
 extern const struct af_info af_info_format;
@@ -55,12 +54,9 @@ extern const struct af_info af_info_karaoke;
 extern const struct af_info af_info_scaletempo;
 extern const struct af_info af_info_bs2b;
 extern const struct af_info af_info_lavfi;
-extern const struct af_info af_info_convert24;
-extern const struct af_info af_info_convertsignendian;
 extern const struct af_info af_info_rubberband;
 
 static const struct af_info *const filter_list[] = {
-    &af_info_dummy,
     &af_info_delay,
     &af_info_channels,
     &af_info_format,
@@ -92,9 +88,6 @@ static const struct af_info *const filter_list[] = {
 #if HAVE_LIBAVFILTER
     &af_info_lavfi,
 #endif
-    // Must come last, because they're fallback format conversion filter
-    &af_info_convert24,
-    &af_info_convertsignendian,
     NULL
 };
 
@@ -334,89 +327,9 @@ static void af_print_filter_chain(struct af_stream *s, struct af_instance *at,
     MP_MSG(s, msg_level, "  [ao] %s\n", mp_audio_config_to_str(&s->output));
 }
 
-static int af_count_filters(struct af_stream *s)
-{
-    int count = 0;
-    for (struct af_instance *af = s->first; af; af = af->next)
-        count++;
-    return count;
-}
-
-// Finds the first conversion filter on the way from srcfmt to dstfmt.
-// Conversions form a DAG: each node is a format/filter pair, and possible
-// conversions are edges. We search the DAG for the shortest path.
-// Some cases visit the same filter multiple times, but with different formats
-// (like u24le->s8), so one node per format or filter separate is not enough.
-// Returns the filter and dest. format for the first conversion step.
-// (So we know what conversion filter with what format to insert next.)
-static char *af_find_conversion_filter(int srcfmt, int *dstfmt)
-{
-#define MAX_NODES (64 * 32)
-    int num_fmt = 0, num_filt = 0;
-    for (int n = 0; filter_list[n]; n++)
-        num_filt = n + 1;
-    for (int n = 0; af_fmtstr_table[n].format; n++)
-        num_fmt = n + 1;
-
-    int num_nodes = num_fmt * num_filt;
-    assert(num_nodes < MAX_NODES);
-
-    bool visited[MAX_NODES] = {0};
-    unsigned char distance[MAX_NODES];
-    short previous[MAX_NODES] = {0};
-    for (int n = 0; n < num_nodes; n++) {
-        distance[n] = 255;
-        if (af_fmtstr_table[n % num_fmt].format == srcfmt)
-            distance[n] = 0;
-    }
-
-    while (1) {
-        int next = -1;
-        for (int n = 0; n < num_nodes; n++) {
-            if (!visited[n] && (next < 0 || (distance[n] < distance[next])))
-                next = n;
-        }
-        if (next < 0 || distance[next] == 255)
-            return NULL;
-        visited[next] = true;
-
-        int fmt = next % num_fmt;
-        if (af_fmtstr_table[fmt].format == *dstfmt) {
-            // Best match found
-            for (int cur = next; cur >= 0; cur = previous[cur] - 1) {
-                if (distance[cur] == 1) {
-                    *dstfmt = af_fmtstr_table[cur % num_fmt].format;
-                    return (char *)filter_list[cur / num_fmt]->name;
-                }
-            }
-            return NULL;
-        }
-
-        for (int n = 0; filter_list[n]; n++) {
-            const struct af_info *af = filter_list[n];
-            if (!af->test_conversion)
-                continue;
-            for (int i = 0; af_fmtstr_table[i].format; i++) {
-                if (i != fmt && af->test_conversion(af_fmtstr_table[fmt].format,
-                                                    af_fmtstr_table[i].format))
-                {
-                    int other = n * num_fmt + i;
-                    int ndist = distance[next] + 1;
-                    if (ndist < distance[other]) {
-                        distance[other] = ndist;
-                        previous[other] = next + 1;
-                    }
-                }
-            }
-        }
-    }
-    assert(0);
-#undef MAX_NODES
-}
-
 static bool af_is_conversion_filter(struct af_instance *af)
 {
-    return af && af->info->test_conversion != NULL;
+    return af && strcmp(af->info->name, "lavrresample") == 0;
 }
 
 // in is what af can take as input - insert a conversion filter if the actual
@@ -436,21 +349,24 @@ static int af_fix_format_conversion(struct af_stream *s,
     if (actual.format == in.format)
         return AF_FALSE;
     int dstfmt = in.format;
-    char *filter = af_find_conversion_filter(actual.format, &dstfmt);
-    if (!filter)
+    char *filter = "lavrresample";
+    if (!af_lavrresample_test_conversion(actual.format, dstfmt))
         return AF_ERROR;
     if (strcmp(filter, prev->info->name) == 0) {
         if (prev->control(prev, AF_CONTROL_SET_FORMAT, &dstfmt) == AF_OK) {
             *p_af = prev;
             return AF_OK;
         }
+        return AF_ERROR;
     }
     struct af_instance *new = af_prepend(s, af, filter, NULL);
     if (new == NULL)
         return AF_ERROR;
     new->auto_inserted = true;
-    if (AF_OK != (rv = new->control(new, AF_CONTROL_SET_FORMAT, &dstfmt)))
+    if (AF_OK != (rv = new->control(new, AF_CONTROL_SET_FORMAT, &dstfmt))) {
+        af_remove(s, new);
         return rv;
+    }
     *p_af = new;
     return AF_OK;
 }
@@ -528,8 +444,8 @@ static int af_reinit(struct af_stream *s)
     // Start with the second filter, as the first filter is the special input
     // filter which needs no initialization.
     struct af_instance *af = s->first->next;
-    // Up to 7 retries per filter (channel, rate, 5x format conversions)
-    int max_retry = af_count_filters(s) * 7;
+    // Up to 4 retries per filter (channel, rate, format conversions)
+    int max_retry = 4;
     int retry = 0;
     while (af) {
         if (retry >= max_retry)
@@ -575,9 +491,9 @@ static int af_reinit(struct af_stream *s)
             int fmt_in1 = af->prev->data->format;
             int fmt_in2 = in.format;
             if (af_fmt_is_valid(fmt_in1) && af_fmt_is_valid(fmt_in2)) {
-                bool spd1 = AF_FORMAT_IS_IEC61937(fmt_in1);
-                bool spd2 = AF_FORMAT_IS_IEC61937(fmt_in2);
-                if (spd1 != spd2) {
+                bool spd1 = af_fmt_is_spdif(fmt_in1);
+                bool spd2 = af_fmt_is_spdif(fmt_in2);
+                if (spd1 != spd2 && af->next) {
                     MP_WARN(af, "Filter %s apparently cannot be used due to "
                                 "spdif passthrough - removing it.\n",
                                 af->info->name);
@@ -601,6 +517,8 @@ static int af_reinit(struct af_stream *s)
                    af->info->name, rv);
             goto error;
         }
+        if (af && !af->auto_inserted)
+            retry = 0;
     }
 
     /* Set previously unset fields in s->output to those of the filter chain
@@ -733,7 +651,7 @@ struct af_instance *af_add(struct af_stream *s, char *name, char *label,
         return NULL;
     new->label = talloc_strdup(new, label);
 
-    // Reinitalize the filter list
+    // Reinitialize the filter list
     if (af_reinit(s) != AF_OK) {
         af_remove_by_label(s, label);
         return NULL;
@@ -832,6 +750,16 @@ static struct mp_audio *af_dequeue_output_frame(struct af_instance *af)
     return res;
 }
 
+static void read_remaining(struct af_instance *af)
+{
+    int num_frames;
+    do {
+        num_frames = af->num_out_queued;
+        if (!af->filter_out || af->filter_out(af) < 0)
+            break;
+    } while (num_frames != af->num_out_queued);
+}
+
 static int af_do_filter(struct af_instance *af, struct mp_audio *frame)
 {
     if (frame)
@@ -871,6 +799,7 @@ int af_output_frame(struct af_stream *s, bool eof)
             // Flush remaining frames on EOF, but do that only if the previous
             // filters have been flushed (i.e. they have no more output).
             if (eof && !last) {
+                read_remaining(cur);
                 int r = af_do_filter(cur, NULL);
                 if (r < 0)
                     return r;

audio/filter/af.h

@@ -48,7 +48,6 @@ struct af_info {
     const char *name;
     const int flags;
     int (*open)(struct af_instance *vf);
-    bool (*test_conversion)(int src_format, int dst_format);
     int priv_size;
     const void *priv_defaults;
     const struct m_option *options;
@@ -158,8 +157,9 @@ double af_calc_delay(struct af_stream *s);
 
 int af_test_output(struct af_instance *af, struct mp_audio *out);
 
-int af_from_dB(int n, float *in, float *out, float k, float mi, float ma);
 int af_from_ms(int n, float *in, int *out, int rate, float mi, float ma);
 float af_softclip(float a);
 
+bool af_lavrresample_test_conversion(int src_format, int dst_format);
+
 #endif /* MPLAYER_AF_H */

audio/filter/af_bs2b.c

@@ -57,12 +57,8 @@ static int filter_##name(struct af_instance *af, struct mp_audio *data) \
 #define FILTERS \
     FILTER(FLOAT,   f)      \
     FILTER(S32,     s32)    \
-    FILTER(U32,     u32)    \
     FILTER(S24,     s24)    \
-    FILTER(U24,     u24)    \
     FILTER(S16,     s16)    \
-    FILTER(U16,     u16)    \
-    FILTER(S8,      s8)     \
     FILTER(U8,      u8)
 
 #define FILTER DEF_FILTER

audio/filter/af_convert24.c

@@ -1,122 +0,0 @@
-/*
- * 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 <stdlib.h>
-#include <assert.h>
-
-#include "audio/format.h"
-#include "af.h"
-#include "osdep/endian.h"
-
-static bool test_conversion(int src_format, int dst_format)
-{
-    return (src_format == AF_FORMAT_U24 && dst_format == AF_FORMAT_U32) ||
-           (src_format == AF_FORMAT_S24 && dst_format == AF_FORMAT_S32) ||
-           (src_format == AF_FORMAT_U32 && dst_format == AF_FORMAT_U24) ||
-           (src_format == AF_FORMAT_S32 && dst_format == AF_FORMAT_S24);
-}
-
-static int control(struct af_instance *af, int cmd, void *arg)
-{
-    switch (cmd) {
-    case AF_CONTROL_REINIT: {
-        struct mp_audio *in = arg;
-        struct mp_audio orig_in = *in;
-        struct mp_audio *out = af->data;
-
-        if (!test_conversion(in->format, out->format))
-            return AF_DETACH;
-
-        if ((in->format & AF_FORMAT_BITS_MASK) == AF_FORMAT_24BIT) {
-            mp_audio_set_format(out, af_fmt_change_bits(in->format, 32));
-        } else if ((in->format & AF_FORMAT_BITS_MASK) == AF_FORMAT_32BIT) {
-            mp_audio_set_format(out, af_fmt_change_bits(in->format, 24));
-        } else {
-            abort();
-        }
-
-        out->rate = in->rate;
-        mp_audio_set_channels(out, &in->channels);
-
-        assert(test_conversion(in->format, out->format));
-
-        return mp_audio_config_equals(in, &orig_in) ? AF_OK : AF_FALSE;
-    }
-    case AF_CONTROL_SET_FORMAT: {
-        mp_audio_set_format(af->data, *(int*)arg);
-        return AF_OK;
-    }
-    }
-    return AF_UNKNOWN;
-}
-
-// The LSB is always ignored.
-#if BYTE_ORDER == BIG_ENDIAN
-#define SHIFT(x) ((3-(x))*8)
-#else
-#define SHIFT(x) (((x)+1)*8)
-#endif
-
-static int filter(struct af_instance *af, struct mp_audio *data)
-{
-    if (!data)
-        return 0;
-    struct mp_audio *out =
-        mp_audio_pool_get(af->out_pool, af->data, data->samples);
-    if (!out) {
-        talloc_free(data);
-        return -1;
-    }
-    mp_audio_copy_attributes(out, data);
-
-    size_t len = mp_audio_psize(data) / data->bps;
-    if (data->bps == 4) {
-        for (int s = 0; s < len; s++) {
-            uint32_t val = *((uint32_t *)data->planes[0] + s);
-            uint8_t *ptr = (uint8_t *)out->planes[0] + s * 3;
-            ptr[0] = val >> SHIFT(0);
-            ptr[1] = val >> SHIFT(1);
-            ptr[2] = val >> SHIFT(2);
-        }
-    } else {
-        for (int s = 0; s < len; s++) {
-            uint8_t *ptr = (uint8_t *)data->planes[0] + s * 3;
-            uint32_t val = ptr[0] << SHIFT(0)
-                         | ptr[1] << SHIFT(1)
-                         | ptr[2] << SHIFT(2);
-            *((uint32_t *)out->planes[0] + s) = val;
-        }
-    }
-
-    talloc_free(data);
-    af_add_output_frame(af, out);
-    return 0;
-}
-
-static int af_open(struct af_instance *af)
-{
-    af->control = control;
-    af->filter_frame = filter;
-    return AF_OK;
-}
-
-const struct af_info af_info_convert24 = {
-    .info = "Convert between 24 and 32 bit sample format",
-    .name = "convert24",
-    .open = af_open,
-    .test_conversion = test_conversion,
-};

audio/filter/af_convertsignendian.c

@@ -1,107 +0,0 @@
-/*
- * 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 <stdlib.h>
-#include <assert.h>
-
-#include "af.h"
-#include "audio/format.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;
-    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;
-}
-
-static int control(struct af_instance *af, int cmd, void *arg)
-{
-    switch (cmd) {
-    case AF_CONTROL_REINIT: {
-        struct mp_audio *in = arg;
-        struct mp_audio orig_in = *in;
-        struct mp_audio *out = af->data;
-
-        if (!test_conversion(in->format, out->format))
-            return AF_DETACH;
-
-        out->rate = in->rate;
-        mp_audio_set_channels(out, &in->channels);
-
-        return mp_audio_config_equals(in, &orig_in) ? AF_OK : AF_FALSE;
-    }
-    case AF_CONTROL_SET_FORMAT: {
-        mp_audio_set_format(af->data, *(int*)arg);
-        return AF_OK;
-    }
-    }
-    return AF_UNKNOWN;
-}
-
-static void si2us(void *data, int len, int bps)
-{
-    ptrdiff_t i = -(len * bps);
-    uint8_t *p = &((uint8_t *)data)[len * bps];
-    if (BYTE_ORDER == LITTLE_ENDIAN && bps > 1)
-        p += bps - 1;
-    if (len <= 0)
-        return;
-    do {
-        p[i] ^= 0x80;
-    } while (i += bps);
-}
-
-static int filter(struct af_instance *af, struct mp_audio *data)
-{
-    if (!data)
-        return 0;
-    if (af_make_writeable(af, data) < 0) {
-        talloc_free(data);
-        return -1;
-    }
-
-    int infmt = data->format;
-    int outfmt = af->data->format;
-    size_t len = data->samples * data->nch;
-
-    if ((infmt & AF_FORMAT_SIGN_MASK) != (outfmt & AF_FORMAT_SIGN_MASK))
-        si2us(data->planes[0], len, data->bps);
-
-    mp_audio_set_format(data, outfmt);
-    af_add_output_frame(af, data);
-    return 0;
-}
-
-static int af_open(struct af_instance *af)
-{
-    af->control = control;
-    af->filter_frame = filter;
-    return AF_OK;
-}
-
-const struct af_info af_info_convertsignendian = {
-    .info = "Convert between sample format sign",
-    .name = "convertsign",
-    .open = af_open,
-    .test_conversion = test_conversion,
-};

audio/filter/af_drc.c

@@ -131,7 +131,7 @@ static void method1_int16(af_drc_t *s, struct mp_audio *c)
     data[i] = tmp;
   }
 
-  // Evaulation of newavg (not 100% accurate because of values clamping)
+  // Evaluation of newavg (not 100% accurate because of values clamping)
   newavg = s->mul * curavg;
 
   // Stores computed values for future smoothing
@@ -168,7 +168,7 @@ static void method1_float(af_drc_t *s, struct mp_audio *c)
   for (i = 0; i < len; i++)
     data[i] *= s->mul;
 
-  // Evaulation of newavg (not 100% accurate because of values clamping)
+  // Evaluation of newavg (not 100% accurate because of values clamping)
   newavg = s->mul * curavg;
 
   // Stores computed values for future smoothing
@@ -216,7 +216,7 @@ static void method2_int16(af_drc_t *s, struct mp_audio *c)
     data[i] = tmp;
   }
 
-  // Evaulation of newavg (not 100% accurate because of values clamping)
+  // Evaluation of newavg (not 100% accurate because of values clamping)
   newavg = s->mul * curavg;
 
   // Stores computed values for future smoothing
@@ -262,7 +262,7 @@ static void method2_float(af_drc_t *s, struct mp_audio *c)
   for (i = 0; i < len; i++)
     data[i] *= s->mul;
 
-  // Evaulation of newavg (not 100% accurate because of values clamping)
+  // Evaluation of newavg (not 100% accurate because of values clamping)
   newavg = s->mul * curavg;
 
   // Stores computed values for future smoothing

audio/filter/af_dummy.c

@@ -1,61 +0,0 @@
-/*
- * The name speaks for itself. This filter is a dummy and will
- * not blow up regardless of what you do with it.
- *
- * Original author: Anders
- *
- * 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 <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-
-#include "af.h"
-
-// Initialization and runtime control
-static int control(struct af_instance* af, int cmd, void* arg)
-{
-  switch(cmd){
-  case AF_CONTROL_REINIT: ;
-    *af->data = *(struct mp_audio*)arg;
-    MP_VERBOSE(af, "Was reinitialized: %iHz/%ich/%s\n",
-        af->data->rate,af->data->nch,af_fmt_to_str(af->data->format));
-    return AF_OK;
-  }
-  return AF_UNKNOWN;
-}
-
-// Filter data through filter
-static int filter(struct af_instance* af, struct mp_audio* data)
-{
-    af_add_output_frame(af, data);
-    return 0;
-}
-
-// Allocate memory and set function pointers
-static int af_open(struct af_instance* af){
-  af->control=control;
-  af->filter_frame=filter;
-  return AF_OK;
-}
-
-// Description of this filter
-const struct af_info af_info_dummy = {
-    .info = "dummy",
-    .name = "dummy",
-    .open = af_open,
-};

audio/filter/af_export.c

@@ -167,7 +167,7 @@ static int filter(struct af_instance *af, struct mp_audio *data)
     return 0;
   struct mp_audio*      c   = data;          // Current working data
   af_export_t*  s   = af->priv;     // Setup for this instance
-  int16_t*      a   = c->planes[0];          // Incomming sound
+  int16_t*      a   = c->planes[0];          // Incoming sound
   int           nch = c->nch;        // Number of channels
   int           len = c->samples*c->nch; // Number of sample in data chunk
   int           sz  = s->sz;         // buffer size (in samples)

audio/filter/af_hrtf.c

@@ -206,7 +206,7 @@ static inline void matrix_decode(short *in, const int k, const int il,
       information about Lt, Rt correlation.  This effectively reshapes
       the front and rear "cones" to concentrate Lt + Rt to C and
       introduce Lt - Rt in L, R. */
-   /* 0.67677 is the emprical lower bound for lpr_gain. */
+   /* 0.67677 is the empirical lower bound for lpr_gain. */
    c_gain = 8 * (*adapt_lpr_gain - 0.67677);
    c_gain = c_gain > 0 ? c_gain : 0;
    /* c_gain should not be too high, not even reaching full

audio/filter/af_ladspa.c

@@ -144,7 +144,7 @@ static int af_ladspa_parse_plugin(struct af_instance *af) {
     LADSPA_PortRangeHint hint;
 
     if (!setup->libhandle)
-        return AF_ERROR; /* only call parse after a succesful load */
+        return AF_ERROR; /* only call parse after a successful load */
     if (!setup->plugin_descriptor)
         return AF_ERROR; /* same as above */
 

audio/filter/af_lavcac3enc.c

@@ -26,7 +26,6 @@
 #include <assert.h>
 
 #include <libavcodec/avcodec.h>
-#include <libavutil/audioconvert.h>
 #include <libavutil/intreadwrite.h>
 #include <libavutil/common.h>
 #include <libavutil/bswap.h>
@@ -75,7 +74,7 @@ static int control(struct af_instance *af, int cmd, void *arg)
         struct mp_audio *in = arg;
         struct mp_audio orig_in = *in;
 
-        if (AF_FORMAT_IS_SPECIAL(in->format) || in->nch < s->cfg_min_channel_num)
+        if (!af_fmt_is_pcm(in->format) || in->nch < s->cfg_min_channel_num)
             return AF_DETACH;
 
         mp_audio_set_format(in, s->in_sampleformat);
@@ -156,6 +155,13 @@ static void uninit(struct af_instance* af)
     }
 }
 
+static void update_delay(struct af_instance *af)
+{
+    af_ac3enc_t *s = af->priv;
+    af->delay = ((s->pending ? s->pending->samples : 0) + s->input->samples) /
+                (double)s->input->rate;
+}
+
 static int filter_frame(struct af_instance *af, struct mp_audio *audio)
 {
     af_ac3enc_t *s = af->priv;
@@ -166,6 +172,7 @@ static int filter_frame(struct af_instance *af, struct mp_audio *audio)
 
     talloc_free(s->pending);
     s->pending = audio;
+    update_delay(af);
     return 0;
 }
 
@@ -177,22 +184,27 @@ static void swap_16(uint16_t *ptr, size_t size)
 
 // Copy data from input frame to encode frame (because libavcodec wants a full
 // AC3 frame for encoding, while filter input frames can be smaller or larger).
-// Return true if the
-static bool fill_buffer(af_ac3enc_t *s)
+// Return true if the frame is complete.
+static bool fill_buffer(struct af_instance *af)
 {
+    af_ac3enc_t *s = af->priv;
+
+    af->delay = 0;
+
     if (s->pending) {
         int copy = MPMIN(s->in_samples - s->input->samples, s->pending->samples);
         s->input->samples += copy;
         mp_audio_copy(s->input, s->input->samples - copy, s->pending, 0, copy);
         mp_audio_skip_samples(s->pending, copy);
     }
+    update_delay(af);
     return s->input->samples >= s->in_samples;
 }
 
 static int filter_out(struct af_instance *af)
 {
     af_ac3enc_t *s = af->priv;
-    if (!fill_buffer(s))
+    if (!fill_buffer(af))
         return 0; // need more input
 
     AVFrame *frame = av_frame_alloc();
@@ -256,6 +268,7 @@ static int filter_out(struct af_instance *af)
     swap_16((uint16_t *)(buf + header_len), s->pkt.size / 2);
     out->samples = frame_size / out->sstride;
     af_add_output_frame(af, out);
+    update_delay(af);
     return 0;
 }
 

audio/filter/af_lavrresample.c

@@ -28,7 +28,6 @@
 #include <assert.h>
 
 #include <libavutil/opt.h>
-#include <libavutil/audioconvert.h>
 #include <libavutil/common.h>
 #include <libavutil/samplefmt.h>
 #include <libavutil/mathematics.h>
@@ -58,6 +57,7 @@
 #include "options/m_option.h"
 #include "audio/filter/af.h"
 #include "audio/fmt-conversion.h"
+#include "osdep/endian.h"
 
 struct af_resample_opts {
     int filter_size;
@@ -78,9 +78,10 @@ struct af_resample {
     int allow_detach;
     char **avopts;
     double playback_speed;
-    struct mp_audio *pending;
-    bool avrctx_ok;
     struct AVAudioResampleContext *avrctx;
+    struct mp_audio avrctx_fmt; // output format of avrctx
+    struct mp_audio pool_fmt; // format used to allocate frames for avrctx output
+    struct mp_audio pre_out_fmt; // format before final conversion (S24)
     struct AVAudioResampleContext *avrctx_out; // for output channel reordering
     struct af_resample_opts ctx;   // opts in the context
     struct af_resample_opts opts;  // opts requested by the user
@@ -91,33 +92,52 @@ struct af_resample {
 };
 
 #if HAVE_LIBAVRESAMPLE
-static int get_delay(struct af_resample *s)
+static double get_delay(struct af_resample *s)
 {
-    return avresample_get_delay(s->avrctx);
+    return avresample_get_delay(s->avrctx) / (double)s->ctx.in_rate +
+           avresample_available(s->avrctx) / (double)s->ctx.out_rate;
 }
 static void drop_all_output(struct af_resample *s)
 {
     while (avresample_read(s->avrctx, NULL, 1000) > 0) {}
 }
-static int get_drain_samples(struct af_resample *s)
+static int get_out_samples(struct af_resample *s, int in_samples)
 {
-    return avresample_get_out_samples(s->avrctx, 0);
+    return avresample_get_out_samples(s->avrctx, in_samples);
 }
 #else
-static int get_delay(struct af_resample *s)
+static double get_delay(struct af_resample *s)
 {
-    return swr_get_delay(s->avrctx, s->ctx.in_rate);
+    int64_t base = s->ctx.in_rate * (int64_t)s->ctx.out_rate;
+    return swr_get_delay(s->avrctx, base) / (double)base;
 }
 static void drop_all_output(struct af_resample *s)
 {
     while (swr_drop_output(s->avrctx, 1000) > 0) {}
 }
-static int get_drain_samples(struct af_resample *s)
+static int get_out_samples(struct af_resample *s, int in_samples)
 {
-    return 4096; // libswscale does not have this
+#if LIBSWRESAMPLE_VERSION_MAJOR > 1 || LIBSWRESAMPLE_VERSION_MINOR >= 2
+    return swr_get_out_samples(s->avrctx, in_samples);
+#else
+    return av_rescale_rnd(in_samples, s->ctx.out_rate, s->ctx.in_rate, AV_ROUND_UP)
+           + swr_get_delay(s->avrctx, s->ctx.out_rate);
+#endif
 }
 #endif
 
+static void close_lavrr(struct af_instance *af)
+{
+    struct af_resample *s = af->priv;
+
+    if (s->avrctx)
+        avresample_close(s->avrctx);
+    avresample_free(&s->avrctx);
+    if (s->avrctx_out)
+        avresample_close(s->avrctx_out);
+    avresample_free(&s->avrctx_out);
+}
+
 static int resample_frame(struct AVAudioResampleContext *r,
                           struct mp_audio *out, struct mp_audio *in)
 {
@@ -140,47 +160,59 @@ static int rate_from_speed(int rate, double speed)
     return lrint(rate * speed);
 }
 
-static bool needs_lavrctx_reconfigure(struct af_resample *s,
-                                      struct mp_audio *in,
-                                      struct mp_audio *out)
+// Return the format libavresample should convert to, given the final output
+// format mp_format. In some cases (S24) we perform an extra conversion step,
+// and signal here what exactly libavresample should output. It will be the
+// input to the final conversion to mp_format.
+static int check_output_conversion(int mp_format)
 {
-    return s->ctx.in_rate_af  != in->rate ||
-           s->ctx.in_format   != in->format ||
-           !mp_chmap_equals(&s->ctx.in_channels, &in->channels) ||
-           s->ctx.out_rate    != out->rate ||
-           s->ctx.out_format  != out->format ||
-           !mp_chmap_equals(&s->ctx.out_channels, &out->channels) ||
-           s->ctx.filter_size != s->opts.filter_size ||
-           s->ctx.phase_shift != s->opts.phase_shift ||
-           s->ctx.linear      != s->opts.linear ||
-           s->ctx.cutoff      != s->opts.cutoff;
-
+    if (mp_format == AF_FORMAT_S24)
+        return AV_SAMPLE_FMT_S32;
+    return af_to_avformat(mp_format);
 }
 
-static bool test_conversion(int src_format, int dst_format)
+bool af_lavrresample_test_conversion(int src_format, int dst_format)
 {
     return af_to_avformat(src_format) != AV_SAMPLE_FMT_NONE &&
-           af_to_avformat(dst_format) != AV_SAMPLE_FMT_NONE;
+           check_output_conversion(dst_format) != AV_SAMPLE_FMT_NONE;
+}
+
+// mp_chmap_get_reorder() performs:
+//  to->speaker[n] = from->speaker[src[n]]
+// but libavresample does:
+//  to->speaker[dst[n]] = from->speaker[n]
+static void transpose_order(int *map, int num)
+{
+    int nmap[MP_NUM_CHANNELS] = {0};
+    for (int n = 0; n < num; n++) {
+        for (int i = 0; i < num; i++) {
+            if (map[n] == i)
+                nmap[i] = n;
+        }
+    }
+    memcpy(map, nmap, sizeof(nmap));
 }
 
 static int configure_lavrr(struct af_instance *af, struct mp_audio *in,
-                           struct mp_audio *out)
+                           struct mp_audio *out, bool verbose)
 {
     struct af_resample *s = af->priv;
 
-    s->avrctx_ok = false;
+    close_lavrr(af);
+
+    s->avrctx = avresample_alloc_context();
+    s->avrctx_out = avresample_alloc_context();
+    if (!s->avrctx || !s->avrctx_out)
+        goto error;
 
     enum AVSampleFormat in_samplefmt = af_to_avformat(in->format);
-    enum AVSampleFormat out_samplefmt = af_to_avformat(out->format);
+    enum AVSampleFormat out_samplefmt = check_output_conversion(out->format);
+    enum AVSampleFormat out_samplefmtp = av_get_planar_sample_fmt(out_samplefmt);
 
-    if (in_samplefmt == AV_SAMPLE_FMT_NONE || out_samplefmt == AV_SAMPLE_FMT_NONE)
-        return AF_ERROR;
-
-    avresample_close(s->avrctx);
-    avresample_close(s->avrctx_out);
-
-    talloc_free(s->pending);
-    s->pending = NULL;
+    if (in_samplefmt == AV_SAMPLE_FMT_NONE ||
+        out_samplefmt == AV_SAMPLE_FMT_NONE ||
+        out_samplefmtp == AV_SAMPLE_FMT_NONE)
+        goto error;
 
     s->ctx.out_rate    = out->rate;
     s->ctx.in_rate_af  = in->rate;
@@ -205,7 +237,7 @@ static int configure_lavrr(struct af_instance *af, struct mp_audio *in,
 #endif
 
     if (mp_set_avopts(af->log, s->avrctx, s->avopts) < 0)
-        return AF_ERROR;
+        goto error;
 
     struct mp_chmap map_in = in->channels;
     struct mp_chmap map_out = out->channels;
@@ -220,27 +252,68 @@ static int configure_lavrr(struct af_instance *af, struct mp_audio *in,
     uint64_t in_ch_layout = mp_chmap_to_lavc_unchecked(&map_in);
     uint64_t out_ch_layout = mp_chmap_to_lavc_unchecked(&map_out);
 
-    av_opt_set_int(s->avrctx, "in_channel_layout",  in_ch_layout, 0);
-    av_opt_set_int(s->avrctx, "out_channel_layout", out_ch_layout, 0);
-
-    av_opt_set_int(s->avrctx, "in_sample_rate",     s->ctx.in_rate, 0);
-    av_opt_set_int(s->avrctx, "out_sample_rate",    s->ctx.out_rate, 0);
-
-    av_opt_set_int(s->avrctx, "in_sample_fmt",      in_samplefmt, 0);
-    av_opt_set_int(s->avrctx, "out_sample_fmt",     out_samplefmt, 0);
-
-    struct mp_chmap in_lavc;
+    struct mp_chmap in_lavc, out_lavc;
     mp_chmap_from_lavc(&in_lavc, in_ch_layout);
-    mp_chmap_get_reorder(s->reorder_in, &map_in, &in_lavc);
-
-    struct mp_chmap out_lavc;
     mp_chmap_from_lavc(&out_lavc, out_ch_layout);
+
+    if (verbose && !mp_chmap_equals(&in_lavc, &out_lavc)) {
+        MP_VERBOSE(af, "Remix: %s -> %s\n", mp_chmap_to_str(&in_lavc),
+                                            mp_chmap_to_str(&out_lavc));
+    }
+
+    if (in_lavc.num != map_in.num) {
+        // For handling NA channels, we would have to add a planarization step.
+        MP_FATAL(af, "Unsupported channel remapping.\n");
+        goto error;
+    }
+
+    mp_chmap_get_reorder(s->reorder_in, &map_in, &in_lavc);
+    transpose_order(s->reorder_in, map_in.num);
+
+    if (mp_chmap_equals(&out_lavc, &map_out)) {
+        // No intermediate step required - output new format directly.
+        out_samplefmtp = out_samplefmt;
+    } else {
+        // Verify that we really just reorder and/or insert NA channels.
+        struct mp_chmap withna = out_lavc;
+        mp_chmap_fill_na(&withna, map_out.num);
+        if (withna.num != map_out.num)
+            goto error;
+    }
     mp_chmap_get_reorder(s->reorder_out, &out_lavc, &map_out);
 
-    // Same configuration; we just reorder.
-    av_opt_set_int(s->avrctx_out, "in_channel_layout",  out_ch_layout, 0);
-    av_opt_set_int(s->avrctx_out, "out_channel_layout", out_ch_layout, 0);
-    av_opt_set_int(s->avrctx_out, "in_sample_fmt",      out_samplefmt, 0);
+    s->avrctx_fmt = *out;
+    mp_audio_set_channels(&s->avrctx_fmt, &out_lavc);
+    mp_audio_set_format(&s->avrctx_fmt, af_from_avformat(out_samplefmtp));
+
+    s->pre_out_fmt = *out;
+    mp_audio_set_format(&s->pre_out_fmt, af_from_avformat(out_samplefmt));
+
+    // If there are NA channels, the final output will have more channels than
+    // the avrctx output. Also, avrctx will output planar (out_samplefmtp was
+    // not overwritten). Allocate the output frame with more channels, so the
+    // NA channels can be trivially added.
+    s->pool_fmt = s->avrctx_fmt;
+    if (map_out.num > out_lavc.num)
+        mp_audio_set_channels(&s->pool_fmt, &map_out);
+
+    // Real conversion; output is input to avrctx_out.
+    av_opt_set_int(s->avrctx, "in_channel_layout",  in_ch_layout, 0);
+    av_opt_set_int(s->avrctx, "out_channel_layout", out_ch_layout, 0);
+    av_opt_set_int(s->avrctx, "in_sample_rate",     s->ctx.in_rate, 0);
+    av_opt_set_int(s->avrctx, "out_sample_rate",    s->ctx.out_rate, 0);
+    av_opt_set_int(s->avrctx, "in_sample_fmt",      in_samplefmt, 0);
+    av_opt_set_int(s->avrctx, "out_sample_fmt",     out_samplefmtp, 0);
+
+    // Just needs the correct number of channels.
+    int fake_out_ch_layout = av_get_default_channel_layout(map_out.num);
+    if (!fake_out_ch_layout)
+        goto error;
+
+    // Deplanarize if needed.
+    av_opt_set_int(s->avrctx_out, "in_channel_layout",  fake_out_ch_layout, 0);
+    av_opt_set_int(s->avrctx_out, "out_channel_layout", fake_out_ch_layout, 0);
+    av_opt_set_int(s->avrctx_out, "in_sample_fmt",      out_samplefmtp, 0);
     av_opt_set_int(s->avrctx_out, "out_sample_fmt",     out_samplefmt, 0);
     av_opt_set_int(s->avrctx_out, "in_sample_rate",     s->ctx.out_rate, 0);
     av_opt_set_int(s->avrctx_out, "out_sample_rate",    s->ctx.out_rate, 0);
@@ -249,16 +322,16 @@ static int configure_lavrr(struct af_instance *af, struct mp_audio *in,
     //  * This function can only be called when the allocated context is not open.
     //  * Also, the input channel layout must have already been set.
     avresample_set_channel_mapping(s->avrctx, s->reorder_in);
-    avresample_set_channel_mapping(s->avrctx_out, s->reorder_out);
 
-    if (avresample_open(s->avrctx) < 0 ||
-        avresample_open(s->avrctx_out) < 0)
-    {
+    if (avresample_open(s->avrctx) < 0 || avresample_open(s->avrctx_out) < 0) {
         MP_ERR(af, "Cannot open Libavresample Context. \n");
-        return AF_ERROR;
+        goto error;
     }
-    s->avrctx_ok = true;
     return AF_OK;
+
+error:
+    close_lavrr(af);
+    return AF_ERROR;
 }
 
 
@@ -286,20 +359,20 @@ static int control(struct af_instance *af, int cmd, void *arg)
 
         if (af_to_avformat(in->format) == AV_SAMPLE_FMT_NONE)
             mp_audio_set_format(in, AF_FORMAT_FLOAT);
-        if (af_to_avformat(out->format) == AV_SAMPLE_FMT_NONE)
+        if (check_output_conversion(out->format) == AV_SAMPLE_FMT_NONE)
             mp_audio_set_format(out, in->format);
 
         int r = ((in->format == orig_in.format) &&
                 mp_chmap_equals(&in->channels, &orig_in.channels))
                 ? AF_OK : AF_FALSE;
 
-        if (r == AF_OK && needs_lavrctx_reconfigure(s, in, out))
-            r = configure_lavrr(af, in, out);
+        if (r == AF_OK)
+            r = configure_lavrr(af, in, out, true);
         return r;
     }
     case AF_CONTROL_SET_FORMAT: {
         int format = *(int *)arg;
-        if (format && af_to_avformat(format) == AV_SAMPLE_FMT_NONE)
+        if (format && check_output_conversion(format) == AV_SAMPLE_FMT_NONE)
             return AF_FALSE;
 
         mp_audio_set_format(af->data, format);
@@ -315,60 +388,71 @@ static int control(struct af_instance *af, int cmd, void *arg)
     case AF_CONTROL_SET_PLAYBACK_SPEED_RESAMPLE: {
         s->playback_speed = *(double *)arg;
         int new_rate = rate_from_speed(s->ctx.in_rate_af, s->playback_speed);
-        if (new_rate != s->ctx.in_rate && s->avrctx_ok && af->fmt_out.format) {
+        if (new_rate != s->ctx.in_rate && s->avrctx && af->fmt_out.format) {
             // Before reconfiguring, drain the audio that is still buffered
             // in the resampler.
-            struct mp_audio *pending = talloc_zero(NULL, struct mp_audio);
-            mp_audio_copy_config(pending, &af->fmt_out);
-            pending->samples = get_drain_samples(s);
-            if (pending->samples > 0) {
-                mp_audio_realloc_min(pending, pending->samples);
-                int r = resample_frame(s->avrctx, pending, NULL);
-                pending->samples = MPMAX(r, 0);
-            }
+            af->filter_frame(af, NULL);
             // Reinitialize resampler.
-            configure_lavrr(af, &af->fmt_in, &af->fmt_out);
-            s->pending = pending;
+            configure_lavrr(af, &af->fmt_in, &af->fmt_out, false);
         }
         return AF_OK;
     }
     case AF_CONTROL_RESET:
-        drop_all_output(s);
+        if (s->avrctx)
+            drop_all_output(s);
         return AF_OK;
     }
     return AF_UNKNOWN;
 }
 
-#undef ctx_opt_set_int
-#undef ctx_opt_set_dbl
-
 static void uninit(struct af_instance *af)
 {
-    struct af_resample *s = af->priv;
-    if (s->avrctx)
-        avresample_close(s->avrctx);
-    avresample_free(&s->avrctx);
-    if (s->avrctx_out)
-        avresample_close(s->avrctx_out);
-    avresample_free(&s->avrctx_out);
-    talloc_free(s->pending);
+    close_lavrr(af);
 }
 
-static bool needs_reorder(int *reorder, int num_ch)
+// The LSB is always ignored.
+#if BYTE_ORDER == BIG_ENDIAN
+#define SHIFT24(x) ((3-(x))*8)
+#else
+#define SHIFT24(x) (((x)+1)*8)
+#endif
+
+static void extra_output_conversion(struct af_instance *af, struct mp_audio *mpa)
 {
-    for (int n = 0; n < num_ch; n++) {
-        if (reorder[n] != n)
-            return true;
+    if (mpa->format == AF_FORMAT_S32 && af->data->format == AF_FORMAT_S24) {
+        size_t len = mp_audio_psize(mpa) / mpa->bps;
+        for (int s = 0; s < len; s++) {
+            uint32_t val = *((uint32_t *)mpa->planes[0] + s);
+            uint8_t *ptr = (uint8_t *)mpa->planes[0] + s * 3;
+            ptr[0] = val >> SHIFT24(0);
+            ptr[1] = val >> SHIFT24(1);
+            ptr[2] = val >> SHIFT24(2);
+        }
+        mp_audio_set_format(mpa, AF_FORMAT_S24);
     }
-    return false;
 }
 
-static void reorder_planes(struct mp_audio *mpa, int *reorder)
+// This relies on the tricky way mpa was allocated.
+static void reorder_planes(struct mp_audio *mpa, int *reorder,
+                           struct mp_chmap *newmap)
 {
     struct mp_audio prev = *mpa;
+    mp_audio_set_channels(mpa, newmap);
+
+    // The trailing planes were never written by avrctx, they're the NA channels.
+    int next_na = prev.num_planes;
+
     for (int n = 0; n < mpa->num_planes; n++) {
-        assert(reorder[n] >= 0 && reorder[n] < mpa->num_planes);
-        mpa->planes[n] = prev.planes[reorder[n]];
+        int src = reorder[n];
+        assert(src >= -1 && src < prev.num_planes);
+        if (src >= 0) {
+            mpa->planes[n] = prev.planes[src];
+        } else {
+            assert(next_na < mpa->num_planes);
+            mpa->planes[n] = prev.planes[next_na++];
+            af_fill_silence(mpa->planes[n], mpa->sstride * mpa->samples,
+                            mpa->format);
+        }
     }
 }
 
@@ -376,27 +460,17 @@ static int filter(struct af_instance *af, struct mp_audio *in)
 {
     struct af_resample *s = af->priv;
 
-    if (s->pending) {
-        if (s->pending->samples) {
-            af_add_output_frame(af, s->pending);
-        } else {
-            talloc_free(s->pending);
-        }
-        s->pending = NULL;
-    }
+    int samples = get_out_samples(s, in ? in->samples : 0);
 
-    int samples = avresample_available(s->avrctx) +
-        av_rescale_rnd(get_delay(s) + (in ? in->samples : 0),
-                       s->ctx.out_rate, s->ctx.in_rate, AV_ROUND_UP);
-    struct mp_audio *out = mp_audio_pool_get(af->out_pool, af->data, samples);
+    struct mp_audio out_format = s->pool_fmt;
+    struct mp_audio *out = mp_audio_pool_get(af->out_pool, &out_format, samples);
     if (!out)
         goto error;
     if (in)
         mp_audio_copy_attributes(out, in);
 
-    mp_audio_realloc_min(out, out->samples);
-
-    af->delay = get_delay(s) / (double)s->ctx.in_rate;
+    if (!s->avrctx)
+        goto error;
 
     if (out->samples) {
         out->samples = resample_frame(s->avrctx, out, in);
@@ -404,11 +478,15 @@ static int filter(struct af_instance *af, struct mp_audio *in)
             goto error;
     }
 
-    if (needs_reorder(s->reorder_out, out->nch)) {
-        if (AF_FORMAT_IS_PLANAR(out->format)) {
-            reorder_planes(out, s->reorder_out);
-        } else if (out->samples) {
-            struct mp_audio *new = mp_audio_pool_get(s->reorder_buffer, out,
+    struct mp_audio real_out = *out;
+    mp_audio_copy_config(out, &s->avrctx_fmt);
+
+    if (out->samples && !mp_audio_config_equals(out, &s->pre_out_fmt)) {
+        assert(af_fmt_is_planar(out->format) && out->format == real_out.format);
+        reorder_planes(out, s->reorder_out, &s->pool_fmt.channels);
+        if (!mp_audio_config_equals(out, &s->pre_out_fmt)) {
+            struct mp_audio *new = mp_audio_pool_get(s->reorder_buffer,
+                                                     &s->pre_out_fmt,
                                                      out->samples);
             if (!new)
                 goto error;
@@ -421,12 +499,17 @@ static int filter(struct af_instance *af, struct mp_audio *in)
         }
     }
 
+    extra_output_conversion(af, out);
+
     talloc_free(in);
     if (out->samples) {
         af_add_output_frame(af, out);
     } else {
         talloc_free(out);
     }
+
+    af->delay = get_delay(s);
+
     return 0;
 error:
     talloc_free(in);
@@ -445,17 +528,9 @@ static int af_open(struct af_instance *af)
     if (s->opts.cutoff <= 0.0)
         s->opts.cutoff = af_resample_default_cutoff(s->opts.filter_size);
 
-    s->avrctx = avresample_alloc_context();
-    s->avrctx_out = avresample_alloc_context();
     s->reorder_buffer = mp_audio_pool_create(s);
 
-    if (s->avrctx && s->avrctx_out) {
-        return AF_OK;
-    } else {
-        MP_ERR(af, "Cannot initialize Libavresample Context. \n");
-        uninit(af);
-        return AF_ERROR;
-    }
+    return AF_OK;
 }
 
 #define OPT_BASE_STRUCT struct af_resample
@@ -464,7 +539,6 @@ const struct af_info af_info_lavrresample = {
     .info = "Sample frequency conversion using libavresample",
     .name = "lavrresample",
     .open = af_open,
-    .test_conversion = test_conversion,
     .priv_size = sizeof(struct af_resample),
     .priv_defaults = &(const struct af_resample) {
         .opts = {

audio/filter/af_volume.c

@@ -30,18 +30,28 @@
 #include "demux/demux.h"
 
 struct priv {
+    float vol;                  // User-specified non-linear volume
     float level;                // User-specified gain level for each channel
     float rgain;                // Replaygain level
     int rgain_track;            // Enable/disable track based replaygain
     int rgain_album;            // Enable/disable album based replaygain
     float rgain_preamp;         // Set replaygain pre-amplification
     int rgain_clip;             // Enable/disable clipping prevention
+    float replaygain_fallback;
     int soft;                   // Enable/disable soft clipping
     int fast;                   // Use fix-point volume control
     int detach;                 // Detach if gain volume is neutral
     float cfg_volume;
 };
 
+// Convert to gain value from dB. input <= -200dB will become 0 gain.
+static float from_dB(float in, float k, float mi, float ma)
+{
+    if (in <= -200)
+        return 0.0;
+    return pow(10.0, MPCLAMP(in, mi, ma) / k);
+}
+
 static int control(struct af_instance *af, int cmd, void *arg)
 {
     struct priv *s = af->priv;
@@ -58,7 +68,7 @@ static int control(struct af_instance *af, int cmd, void *arg)
         } else {
             mp_audio_set_format(af->data, AF_FORMAT_FLOAT);
         }
-        if (AF_FORMAT_IS_PLANAR(in->format))
+        if (af_fmt_is_planar(in->format))
             mp_audio_set_format(af->data, af_fmt_to_planar(af->data->format));
         s->rgain = 1.0;
         if ((s->rgain_track || s->rgain_album) && af->replaygain_data) {
@@ -73,7 +83,7 @@ static int control(struct af_instance *af, int cmd, void *arg)
             }
 
             gain += s->rgain_preamp;
-            af_from_dB(1, &gain, &s->rgain, 20.0, -200.0, 60.0);
+            s->rgain = from_dB(gain, 20.0, -200.0, 60.0);
 
             MP_VERBOSE(af, "Applying replay-gain: %f\n", s->rgain);
 
@@ -81,16 +91,21 @@ static int control(struct af_instance *af, int cmd, void *arg)
                 s->rgain = MPMIN(s->rgain, 1.0 / peak);
                 MP_VERBOSE(af, "...with clipping prevention: %f\n", s->rgain);
             }
+        } else if (s->replaygain_fallback) {
+            s->rgain = from_dB(s->replaygain_fallback, 20.0, -200.0, 60.0);
+            MP_VERBOSE(af, "Applying fallback gain: %f\n", s->rgain);
         }
         if (s->detach && fabs(s->level * s->rgain - 1.0) < 0.00001)
             return AF_DETACH;
         return af_test_output(af, in);
     }
     case AF_CONTROL_SET_VOLUME:
-        s->level = *(float *)arg;
+        s->vol = *(float *)arg;
+        s->level = pow(s->vol, 3);
+        MP_VERBOSE(af, "volume gain: %f\n", s->level);
         return AF_OK;
     case AF_CONTROL_GET_VOLUME:
-        *(float *)arg = s->level;
+        *(float *)arg = s->vol;
         return AF_OK;
     }
     return AF_UNKNOWN;
@@ -143,7 +158,7 @@ static int af_open(struct af_instance *af)
     struct priv *s = af->priv;
     af->control = control;
     af->filter_frame = filter;
-    af_from_dB(1, &s->cfg_volume, &s->level, 20.0, -200.0, 60.0);
+    s->level = from_dB(s->cfg_volume, 20.0, -200.0, 60.0);
     return AF_OK;
 }
 
@@ -162,6 +177,7 @@ const struct af_info af_info_volume = {
         OPT_FLAG("replaygain-album", rgain_album, 0),
         OPT_FLOATRANGE("replaygain-preamp", rgain_preamp, 0, -15, 15),
         OPT_FLAG("replaygain-clip", rgain_clip, 0),
+        OPT_FLOATRANGE("replaygain-fallback", replaygain_fallback, 0, -200, 60),
         OPT_FLAG("softclip", soft, 0),
         OPT_FLAG("s16", fast, 0),
         OPT_FLAG("detach", detach, 0),

audio/filter/tools.c

@@ -21,24 +21,6 @@
 #include "common/common.h"
 #include "af.h"
 
-/* Convert to gain value from dB. Returns AF_OK if of and AF_ERROR if
- * fail. input <= -200dB will become 0 gain. */
-int af_from_dB(int n, float* in, float* out, float k, float mi, float ma)
-{
-  int i = 0;
-  // Sanity check
-  if(!in || !out)
-    return AF_ERROR;
-
-  for(i=0;i<n;i++){
-    if(in[i]<=-200)
-      out[i]=0.0;
-    else
-      out[i]=pow(10.0,MPCLAMP(in[i],mi,ma)/k);
-  }
-  return AF_OK;
-}
-
 /* Convert from ms to sample time */
 int af_from_ms(int n, float* in, int* out, int rate, float mi, float ma)
 {

audio/format.c

@@ -27,42 +27,68 @@
 #include "common/common.h"
 #include "audio/filter/af.h"
 
-int af_fmt2bps(int format)
+// number of bytes per sample, 0 if invalid/unknown
+int af_fmt_to_bytes(int format)
 {
-    switch (format & AF_FORMAT_BITS_MASK) {
-    case AF_FORMAT_8BIT:  return 1;
-    case AF_FORMAT_16BIT: return 2;
-    case AF_FORMAT_24BIT: return 3;
-    case AF_FORMAT_32BIT: return 4;
-    case AF_FORMAT_64BIT: return 8;
+    switch (af_fmt_from_planar(format)) {
+    case AF_FORMAT_U8:      return 1;
+    case AF_FORMAT_S16:     return 2;
+    case AF_FORMAT_S24:     return 3;
+    case AF_FORMAT_S32:     return 4;
+    case AF_FORMAT_FLOAT:   return 4;
+    case AF_FORMAT_DOUBLE:  return 8;
     }
+    if (af_fmt_is_spdif(format))
+        return 2;
     return 0;
 }
 
-int af_fmt2bits(int format)
+int af_fmt_change_bytes(int format, int bytes)
 {
-    return af_fmt2bps(format) * 8;
-}
-
-static int bits_to_mask(int bits)
-{
-    switch (bits) {
-    case 8:  return AF_FORMAT_8BIT;
-    case 16: return AF_FORMAT_16BIT;
-    case 24: return AF_FORMAT_24BIT;
-    case 32: return AF_FORMAT_32BIT;
-    case 64: return AF_FORMAT_64BIT;
-    }
-    return 0;
-}
-
-int af_fmt_change_bits(int format, int bits)
-{
-    if (!af_fmt_is_valid(format))
+    if (!af_fmt_is_valid(format) || !bytes)
         return 0;
-    int mask = bits_to_mask(bits);
-    format = (format & ~AF_FORMAT_BITS_MASK) | mask;
-    return af_fmt_is_valid(format) ? format : 0;
+    for (int fmt = 1; fmt < AF_FORMAT_COUNT; fmt++) {
+        if (af_fmt_to_bytes(fmt) == bytes &&
+            af_fmt_is_float(fmt) == af_fmt_is_float(format) &&
+            af_fmt_is_planar(fmt) == af_fmt_is_planar(format) &&
+            af_fmt_is_spdif(fmt) == af_fmt_is_spdif(format))
+            return fmt;
+    }
+    return 0;
+}
+
+// All formats are considered signed, except explicitly unsigned int formats.
+bool af_fmt_is_unsigned(int format)
+{
+    return format == AF_FORMAT_U8 || format == AF_FORMAT_U8P;
+}
+
+bool af_fmt_is_float(int format)
+{
+    format = af_fmt_from_planar(format);
+    return format == AF_FORMAT_FLOAT || format == AF_FORMAT_DOUBLE;
+}
+
+// true for both unsigned and signed ints
+bool af_fmt_is_int(int format)
+{
+    return format && !af_fmt_is_spdif(format) && !af_fmt_is_float(format);
+}
+
+// false for interleaved and AF_FORMAT_UNKNOWN
+bool af_fmt_is_planar(int format)
+{
+    return format && af_fmt_to_planar(format) == format;
+}
+
+bool af_fmt_is_spdif(int format)
+{
+    return af_format_sample_alignment(format) > 1;
+}
+
+bool af_fmt_is_pcm(int format)
+{
+    return af_fmt_is_valid(format) && !af_fmt_is_spdif(format);
 }
 
 static const int planar_formats[][2] = {
@@ -99,58 +125,40 @@ int af_fmt_from_planar(int format)
     return format;
 }
 
-const struct af_fmt_entry af_fmtstr_table[] = {
-    {"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},
-
-    {"u8p",         AF_FORMAT_U8P},
-    {"s16p",        AF_FORMAT_S16P},
-    {"s32p",        AF_FORMAT_S32P},
-    {"floatp",      AF_FORMAT_FLOATP},
-    {"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}
-};
-
 bool af_fmt_is_valid(int format)
 {
-    for (int i = 0; af_fmtstr_table[i].name; i++) {
-        if (af_fmtstr_table[i].format == format)
-            return true;
-    }
-    return false;
+    return format > 0 && format < AF_FORMAT_COUNT;
 }
 
 const char *af_fmt_to_str(int format)
 {
-    for (int i = 0; af_fmtstr_table[i].name; i++) {
-        if (af_fmtstr_table[i].format == format)
-            return af_fmtstr_table[i].name;
+    switch (format) {
+    case AF_FORMAT_U8:          return "u8";
+    case AF_FORMAT_S16:         return "s16";
+    case AF_FORMAT_S24:         return "s24";
+    case AF_FORMAT_S32:         return "s32";
+    case AF_FORMAT_FLOAT:       return "float";
+    case AF_FORMAT_DOUBLE:      return "double";
+    case AF_FORMAT_U8P:         return "u8p";
+    case AF_FORMAT_S16P:        return "s16p";
+    case AF_FORMAT_S32P:        return "s32p";
+    case AF_FORMAT_FLOATP:      return "floatp";
+    case AF_FORMAT_DOUBLEP:     return "doublep";
+    case AF_FORMAT_S_AAC:       return "spdif-aac";
+    case AF_FORMAT_S_AC3:       return "spdif-ac3";
+    case AF_FORMAT_S_DTS:       return "spdif-dts";
+    case AF_FORMAT_S_DTSHD:     return "spdif-dtshd";
+    case AF_FORMAT_S_EAC3:      return "spdif-eac3";
+    case AF_FORMAT_S_MP3:       return "spdif-mp3";
+    case AF_FORMAT_S_TRUEHD:    return "spdif-truehd";
     }
-
     return "??";
 }
 
 int af_fmt_seconds_to_bytes(int format, float seconds, int channels, int samplerate)
 {
-    assert(!AF_FORMAT_IS_PLANAR(format));
-    int bps      = af_fmt2bps(format);
+    assert(!af_fmt_is_planar(format));
+    int bps      = af_fmt_to_bytes(format);
     int framelen = channels * bps;
     int bytes    = seconds  * bps * samplerate;
     if (bytes % framelen)
@@ -158,19 +166,9 @@ int af_fmt_seconds_to_bytes(int format, float seconds, int channels, int sampler
     return bytes;
 }
 
-int af_str2fmt_short(bstr str)
-{
-    for (int i = 0; af_fmtstr_table[i].name; i++) {
-        if (!bstrcasecmp0(str, af_fmtstr_table[i].name))
-            return af_fmtstr_table[i].format;
-    }
-    return 0;
-}
-
 void af_fill_silence(void *dst, size_t bytes, int format)
 {
-    bool us = (format & AF_FORMAT_SIGN_MASK) == AF_FORMAT_US;
-    memset(dst, us ? 0x80 : 0, bytes);
+    memset(dst, af_fmt_is_unsigned(format) ? 0x80 : 0, bytes);
 }
 
 #define FMT_DIFF(type, a, b) (((a) & type) - ((b) & type))
@@ -186,33 +184,30 @@ int af_format_conversion_score(int dst_format, int src_format)
     if (dst_format == src_format)
         return 1024;
     // Can't be normally converted
-    if (AF_FORMAT_IS_SPECIAL(dst_format) || AF_FORMAT_IS_SPECIAL(src_format))
+    if (!af_fmt_is_pcm(dst_format) || !af_fmt_is_pcm(src_format))
         return INT_MIN;
     int score = 1024;
-    if (FMT_DIFF(AF_FORMAT_INTERLEAVING_MASK, dst_format, src_format))
+    if (af_fmt_is_planar(dst_format) != af_fmt_is_planar(src_format))
         score -= 1;     // has to (de-)planarize
-    if (FMT_DIFF(AF_FORMAT_SIGN_MASK, dst_format, src_format))
-        score -= 4;     // has to swap sign
-    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_TYPE_MASK) == AF_FORMAT_F) {
+    if (af_fmt_is_float(dst_format) != af_fmt_is_float(src_format)) {
+        int dst_bytes = af_fmt_to_bytes(dst_format);
+        if (af_fmt_is_float(dst_format)) {
             // For int->float, always prefer 32 bit float.
-            score -= dst_bits == AF_FORMAT_32BIT ? 8 : 0;
+            score -= dst_bytes == 4 ? 1 : 0;
         } else {
             // For float->int, always prefer highest bit depth int
-            score -= 8 * (AF_FORMAT_64BIT - dst_bits);
+            score -= 8 - dst_bytes;
         }
+        // Has to convert float<->int - Consider this the worst case.
+        score -= 2048;
     } else {
-        int bits = FMT_DIFF(AF_FORMAT_BITS_MASK, dst_format, src_format);
-        if (bits > 0) {
-            score -= 8 * bits;          // has to add padding
-        } else if (bits < 0) {
-            score -= 1024 - 8 * bits;   // has to reduce bit depth
+        int bytes = af_fmt_to_bytes(dst_format) - af_fmt_to_bytes(src_format);
+        if (bytes > 0) {
+            score -= bytes;             // has to add padding
+        } else if (bytes < 0) {
+            score -= 1024 - bytes;      // has to reduce bit depth
         }
     }
-    // Consider this the worst case.
-    if (FMT_DIFF(AF_FORMAT_TYPE_MASK, dst_format, src_format))
-        score -= 2048;  // has to convert float<->int
     return score;
 }
 

audio/format.h

@@ -22,96 +22,50 @@
 #ifndef MPLAYER_AF_FORMAT_H
 #define MPLAYER_AF_FORMAT_H
 
+#include <stddef.h>
 #include <stdbool.h>
 
-#include "misc/bstr.h"
-
-// Signed/unsigned
-#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<<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)
-
-// 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<<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_UNKNOWN = 0,
 
-    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     = (AF_FORMAT_F|AF_FORMAT_32BIT),
-    AF_FORMAT_DOUBLE    = (AF_FORMAT_F|AF_FORMAT_64BIT),
+    AF_FORMAT_U8,
+    AF_FORMAT_S16,
+    AF_FORMAT_S24,
+    AF_FORMAT_S32,
+    AF_FORMAT_FLOAT,
+    AF_FORMAT_DOUBLE,
 
     // Planar variants
-    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),
+    AF_FORMAT_U8P,
+    AF_FORMAT_S16P,
+    AF_FORMAT_S32P,
+    AF_FORMAT_FLOATP,
+    AF_FORMAT_DOUBLEP,
 
     // 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),
+    AF_FORMAT_S_AAC,
+    AF_FORMAT_S_AC3,
+    AF_FORMAT_S_DTS,
+    AF_FORMAT_S_DTSHD,
+    AF_FORMAT_S_EAC3,
+    AF_FORMAT_S_MP3,
+    AF_FORMAT_S_TRUEHD,
+
+    AF_FORMAT_COUNT
 };
 
-#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)
-#define AF_FORMAT_IS_FLOAT(f) (!!((f) & AF_FORMAT_F))
-// false for interleaved and AF_FORMAT_UNKNOWN
-#define AF_FORMAT_IS_PLANAR(f) (!!((f) & AF_FORMAT_PLANAR))
-
-struct af_fmt_entry {
-    const char *name;
-    int format;
-};
-
-extern const struct af_fmt_entry af_fmtstr_table[];
-
-int af_str2fmt_short(bstr str);
 const char *af_fmt_to_str(int format);
 
-int af_fmt2bps(int format);
-int af_fmt2bits(int format);
-int af_fmt_change_bits(int format, int bits);
+int af_fmt_to_bytes(int format);
+int af_fmt_change_bytes(int format, int bytes);
+
+bool af_fmt_is_valid(int format);
+bool af_fmt_is_unsigned(int format);
+bool af_fmt_is_float(int format);
+bool af_fmt_is_int(int format);
+bool af_fmt_is_planar(int format);
+bool af_fmt_is_spdif(int format);
+bool af_fmt_is_pcm(int format);
 
 int af_fmt_to_planar(int format);
 int af_fmt_from_planar(int format);
@@ -119,8 +73,6 @@ int af_fmt_from_planar(int format);
 // Amount of bytes that contain audio of the given duration, aligned to frames.
 int af_fmt_seconds_to_bytes(int format, float seconds, int channels, int samplerate);
 
-bool af_fmt_is_valid(int format);
-
 void af_fill_silence(void *dst, size_t bytes, int format);
 
 int af_format_conversion_score(int dst_format, int src_format);

audio/mixer.c

@@ -69,10 +69,10 @@ bool mixer_audio_initialized(struct mixer *mixer)
     return !!mixer->ao;
 }
 
-float mixer_getneutralvolume(struct mixer *mixer)
+float mixer_getmaxvolume(struct mixer *mixer)
 {
     // gain == 1
-    return mixer->softvol ? 1.0 / mixer->opts->softvol_max * 100.0 * 100.0 : 100;
+    return mixer->softvol ? mixer->opts->softvol_max : 100;
 }
 
 static void checkvolume(struct mixer *mixer)
@@ -85,8 +85,8 @@ static void checkvolume(struct mixer *mixer)
         float gain;
         if (!af_control_any_rev(mixer->af, AF_CONTROL_GET_VOLUME, &gain))
             gain = 1.0;
-        vol.left = (gain / (mixer->opts->softvol_max / 100.0)) * 100.0;
-        vol.right = (gain / (mixer->opts->softvol_max / 100.0)) * 100.0;
+        vol.left = gain * 100.0;
+        vol.right = gain * 100.0;
     } else {
         MP_DBG(mixer, "Reading volume from AO.\n");
         // Rely on the values not changing if the query is not supported
@@ -122,14 +122,14 @@ void mixer_getvolume(struct mixer *mixer, float *l, float *r)
 
 static void setvolume_internal(struct mixer *mixer, float l, float r)
 {
-    struct ao_control_vol vol = {.left = l, .right = r};
     if (!mixer->softvol) {
         MP_DBG(mixer, "Setting volume on AO.\n");
+        struct ao_control_vol vol = {.left = l, .right = r};
         if (ao_control(mixer->ao, AOCONTROL_SET_VOLUME, &vol) != CONTROL_OK)
             MP_ERR(mixer, "Failed to change audio output volume.\n");
         return;
     }
-    float gain = (l + r) / 2.0 / 100.0 * mixer->opts->softvol_max / 100.0;
+    float gain = (l + r) / 2.0 / 100.0;
     if (!af_control_any_rev(mixer->af, AF_CONTROL_SET_VOLUME, &gain)) {
         if (gain == 1.0)
             return;
@@ -144,8 +144,9 @@ void mixer_setvolume(struct mixer *mixer, float l, float r)
 {
     checkvolume(mixer);  // to check mute status
 
-    mixer->vol_l = av_clipf(l, 0, 100);
-    mixer->vol_r = av_clipf(r, 0, 100);
+    float max = mixer_getmaxvolume(mixer);
+    mixer->vol_l = MPCLAMP(l, 0, max);
+    mixer->vol_r = MPCLAMP(r, 0, max);
     if (mixer->ao && !(mixer->emulate_mute && mixer->muted))
         setvolume_internal(mixer, mixer->vol_l, mixer->vol_r);
 }
@@ -226,7 +227,7 @@ void mixer_setbalance(struct mixer *mixer, float val)
         return;
     }
 
-    /* make all other channels pass thru since by default pan blocks all */
+    /* make all other channels pass through since by default pan blocks all */
     for (int i = 2; i < AF_NCH; i++) {
         float level[AF_NCH] = {0};
         level[i] = 1.f;
@@ -242,9 +243,8 @@ char *mixer_get_volume_restore_data(struct mixer *mixer)
 {
     if (!mixer->driver[0])
         return NULL;
-    return talloc_asprintf(NULL, "%s:%f:%f:%d:%f", mixer->driver, mixer->vol_l,
-                           mixer->vol_r, mixer->muted_by_us,
-                           mixer->opts->softvol_max);
+    return talloc_asprintf(NULL, "%s:%f:%f:%d", mixer->driver, mixer->vol_l,
+                           mixer->vol_r, mixer->muted_by_us);
 }
 
 static void probe_softvol(struct mixer *mixer)
@@ -315,11 +315,10 @@ static void restore_volume(struct mixer *mixer)
     char *data = mixer->opts->mixer_restore_volume_data;
     if (!mixer->persistent_volume && data && data[0]) {
         char drv[40];
-        float v_l, v_r, s;
+        float v_l, v_r;
         int m;
-        if (sscanf(data, "%39[^:]:%f:%f:%d:%f", drv, &v_l, &v_r, &m, &s) == 5) {
-            float diff = fabs(mixer->opts->softvol_max - s);
-            if (strcmp(mixer->driver, drv) == 0 && diff < 0.01) {
+        if (sscanf(data, "%39[^:]:%f:%f:%d", drv, &v_l, &v_r, &m) == 4) {
+            if (strcmp(mixer->driver, drv) == 0) {
                 force_vol_l = v_l;
                 force_vol_r = v_r;
                 force_mute = !!m;

audio/mixer.h

@@ -44,7 +44,7 @@ void mixer_setmute(struct mixer *mixer, bool mute);
 bool mixer_getmute(struct mixer *mixer);
 void mixer_getbalance(struct mixer *mixer, float *bal);
 void mixer_setbalance(struct mixer *mixer, float bal);
-float mixer_getneutralvolume(struct mixer *mixer);
+float mixer_getmaxvolume(struct mixer *mixer);
 char *mixer_get_volume_restore_data(struct mixer *mixer);
 
 #endif /* MPLAYER_MIXER_H */

audio/out/ao.c

@@ -191,14 +191,14 @@ static struct ao *ao_init(bool probing, struct mpv_global *global,
             snprintf(rdevice, sizeof(rdevice), "%s", ao->device ? ao->device : "");
             talloc_free(ao);
             return ao_init(probing, global, input_ctx, encode_lavc_ctx,
-                           samplerate, format, channels, rdevice, redirect, args);
+                           samplerate, format, channels, rdevice, redirect, NULL);
         }
         goto fail;
     }
 
-    ao->sstride = af_fmt2bps(ao->format);
+    ao->sstride = af_fmt_to_bytes(ao->format);
     ao->num_planes = 1;
-    if (AF_FORMAT_IS_PLANAR(ao->format)) {
+    if (af_fmt_is_planar(ao->format)) {
         ao->num_planes = ao->channels.num;
     } else {
         ao->sstride *= ao->channels.num;
@@ -307,7 +307,8 @@ done:
 // Uninitialize and destroy the AO. Remaining audio must be dropped.
 void ao_uninit(struct ao *ao)
 {
-    ao->api->uninit(ao);
+    if (ao)
+        ao->api->uninit(ao);
     talloc_free(ao);
 }
 
@@ -385,28 +386,26 @@ bool ao_eof_reached(struct ao *ao)
 // 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;
-    if (atomic_load(&ao->request_hotplug))
-        actual_events |= AO_EVENT_HOTPLUG;
-    return actual_events & events;
+    return atomic_fetch_and(&ao->events_, ~(unsigned)events) & events;
+}
+
+static void ao_add_events(struct ao *ao, int events)
+{
+    atomic_fetch_or(&ao->events_, events);
+    if (ao->input_ctx)
+        mp_input_wakeup(ao->input_ctx);
 }
 
 // Request that the player core destroys and recreates the AO. Fully thread-safe.
 void ao_request_reload(struct ao *ao)
 {
-    atomic_store(&ao->request_reload, true);
-    if (ao->input_ctx)
-        mp_input_wakeup(ao->input_ctx);
+    ao_add_events(ao, AO_EVENT_RELOAD);
 }
 
 // Notify the player that the device list changed. Fully thread-safe.
 void ao_hotplug_event(struct ao *ao)
 {
-    atomic_store(&ao->request_hotplug, true);
-    if (ao->input_ctx)
-        mp_input_wakeup(ao->input_ctx);
+    ao_add_events(ao, AO_EVENT_HOTPLUG);
 }
 
 bool ao_chmap_sel_adjust(struct ao *ao, const struct mp_chmap_sel *s,
@@ -415,12 +414,15 @@ bool ao_chmap_sel_adjust(struct ao *ao, const struct mp_chmap_sel *s,
     if (mp_msg_test(ao->log, MSGL_DEBUG)) {
         for (int i = 0; i < s->num_chmaps; i++) {
             struct mp_chmap c = s->chmaps[i];
-            MP_DBG(ao, "chmap_sel #%d: %s\n", i, mp_chmap_to_str(&c));
+            MP_DBG(ao, "chmap_sel #%d: %s (%s)\n", i, mp_chmap_to_str(&c),
+                   mp_chmap_to_str_hr(&c));
         }
     }
     bool r = mp_chmap_sel_adjust(s, map);
-    if (r)
-        MP_DBG(ao, "result: %s\n", mp_chmap_to_str(map));
+    if (r) {
+        MP_DBG(ao, "result: %s (%s)\n", mp_chmap_to_str(map),
+               mp_chmap_to_str_hr(map));
+    }
     return r;
 }
 
@@ -496,7 +498,6 @@ bool ao_hotplug_check_update(struct ao_hotplug *hp)
 {
     if (hp->ao && ao_query_and_reset_events(hp->ao, AO_EVENT_HOTPLUG)) {
         hp->needs_update = true;
-        atomic_store(&hp->ao->request_hotplug, false);
         return true;
     }
     return false;

audio/out/ao_alsa.c

@@ -101,7 +101,7 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
         long get_vol, set_vol;
         float f_multi;
 
-        if (AF_FORMAT_IS_SPECIAL(ao->format))
+        if (!af_fmt_is_pcm(ao->format))
             return CONTROL_FALSE;
 
         snd_mixer_selem_id_alloca(&sid);
@@ -204,14 +204,9 @@ 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,         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},
@@ -251,7 +246,10 @@ static const int alsa_to_mp_channels[][2] = {
     {SND_CHMAP_TRL,     MP_SP(TBL)},
     {SND_CHMAP_TRR,     MP_SP(TBR)},
     {SND_CHMAP_TRC,     MP_SP(TBC)},
+    {SND_CHMAP_RRC,     MP_SP(SDR)},
+    {SND_CHMAP_RLC,     MP_SP(SDL)},
     {SND_CHMAP_MONO,    MP_SP(FC)},
+    {SND_CHMAP_NA,      MP_SPEAKER_ID_NA},
     {SND_CHMAP_LAST,    MP_SPEAKER_ID_COUNT}
 };
 
@@ -275,6 +273,20 @@ static int find_alsa_channel(int mp_channel)
     return SND_CHMAP_UNKNOWN;
 }
 
+static int mp_chmap_from_alsa(struct mp_chmap *dst, snd_pcm_chmap_t *src)
+{
+    *dst = (struct mp_chmap) {0};
+
+    if (src->channels > MP_NUM_CHANNELS)
+        return -1;
+
+    dst->num = src->channels;
+    for (int c = 0; c < dst->num; c++)
+        dst->speaker[c] = find_mp_channel(src->pos[c]);
+
+    return 0;
+}
+
 static bool query_chmaps(struct ao *ao, struct mp_chmap *chmap)
 {
     struct priv *p = ao->priv;
@@ -285,16 +297,12 @@ static bool query_chmaps(struct ao *ao, struct mp_chmap *chmap)
         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]);
+        struct mp_chmap entry;
+        mp_chmap_from_alsa(&entry, &maps[i]->map);
 
         if (mp_chmap_is_valid(&entry)) {
+            if (maps[i]->type == SND_CHMAP_TYPE_VAR)
+                mp_chmap_reorder_norm(&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));
@@ -366,8 +374,9 @@ static char *append_params(void *ta_parent, const char *device, const char *p)
 static int try_open_device(struct ao *ao, const char *device)
 {
     struct priv *p = ao->priv;
+    int err;
 
-    if (AF_FORMAT_IS_IEC61937(ao->format)) {
+    if (af_fmt_is_spdif(ao->format)) {
         void *tmp = talloc_new(NULL);
         char *params = talloc_asprintf(tmp,
                         "AES0=%d,AES1=%d,AES2=0,AES3=%d",
@@ -376,15 +385,29 @@ static int try_open_device(struct ao *ao, const char *device)
                         map_iec958_srate(ao->samplerate));
         const char *ac3_device = append_params(tmp, device, params);
         MP_VERBOSE(ao, "opening device '%s' => '%s'\n", device, ac3_device);
-        int err = snd_pcm_open
-                    (&p->alsa, ac3_device, SND_PCM_STREAM_PLAYBACK, 0);
+        err = snd_pcm_open(&p->alsa, ac3_device, SND_PCM_STREAM_PLAYBACK, 0);
+        if (err < 0) {
+            // Some spdif-capable devices do not accept the AES0 parameter,
+            // and instead require the iec958 pseudo-device (they will play
+            // noise otherwise). Unfortunately, ALSA gives us no way to map
+            // these devices, so try it for the default device only.
+            bstr dev;
+            bstr_split_tok(bstr0(device), ":", &dev, &(bstr){0});
+            if (bstr_equals0(dev, "default")) {
+                ac3_device = append_params(tmp, "iec958", params);
+                MP_VERBOSE(ao, "got error %d; opening iec fallback device '%s'\n",
+                           err, ac3_device);
+                err = snd_pcm_open
+                            (&p->alsa, ac3_device, SND_PCM_STREAM_PLAYBACK, 0);
+            }
+        }
         talloc_free(tmp);
-        if (!err)
-            return 0;
+    } else {
+        MP_VERBOSE(ao, "opening device '%s'\n", device);
+        err = snd_pcm_open(&p->alsa, device, SND_PCM_STREAM_PLAYBACK, 0);
     }
 
-    MP_VERBOSE(ao, "opening device '%s'\n", device);
-    return snd_pcm_open(&p->alsa, device, SND_PCM_STREAM_PLAYBACK, 0);
+    return err;
 }
 
 static void uninit(struct ao *ao)
@@ -404,14 +427,12 @@ alsa_error: ;
 #define INIT_OK 0
 #define INIT_ERROR -1
 #define INIT_BRAINDEATH -2
-static int init_device(struct ao *ao)
+static int init_device(struct ao *ao, bool second_try)
 {
     struct priv *p = ao->priv;
     int err;
 
     const char *device = "default";
-    if (AF_FORMAT_IS_IEC61937(ao->format))
-        device = "iec958";
     if (ao->device)
         device = ao->device;
     if (p->cfg_device && p->cfg_device[0])
@@ -432,7 +453,7 @@ static int init_device(struct ao *ao)
     err = snd_pcm_hw_params_any(p->alsa, alsa_hwparams);
     CHECK_ALSA_ERROR("Unable to get initial parameters");
 
-    if (AF_FORMAT_IS_IEC61937(ao->format)) {
+    if (af_fmt_is_spdif(ao->format)) {
         if (ao->format == AF_FORMAT_S_MP3) {
             p->alsa_fmt = SND_PCM_FORMAT_MPEG;
         } else {
@@ -448,7 +469,7 @@ static int init_device(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))
+        if (af_fmt_is_spdif(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));
@@ -459,11 +480,11 @@ static int init_device(struct ao *ao)
     err = snd_pcm_hw_params_set_format(p->alsa, alsa_hwparams, p->alsa_fmt);
     CHECK_ALSA_ERROR("Unable to set format");
 
-    snd_pcm_access_t access = AF_FORMAT_IS_PLANAR(ao->format)
+    snd_pcm_access_t access = af_fmt_is_planar(ao->format)
                                     ? SND_PCM_ACCESS_RW_NONINTERLEAVED
                                     : SND_PCM_ACCESS_RW_INTERLEAVED;
     err = snd_pcm_hw_params_set_access(p->alsa, alsa_hwparams, access);
-    if (err < 0 && AF_FORMAT_IS_PLANAR(ao->format)) {
+    if (err < 0 && af_fmt_is_planar(ao->format)) {
         ao->format = af_fmt_from_planar(ao->format);
         access = SND_PCM_ACCESS_RW_INTERLEAVED;
         err = snd_pcm_hw_params_set_access(p->alsa, alsa_hwparams, access);
@@ -471,8 +492,11 @@ static int init_device(struct ao *ao)
     CHECK_ALSA_ERROR("Unable to set access type");
 
     struct mp_chmap dev_chmap = ao->channels;
-    if (AF_FORMAT_IS_IEC61937(ao->format) || p->cfg_ignore_chmap) {
+    if (af_fmt_is_spdif(ao->format) || p->cfg_ignore_chmap) {
         dev_chmap.num = 0; // disable chmap API
+    } else if (dev_chmap.num == 1 && dev_chmap.speaker[0] == MP_SPEAKER_ID_FC) {
+        // As yet another ALSA API inconsistency, mono is not reported correctly.
+        dev_chmap.num = 0;
     } else if (query_chmaps(ao, &dev_chmap)) {
         ao->channels = dev_chmap;
     } else {
@@ -492,11 +516,12 @@ static int init_device(struct ao *ao)
     }
 
     if (num_channels != ao->channels.num) {
+        int req = ao->channels.num;
         mp_chmap_from_channels_alsa(&ao->channels, num_channels);
         if (!mp_chmap_is_valid(&ao->channels))
             mp_chmap_from_channels(&ao->channels, 2);
-        MP_ERR(ao, "Couldn't get requested number of channels, fallback to %s.\n",
-               mp_chmap_to_str(&ao->channels));
+        MP_ERR(ao, "Asked for %d channels, got %d - fallback to %s.\n", req,
+               num_channels, mp_chmap_to_str(&ao->channels));
     }
 
     // Some ALSA drivers have broken delay reporting, so disable the ALSA
@@ -546,10 +571,11 @@ static int init_device(struct ao *ao)
 
         err = snd_pcm_set_chmap(p->alsa, alsa_chmap);
         if (err == -ENXIO) {
-            // I consider this an ALSA bug: the channel map was reported as
-            // supported, but we still can't set it. It happens virtually
-            // always with dmix, though.
-            MP_VERBOSE(ao, "Device does not support requested channel map (%s)\n",
+            // A device my not be able to set any channel map, even channel maps
+            // that were reported as supported. This is either because the ALSA
+            // device is broken (dmix), or because the driver has only 1
+            // channel map per channel count, and setting the map is not needed.
+            MP_VERBOSE(ao, "device returned ENXIO when setting channel map %s\n",
                        mp_chmap_to_str(&dev_chmap));
         } else {
             CHECK_ALSA_WARN("Channel map setup failed");
@@ -564,17 +590,45 @@ static int init_device(struct ao *ao)
         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]);
+        struct mp_chmap chmap;
+        mp_chmap_from_alsa(&chmap, alsa_chmap);
 
         MP_VERBOSE(ao, "which we understand as: %s\n", mp_chmap_to_str(&chmap));
 
         if (p->cfg_ignore_chmap) {
             MP_VERBOSE(ao, "user set ignore-chmap; ignoring the channel map.\n");
-        } else if (AF_FORMAT_IS_IEC61937(ao->format)) {
+        } else if (af_fmt_is_spdif(ao->format)) {
             MP_VERBOSE(ao, "using spdif passthrough; ignoring the channel map.\n");
         } else if (mp_chmap_is_valid(&chmap)) {
+            // Is it one that contains NA channels?
+            struct mp_chmap without_na = chmap;
+            mp_chmap_remove_na(&without_na);
+
+            if (mp_chmap_is_valid(&without_na) &&
+                !mp_chmap_equals(&without_na, &chmap) &&
+                !mp_chmap_equals(&chmap, &ao->channels) &&
+                !second_try)
+            {
+                // Sometimes, ALSA will advertise certain chmaps, but it's not
+                // possible to set them. This can happen with dmix: as of
+                // alsa 1.0.28, dmix can do stereo only, but advertises the
+                // surround chmaps of the underlying device. In this case,
+                // e.g. setting 6 channels will succeed, but requesting  5.1
+                // afterwards will fail. Then it will return something like
+                // "FL FR NA NA NA NA" as channel map. This means we would
+                // have to pad stereo output to 6 channels with silence, which
+                // would require lots of extra processing. You can't change
+                // the number of channels to 2 either, because the hw params
+                // are already set! So just fuck it and reopen the device with
+                // the chmap "cleaned out" of NA entries.
+                MP_VERBOSE(ao, "Working around braindead ALSA behavior.\n");
+                err = snd_pcm_close(p->alsa);
+                p->alsa = NULL;
+                CHECK_ALSA_ERROR("pcm close error");
+                ao->channels = without_na;
+                return INIT_BRAINDEATH;
+            }
+
             if (mp_chmap_equals(&chmap, &ao->channels)) {
                 MP_VERBOSE(ao, "which is what we requested.\n");
             } else if (chmap.num == ao->channels.num) {
@@ -584,41 +638,14 @@ static int init_device(struct ao *ao)
                 MP_WARN(ao, "ALSA channel map conflicts with channel count!\n");
             }
         } else {
-            // Is it one that contains NA channels?
-            struct mp_chmap chmap2 = {0};
-            for (int c = 0; c < alsa_chmap->channels; c++) {
-                int alsa_ch = alsa_chmap->pos[c];
-                if (alsa_ch != SND_CHMAP_NA)
-                    chmap2.speaker[chmap2.num++] = find_mp_channel(alsa_ch);
-            }
-
-            if (mp_chmap_is_valid(&chmap2)) {
-                // Sometimes, ALSA will advertise certain chmaps, but it's not
-                // possible to set them. This can happen with dmix: as of
-                // alsa 1.0.28, dmix can do stereo only, but advertises the
-                // surround chmaps of the underlying device. In this case,
-                // requesting e.g. 5.1 will fail, but it will still allow
-                // setting 6 channels. Then it will return something like
-                // "FL FR NA NA NA NA" as channel map. This means we would
-                // have to pad stereo output to 6 channels with silence, which
-                // is way too complicated in the general case. You can't change
-                // the number of channels to 2 either, because the hw params
-                // are already set! So just fuck it and reopen the device with
-                // the chmap "cleaned out" of NA entries.
-                MP_VERBOSE(ao, "Working around braindead ALSA behavior.\n");
-                err = snd_pcm_close(p->alsa);
-                p->alsa = NULL;
-                CHECK_ALSA_ERROR("pcm close error");
-                ao->channels = chmap2;
-                return INIT_BRAINDEATH;
-            }
-
             MP_WARN(ao, "Got unknown channel map from ALSA.\n");
         }
 
         // mpv and ALSA use different conventions for mono
-        if (ao->channels.num == 1)
+        if (ao->channels.num == 1) {
+            MP_VERBOSE(ao, "assuming we actually got MONO from ALSA.\n");
             ao->channels.speaker[0] = MP_SP(FC);
+        }
 
         free(alsa_chmap);
     }
@@ -683,9 +710,9 @@ static int init(struct ao *ao)
 
     MP_VERBOSE(ao, "using ALSA version: %s\n", snd_asoundlib_version());
 
-    int r = init_device(ao);
+    int r = init_device(ao, false);
     if (r == INIT_BRAINDEATH)
-        r = init_device(ao); // retry with normalized channel layout
+        r = init_device(ao, true); // retry with normalized channel layout
     return r == INIT_OK ? 0 : -1;
 }
 
@@ -822,7 +849,7 @@ static int play(struct ao *ao, void **data, int samples, int flags)
         return 0;
 
     do {
-        if (AF_FORMAT_IS_PLANAR(ao->format)) {
+        if (af_fmt_is_planar(ao->format)) {
             res = snd_pcm_writen(p->alsa, data, samples);
         } else {
             res = snd_pcm_writei(p->alsa, data[0], samples);
@@ -896,15 +923,18 @@ static void list_devs(struct ao *ao, struct ao_device_list *list)
         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] = '/';
+        if (io && strcmp(io, "Output") == 0) {
+            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});
         }
-        ao_device_list_add(list, ao, &(struct ao_device_desc){name, desc2});
+        free(name);
+        free(desc);
+        free(io);
     }
 
     snd_device_name_free_hint(hints);

audio/out/ao_coreaudio.c

@@ -25,18 +25,22 @@
 #include "options/m_option.h"
 #include "misc/ring.h"
 #include "common/msg.h"
-#include "audio/out/ao_coreaudio_properties.h"
-#include "audio/out/ao_coreaudio_utils.h"
+#include "ao_coreaudio_chmap.h"
+#include "ao_coreaudio_properties.h"
+#include "ao_coreaudio_utils.h"
 
 struct priv {
     AudioDeviceID device;
     AudioUnit audio_unit;
 
     uint64_t hw_latency_us;
-};
 
-bool ca_layout_to_mp_chmap(struct ao *ao, AudioChannelLayout *layout,
-                           struct mp_chmap *chmap);
+    AudioStreamBasicDescription original_asbd;
+    AudioStreamID original_asbd_stream;
+
+    int change_physical_format;
+    int exclusive;
+};
 
 static int64_t ca_get_hardware_latency(struct ao *ao) {
     struct priv *p = ao->priv;
@@ -52,12 +56,8 @@ static int64_t ca_get_hardware_latency(struct ao *ao) {
             &size);
     CHECK_CA_ERROR("cannot get audio unit latency");
 
-    uint32_t frames = 0;
-    err = CA_GET_O(p->device, kAudioDevicePropertyLatency, &frames);
-    CHECK_CA_ERROR("cannot get device latency");
-
     uint64_t audiounit_latency_us = audiounit_latency_sec * 1e6;
-    uint64_t device_latency_us    = ca_frames_to_us(ao, frames);
+    uint64_t device_latency_us    = ca_get_device_latency_us(ao, p->device);
 
     MP_VERBOSE(ao, "audiounit latency [us]: %lld\n", audiounit_latency_us);
     MP_VERBOSE(ao, "device latency [us]: %lld\n", device_latency_us);
@@ -74,12 +74,14 @@ static OSStatus render_cb_lpcm(void *ctx, AudioUnitRenderActionFlags *aflags,
 {
     struct ao *ao   = ctx;
     struct priv *p  = ao->priv;
-    AudioBuffer buf = buffer_list->mBuffers[0];
+    void *planes[MP_NUM_CHANNELS] = {0};
+
+    for (int n = 0; n < ao->num_planes; n++)
+        planes[n] = buffer_list->mBuffers[n].mData;
 
     int64_t end = mp_time_us();
     end += p->hw_latency_us + ca_get_latency(ts) + ca_frames_to_us(ao, frames);
-
-    ao_read_data(ao, &buf.mData, frames, end);
+    ao_read_data(ao, planes, frames, end);
     return noErr;
 }
 
@@ -122,8 +124,8 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
     return CONTROL_UNKNOWN;
 }
 
-static bool init_chmap(struct ao *ao);
 static bool init_audiounit(struct ao *ao, AudioStreamBasicDescription asbd);
+static void init_physical_format(struct ao *ao);
 
 static bool reinit_device(struct ao *ao) {
     struct priv *p = ao->priv;
@@ -144,8 +146,10 @@ coreaudio_error:
 
 static int init(struct ao *ao)
 {
-    if (AF_FORMAT_IS_IEC61937(ao->format)) {
-        MP_WARN(ao, "detected IEC61937, redirecting to coreaudio_exclusive\n");
+    struct priv *p = ao->priv;
+
+    if (!af_fmt_is_pcm(ao->format) || p->exclusive) {
+        MP_VERBOSE(ao, "redirecting to coreaudio_exclusive\n");
         ao->redirect = "coreaudio_exclusive";
         return CONTROL_ERROR;
     }
@@ -153,10 +157,11 @@ static int init(struct ao *ao)
     if (!reinit_device(ao))
         goto coreaudio_error;
 
-    if (!init_chmap(ao))
-        goto coreaudio_error;
+    if (p->change_physical_format)
+        init_physical_format(ao);
 
-    ao->format = af_fmt_from_planar(ao->format);
+    if (!ca_init_chmap(ao, p->device))
+        goto coreaudio_error;
 
     AudioStreamBasicDescription asbd;
     ca_fill_asbd(ao, &asbd);
@@ -170,95 +175,71 @@ coreaudio_error:
     return CONTROL_ERROR;
 }
 
-static AudioChannelLayout* ca_query_layout(struct ao *ao, void *talloc_ctx)
+static void init_physical_format(struct ao *ao)
 {
     struct priv *p = ao->priv;
-    OSStatus err;
-    uint32_t psize;
-    AudioChannelLayout *r = NULL;
+    OSErr err;
 
-    AudioObjectPropertyAddress p_addr = (AudioObjectPropertyAddress) {
-        .mSelector = kAudioDevicePropertyPreferredChannelLayout,
-        .mScope    = kAudioDevicePropertyScopeOutput,
-        .mElement  = kAudioObjectPropertyElementWildcard,
-    };
+    AudioStreamBasicDescription asbd;
+    ca_fill_asbd(ao, &asbd);
 
-    err = AudioObjectGetPropertyDataSize(p->device, &p_addr, 0, NULL, &psize);
-    CHECK_CA_ERROR("could not get device preferred layout (size)");
+    AudioStreamID *streams;
+    size_t n_streams;
 
-    r = talloc_size(talloc_ctx, psize);
+    err = CA_GET_ARY_O(p->device, kAudioDevicePropertyStreams,
+                       &streams, &n_streams);
+    CHECK_CA_ERROR("could not get number of streams");
 
-    err = AudioObjectGetPropertyData(p->device, &p_addr, 0, NULL, &psize, r);
-    CHECK_CA_ERROR("could not get device preferred layout (get)");
+    MP_VERBOSE(ao, "Found %zd substream(s).\n", n_streams);
 
-coreaudio_error:
-    return r;
-}
+    for (int i = 0; i < n_streams; i++) {
+        AudioStreamRangedDescription *formats;
+        size_t n_formats;
 
-static AudioChannelLayout* ca_query_stereo_layout(struct ao *ao, void *talloc_ctx)
-{
-    struct priv *p = ao->priv;
-    OSStatus err;
-    const int nch = 2;
-    uint32_t channels[nch];
-    AudioChannelLayout *r = NULL;
+        MP_VERBOSE(ao, "Looking at formats in substream %d...\n", i);
 
-    AudioObjectPropertyAddress p_addr = (AudioObjectPropertyAddress) {
-        .mSelector = kAudioDevicePropertyPreferredChannelsForStereo,
-        .mScope    = kAudioDevicePropertyScopeOutput,
-        .mElement  = kAudioObjectPropertyElementWildcard,
-    };
+        err = CA_GET_ARY(streams[i], kAudioStreamPropertyAvailablePhysicalFormats,
+                         &formats, &n_formats);
 
-    uint32_t psize = sizeof(channels);
-    err = AudioObjectGetPropertyData(p->device, &p_addr, 0, NULL, &psize, channels);
-    CHECK_CA_ERROR("could not get device preferred stereo layout");
+        if (!CHECK_CA_WARN("could not get number of stream formats"))
+            continue; // try next one
 
-    psize = sizeof(AudioChannelLayout) + nch * sizeof(AudioChannelDescription);
-    r = talloc_zero_size(talloc_ctx, psize);
-    r->mChannelLayoutTag = kAudioChannelLayoutTag_UseChannelDescriptions;
-    r->mNumberChannelDescriptions = nch;
 
-    AudioChannelDescription desc = {0};
-    desc.mChannelFlags = kAudioChannelFlags_AllOff;
+        uint32_t direction;
+        err = CA_GET(streams[i], kAudioStreamPropertyDirection, &direction);
+        CHECK_CA_ERROR("could not get stream direction");
+        if (direction != 0) {
+            MP_VERBOSE(ao, "Not an output stream.\n");
+            continue;
+        }
 
-    for(int i = 0; i < nch; i++) {
-        desc.mChannelLabel = channels[i];
-        r->mChannelDescriptions[i] = desc;
+        AudioStreamBasicDescription best_asbd = {0};
+
+        for (int j = 0; j < n_formats; j++) {
+            AudioStreamBasicDescription *stream_asbd = &formats[j].mFormat;
+
+            ca_print_asbd(ao, "- ", stream_asbd);
+
+            if (!best_asbd.mFormatID || ca_asbd_is_better(&asbd, &best_asbd,
+                                                          stream_asbd))
+                best_asbd = *stream_asbd;
+        }
+
+        if (best_asbd.mFormatID) {
+            p->original_asbd_stream = streams[i];
+            err = CA_GET(p->original_asbd_stream,
+                         kAudioStreamPropertyPhysicalFormat,
+                         &p->original_asbd);
+            CHECK_CA_WARN("could not get current physical stream format");
+
+            if (!ca_change_physical_format_sync(ao, streams[i], best_asbd))
+                p->original_asbd = (AudioStreamBasicDescription){0};
+            break;
+        }
     }
 
 coreaudio_error:
-    return r;
-}
-
-static bool init_chmap(struct ao *ao)
-{
-    struct priv *p = ao->priv;
-    void *ta_ctx = talloc_new(NULL);
-
-    struct mp_chmap_sel chmap_sel = {.tmp = p};
-    struct mp_chmap chmap = {0};
-
-    AudioChannelLayout *ml = ca_query_layout(ao, ta_ctx);
-    if (ml && ca_layout_to_mp_chmap(ao, ml, &chmap))
-        mp_chmap_sel_add_map(&chmap_sel, &chmap);
-
-    AudioChannelLayout *sl = ca_query_stereo_layout(ao, ta_ctx);
-    if (sl && ca_layout_to_mp_chmap(ao, sl, &chmap))
-        mp_chmap_sel_add_map(&chmap_sel, &chmap);
-
-    talloc_free(ta_ctx);
-
-    if (!ao_chmap_sel_adjust(ao, &chmap_sel, &ao->channels)) {
-        MP_ERR(ao, "could not select a suitable channel map among the "
-                   "hardware supported ones. Make sure to configure your "
-                   "output device correctly in 'Audio MIDI Setup.app'\n");
-        goto coreaudio_error;
-    }
-
-    return true;
-
-coreaudio_error:
-    return false;
+    return;
 }
 
 static bool init_audiounit(struct ao *ao, AudioStreamBasicDescription asbd)
@@ -351,13 +332,23 @@ static void uninit(struct ao *ao)
     AudioOutputUnitStop(p->audio_unit);
     AudioUnitUninitialize(p->audio_unit);
     AudioComponentInstanceDispose(p->audio_unit);
+
+    if (p->original_asbd.mFormatID) {
+        OSStatus err = CA_SET(p->original_asbd_stream,
+                              kAudioStreamPropertyPhysicalFormat,
+                              &p->original_asbd);
+        CHECK_CA_WARN("could not restore physical stream format");
+    }
 }
 
 static OSStatus hotplug_cb(AudioObjectID id, UInt32 naddr,
                            const AudioObjectPropertyAddress addr[],
-                           void *ctx) {
-    reinit_device(ctx);
-    ao_hotplug_event(ctx);
+                           void *ctx)
+{
+    struct ao *ao = ctx;
+    MP_VERBOSE(ao, "Handling potential hotplug event...\n");
+    reinit_device(ao);
+    ao_hotplug_event(ao);
     return noErr;
 }
 
@@ -381,8 +372,8 @@ static int hotplug_init(struct ao *ao)
         err = AudioObjectAddPropertyListener(
             kAudioObjectSystemObject, &addr, hotplug_cb, (void *)ao);
         if (err != noErr) {
-            char *c1 = fourcc_repr(ao, hotplug_properties[i]);
-            char *c2 = fourcc_repr(ao, err);
+            char *c1 = fourcc_repr(hotplug_properties[i]);
+            char *c2 = fourcc_repr(err);
             MP_ERR(ao, "failed to set device listener %s (%s)", c1, c2);
             goto coreaudio_error;
         }
@@ -406,163 +397,13 @@ static void hotplug_uninit(struct ao *ao)
         err = AudioObjectRemovePropertyListener(
             kAudioObjectSystemObject, &addr, hotplug_cb, (void *)ao);
         if (err != noErr) {
-            char *c1 = fourcc_repr(ao, hotplug_properties[i]);
-            char *c2 = fourcc_repr(ao, err);
+            char *c1 = fourcc_repr(hotplug_properties[i]);
+            char *c2 = fourcc_repr(err);
             MP_ERR(ao, "failed to set device listener %s (%s)", c1, c2);
         }
     }
 }
 
-
-// Channel Mapping functions
-static const int speaker_map[][2] = {
-    { kAudioChannelLabel_Left,                 MP_SPEAKER_ID_FL   },
-    { kAudioChannelLabel_Right,                MP_SPEAKER_ID_FR   },
-    { kAudioChannelLabel_Center,               MP_SPEAKER_ID_FC   },
-    { kAudioChannelLabel_LFEScreen,            MP_SPEAKER_ID_LFE  },
-    { kAudioChannelLabel_LeftSurround,         MP_SPEAKER_ID_BL   },
-    { kAudioChannelLabel_RightSurround,        MP_SPEAKER_ID_BR   },
-    { kAudioChannelLabel_LeftCenter,           MP_SPEAKER_ID_FLC  },
-    { kAudioChannelLabel_RightCenter,          MP_SPEAKER_ID_FRC  },
-    { kAudioChannelLabel_CenterSurround,       MP_SPEAKER_ID_BC   },
-    { kAudioChannelLabel_LeftSurroundDirect,   MP_SPEAKER_ID_SL   },
-    { kAudioChannelLabel_RightSurroundDirect,  MP_SPEAKER_ID_SR   },
-    { kAudioChannelLabel_TopCenterSurround,    MP_SPEAKER_ID_TC   },
-    { kAudioChannelLabel_VerticalHeightLeft,   MP_SPEAKER_ID_TFL  },
-    { kAudioChannelLabel_VerticalHeightCenter, MP_SPEAKER_ID_TFC  },
-    { kAudioChannelLabel_VerticalHeightRight,  MP_SPEAKER_ID_TFR  },
-    { kAudioChannelLabel_TopBackLeft,          MP_SPEAKER_ID_TBL  },
-    { kAudioChannelLabel_TopBackCenter,        MP_SPEAKER_ID_TBC  },
-    { kAudioChannelLabel_TopBackRight,         MP_SPEAKER_ID_TBR  },
-
-    // unofficial extensions
-    { kAudioChannelLabel_RearSurroundLeft,     MP_SPEAKER_ID_SDL  },
-    { kAudioChannelLabel_RearSurroundRight,    MP_SPEAKER_ID_SDR  },
-    { kAudioChannelLabel_LeftWide,             MP_SPEAKER_ID_WL   },
-    { kAudioChannelLabel_RightWide,            MP_SPEAKER_ID_WR   },
-    { kAudioChannelLabel_LFE2,                 MP_SPEAKER_ID_LFE2 },
-
-    { kAudioChannelLabel_HeadphonesLeft,       MP_SPEAKER_ID_DL   },
-    { kAudioChannelLabel_HeadphonesRight,      MP_SPEAKER_ID_DR   },
-
-    { kAudioChannelLabel_Unknown,              MP_SPEAKER_ID_UNKNOWN0 },
-};
-
-static int ca_label_to_mp_speaker_id(AudioChannelLabel label)
-{
-    for (int i = 0; speaker_map[i][1] >= 0; i++)
-        if (speaker_map[i][0] == label)
-            return speaker_map[i][1];
-    return -1;
-}
-
-static void ca_log_layout(struct ao *ao, int l, AudioChannelLayout *layout)
-{
-    if (!mp_msg_test(ao->log, l))
-        return;
-
-    AudioChannelDescription *descs = layout->mChannelDescriptions;
-
-    mp_msg(ao->log, l, "layout: tag: <%u>, bitmap: <%u>, "
-                       "descriptions <%u>\n",
-                       (unsigned) layout->mChannelLayoutTag,
-                       (unsigned) layout->mChannelBitmap,
-                       (unsigned) layout->mNumberChannelDescriptions);
-
-    for (int i = 0; i < layout->mNumberChannelDescriptions; i++) {
-        AudioChannelDescription d = descs[i];
-        mp_msg(ao->log, l, " - description %d: label <%u, %u>, "
-            " flags: <%u>, coords: <%f, %f, %f>\n", i,
-            (unsigned) d.mChannelLabel,
-            (unsigned) ca_label_to_mp_speaker_id(d.mChannelLabel),
-            (unsigned) d.mChannelFlags,
-            d.mCoordinates[0],
-            d.mCoordinates[1],
-            d.mCoordinates[2]);
-    }
-}
-
-static AudioChannelLayout *ca_layout_to_custom_layout(
-    struct ao *ao, void *talloc_ctx, AudioChannelLayout *l)
-{
-    AudioChannelLayoutTag tag = l->mChannelLayoutTag;
-    AudioChannelLayout *r;
-    OSStatus err;
-
-    if (tag == kAudioChannelLayoutTag_UseChannelBitmap) {
-        uint32_t psize;
-        err = AudioFormatGetPropertyInfo(
-            kAudioFormatProperty_ChannelLayoutForBitmap,
-            sizeof(uint32_t), &l->mChannelBitmap, &psize);
-        CHECK_CA_ERROR("failed to convert channel bitmap to descriptions (info)");
-        r = talloc_size(NULL, psize);
-        err = AudioFormatGetProperty(
-            kAudioFormatProperty_ChannelLayoutForBitmap,
-            sizeof(uint32_t), &l->mChannelBitmap, &psize, r);
-        CHECK_CA_ERROR("failed to convert channel bitmap to descriptions (get)");
-    } else if (tag != kAudioChannelLayoutTag_UseChannelDescriptions) {
-        uint32_t psize;
-        err = AudioFormatGetPropertyInfo(
-            kAudioFormatProperty_ChannelLayoutForTag,
-            sizeof(AudioChannelLayoutTag), &l->mChannelLayoutTag, &psize);
-        r = talloc_size(NULL, psize);
-        CHECK_CA_ERROR("failed to convert channel tag to descriptions (info)");
-        err = AudioFormatGetProperty(
-            kAudioFormatProperty_ChannelLayoutForTag,
-            sizeof(AudioChannelLayoutTag), &l->mChannelLayoutTag, &psize, r);
-        CHECK_CA_ERROR("failed to convert channel tag to descriptions (get)");
-    } else {
-        r = l;
-    }
-
-    return r;
-coreaudio_error:
-    return NULL;
-}
-
-bool ca_layout_to_mp_chmap(struct ao *ao, AudioChannelLayout *layout,
-                           struct mp_chmap *chmap)
-{
-    void *talloc_ctx = talloc_new(NULL);
-
-    MP_DBG(ao, "input channel layout:\n");
-    ca_log_layout(ao, MSGL_DEBUG, layout);
-
-    AudioChannelLayout *l = ca_layout_to_custom_layout(ao, talloc_ctx, layout);
-    if (!l)
-        goto coreaudio_error;
-
-    MP_VERBOSE(ao, "converted input channel layout:\n");
-    ca_log_layout(ao, MSGL_V, l);
-
-    if (l->mNumberChannelDescriptions > MP_NUM_CHANNELS) {
-        MP_VERBOSE(ao, "layout has too many descriptions (%u, max: %d)\n",
-                   (unsigned) l->mNumberChannelDescriptions, MP_NUM_CHANNELS);
-        return false;
-    }
-
-    for (int n = 0; n < l->mNumberChannelDescriptions; n++) {
-        AudioChannelLabel label = l->mChannelDescriptions[n].mChannelLabel;
-        uint8_t speaker = ca_label_to_mp_speaker_id(label);
-        if (speaker < 0) {
-            MP_VERBOSE(ao, "channel label=%u unusable to build channel "
-                           "bitmap, skipping layout\n", (unsigned) label);
-            goto coreaudio_error;
-        } else {
-            chmap->speaker[n] = speaker;
-            chmap->num = n + 1;
-        }
-    }
-
-    talloc_free(talloc_ctx);
-    return chmap->num > 0;
-coreaudio_error:
-    MP_VERBOSE(ao, "converted input channel layout (failed):\n");
-    ca_log_layout(ao, MSGL_V, layout);
-    talloc_free(talloc_ctx);
-    return false;
-}
-
 #define OPT_BASE_STRUCT struct priv
 
 const struct ao_driver audio_out_coreaudio = {
@@ -577,4 +418,9 @@ const struct ao_driver audio_out_coreaudio = {
     .hotplug_uninit = hotplug_uninit,
     .list_devs      = ca_get_device_list,
     .priv_size      = sizeof(struct priv),
+    .options = (const struct m_option[]){
+        OPT_FLAG("change-physical-format", change_physical_format, 0),
+        OPT_FLAG("exclusive", exclusive, 0),
+        {0}
+    },
 };

audio/out/ao_coreaudio_chmap.c

@@ -0,0 +1,267 @@
+/*
+ * 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 "common/common.h"
+
+#include "ao_coreaudio_utils.h"
+
+#include "ao_coreaudio_chmap.h"
+
+static const int speaker_map[][2] = {
+    { kAudioChannelLabel_Left,                 MP_SPEAKER_ID_FL   },
+    { kAudioChannelLabel_Right,                MP_SPEAKER_ID_FR   },
+    { kAudioChannelLabel_Center,               MP_SPEAKER_ID_FC   },
+    { kAudioChannelLabel_LFEScreen,            MP_SPEAKER_ID_LFE  },
+    { kAudioChannelLabel_LeftSurround,         MP_SPEAKER_ID_BL   },
+    { kAudioChannelLabel_RightSurround,        MP_SPEAKER_ID_BR   },
+    { kAudioChannelLabel_LeftCenter,           MP_SPEAKER_ID_FLC  },
+    { kAudioChannelLabel_RightCenter,          MP_SPEAKER_ID_FRC  },
+    { kAudioChannelLabel_CenterSurround,       MP_SPEAKER_ID_BC   },
+    { kAudioChannelLabel_LeftSurroundDirect,   MP_SPEAKER_ID_SL   },
+    { kAudioChannelLabel_RightSurroundDirect,  MP_SPEAKER_ID_SR   },
+    { kAudioChannelLabel_TopCenterSurround,    MP_SPEAKER_ID_TC   },
+    { kAudioChannelLabel_VerticalHeightLeft,   MP_SPEAKER_ID_TFL  },
+    { kAudioChannelLabel_VerticalHeightCenter, MP_SPEAKER_ID_TFC  },
+    { kAudioChannelLabel_VerticalHeightRight,  MP_SPEAKER_ID_TFR  },
+    { kAudioChannelLabel_TopBackLeft,          MP_SPEAKER_ID_TBL  },
+    { kAudioChannelLabel_TopBackCenter,        MP_SPEAKER_ID_TBC  },
+    { kAudioChannelLabel_TopBackRight,         MP_SPEAKER_ID_TBR  },
+
+    // unofficial extensions
+    { kAudioChannelLabel_RearSurroundLeft,     MP_SPEAKER_ID_SDL  },
+    { kAudioChannelLabel_RearSurroundRight,    MP_SPEAKER_ID_SDR  },
+    { kAudioChannelLabel_LeftWide,             MP_SPEAKER_ID_WL   },
+    { kAudioChannelLabel_RightWide,            MP_SPEAKER_ID_WR   },
+    { kAudioChannelLabel_LFE2,                 MP_SPEAKER_ID_LFE2 },
+
+    { kAudioChannelLabel_HeadphonesLeft,       MP_SPEAKER_ID_DL   },
+    { kAudioChannelLabel_HeadphonesRight,      MP_SPEAKER_ID_DR   },
+
+    { kAudioChannelLabel_Unknown,              MP_SPEAKER_ID_NA   },
+
+    { 0,                                       -1                 },
+};
+
+static int ca_label_to_mp_speaker_id(AudioChannelLabel label)
+{
+    for (int i = 0; speaker_map[i][1] >= 0; i++)
+        if (speaker_map[i][0] == label)
+            return speaker_map[i][1];
+    return -1;
+}
+
+static void ca_log_layout(struct ao *ao, int l, AudioChannelLayout *layout)
+{
+    if (!mp_msg_test(ao->log, l))
+        return;
+
+    AudioChannelDescription *descs = layout->mChannelDescriptions;
+
+    mp_msg(ao->log, l, "layout: tag: <%u>, bitmap: <%u>, "
+                       "descriptions <%u>\n",
+                       (unsigned) layout->mChannelLayoutTag,
+                       (unsigned) layout->mChannelBitmap,
+                       (unsigned) layout->mNumberChannelDescriptions);
+
+    for (int i = 0; i < layout->mNumberChannelDescriptions; i++) {
+        AudioChannelDescription d = descs[i];
+        mp_msg(ao->log, l, " - description %d: label <%u, %u>, "
+            " flags: <%u>, coords: <%f, %f, %f>\n", i,
+            (unsigned) d.mChannelLabel,
+            (unsigned) ca_label_to_mp_speaker_id(d.mChannelLabel),
+            (unsigned) d.mChannelFlags,
+            d.mCoordinates[0],
+            d.mCoordinates[1],
+            d.mCoordinates[2]);
+    }
+}
+
+static AudioChannelLayout *ca_layout_to_custom_layout(struct ao *ao,
+                                                      void *talloc_ctx,
+                                                      AudioChannelLayout *l)
+{
+    AudioChannelLayoutTag tag = l->mChannelLayoutTag;
+    AudioChannelLayout *r;
+    OSStatus err;
+
+    if (tag == kAudioChannelLayoutTag_UseChannelBitmap) {
+        uint32_t psize;
+        err = AudioFormatGetPropertyInfo(
+            kAudioFormatProperty_ChannelLayoutForBitmap,
+            sizeof(uint32_t), &l->mChannelBitmap, &psize);
+        CHECK_CA_ERROR("failed to convert channel bitmap to descriptions (info)");
+        r = talloc_size(NULL, psize);
+        err = AudioFormatGetProperty(
+            kAudioFormatProperty_ChannelLayoutForBitmap,
+            sizeof(uint32_t), &l->mChannelBitmap, &psize, r);
+        CHECK_CA_ERROR("failed to convert channel bitmap to descriptions (get)");
+    } else if (tag != kAudioChannelLayoutTag_UseChannelDescriptions) {
+        uint32_t psize;
+        err = AudioFormatGetPropertyInfo(
+            kAudioFormatProperty_ChannelLayoutForTag,
+            sizeof(AudioChannelLayoutTag), &l->mChannelLayoutTag, &psize);
+        r = talloc_size(NULL, psize);
+        CHECK_CA_ERROR("failed to convert channel tag to descriptions (info)");
+        err = AudioFormatGetProperty(
+            kAudioFormatProperty_ChannelLayoutForTag,
+            sizeof(AudioChannelLayoutTag), &l->mChannelLayoutTag, &psize, r);
+        CHECK_CA_ERROR("failed to convert channel tag to descriptions (get)");
+    } else {
+        r = l;
+    }
+
+    return r;
+coreaudio_error:
+    return NULL;
+}
+
+static bool ca_layout_to_mp_chmap(struct ao *ao, AudioChannelLayout *layout,
+                                  struct mp_chmap *chmap)
+{
+    void *talloc_ctx = talloc_new(NULL);
+
+    MP_DBG(ao, "input channel layout:\n");
+    ca_log_layout(ao, MSGL_DEBUG, layout);
+
+    AudioChannelLayout *l = ca_layout_to_custom_layout(ao, talloc_ctx, layout);
+    if (!l)
+        goto coreaudio_error;
+
+    MP_VERBOSE(ao, "converted input channel layout:\n");
+    ca_log_layout(ao, MSGL_V, l);
+
+    if (l->mNumberChannelDescriptions > MP_NUM_CHANNELS) {
+        MP_VERBOSE(ao, "layout has too many descriptions (%u, max: %d)\n",
+                   (unsigned) l->mNumberChannelDescriptions, MP_NUM_CHANNELS);
+        return false;
+    }
+
+    for (int n = 0; n < l->mNumberChannelDescriptions; n++) {
+        AudioChannelLabel label = l->mChannelDescriptions[n].mChannelLabel;
+        int speaker = ca_label_to_mp_speaker_id(label);
+        if (speaker < 0) {
+            MP_VERBOSE(ao, "channel label=%u unusable to build channel "
+                           "bitmap, skipping layout\n", (unsigned) label);
+            goto coreaudio_error;
+        } else {
+            chmap->speaker[n] = speaker;
+            chmap->num = n + 1;
+        }
+    }
+
+    talloc_free(talloc_ctx);
+    return chmap->num > 0;
+coreaudio_error:
+    MP_VERBOSE(ao, "converted input channel layout (failed):\n");
+    ca_log_layout(ao, MSGL_V, layout);
+    talloc_free(talloc_ctx);
+    return false;
+}
+
+static AudioChannelLayout* ca_query_layout(struct ao *ao,
+                                           AudioDeviceID device,
+                                           void *talloc_ctx)
+{
+    OSStatus err;
+    uint32_t psize;
+    AudioChannelLayout *r = NULL;
+
+    AudioObjectPropertyAddress p_addr = (AudioObjectPropertyAddress) {
+        .mSelector = kAudioDevicePropertyPreferredChannelLayout,
+        .mScope    = kAudioDevicePropertyScopeOutput,
+        .mElement  = kAudioObjectPropertyElementWildcard,
+    };
+
+    err = AudioObjectGetPropertyDataSize(device, &p_addr, 0, NULL, &psize);
+    CHECK_CA_ERROR("could not get device preferred layout (size)");
+
+    r = talloc_size(talloc_ctx, psize);
+
+    err = AudioObjectGetPropertyData(device, &p_addr, 0, NULL, &psize, r);
+    CHECK_CA_ERROR("could not get device preferred layout (get)");
+
+coreaudio_error:
+    return r;
+}
+
+static AudioChannelLayout* ca_query_stereo_layout(struct ao *ao,
+                                                  AudioDeviceID device,
+                                                  void *talloc_ctx)
+{
+    OSStatus err;
+    const int nch = 2;
+    uint32_t channels[nch];
+    AudioChannelLayout *r = NULL;
+
+    AudioObjectPropertyAddress p_addr = (AudioObjectPropertyAddress) {
+        .mSelector = kAudioDevicePropertyPreferredChannelsForStereo,
+        .mScope    = kAudioDevicePropertyScopeOutput,
+        .mElement  = kAudioObjectPropertyElementWildcard,
+    };
+
+    uint32_t psize = sizeof(channels);
+    err = AudioObjectGetPropertyData(device, &p_addr, 0, NULL, &psize, channels);
+    CHECK_CA_ERROR("could not get device preferred stereo layout");
+
+    psize = sizeof(AudioChannelLayout) + nch * sizeof(AudioChannelDescription);
+    r = talloc_zero_size(talloc_ctx, psize);
+    r->mChannelLayoutTag = kAudioChannelLayoutTag_UseChannelDescriptions;
+    r->mNumberChannelDescriptions = nch;
+
+    AudioChannelDescription desc = {0};
+    desc.mChannelFlags = kAudioChannelFlags_AllOff;
+
+    for(int i = 0; i < nch; i++) {
+        desc.mChannelLabel = channels[i];
+        r->mChannelDescriptions[i] = desc;
+    }
+
+coreaudio_error:
+    return r;
+}
+
+bool ca_init_chmap(struct ao *ao, AudioDeviceID device)
+{
+    void *ta_ctx = talloc_new(NULL);
+
+    struct mp_chmap_sel chmap_sel = {.tmp = ta_ctx};
+    struct mp_chmap chmap = {0};
+
+    mp_chmap_sel_add_map(&chmap_sel, &(struct mp_chmap)MP_CHMAP_INIT_MONO);
+
+    AudioChannelLayout *ml = ca_query_layout(ao, device, ta_ctx);
+    if (ml && ca_layout_to_mp_chmap(ao, ml, &chmap))
+        mp_chmap_sel_add_map(&chmap_sel, &chmap);
+
+    AudioChannelLayout *sl = ca_query_stereo_layout(ao, device, ta_ctx);
+    if (sl && ca_layout_to_mp_chmap(ao, sl, &chmap))
+        mp_chmap_sel_add_map(&chmap_sel, &chmap);
+
+    if (!ao_chmap_sel_adjust(ao, &chmap_sel, &ao->channels)) {
+        MP_ERR(ao, "could not select a suitable channel map among the "
+                   "hardware supported ones. Make sure to configure your "
+                   "output device correctly in 'Audio MIDI Setup.app'\n");
+        goto coreaudio_error;
+    }
+
+    talloc_free(ta_ctx);
+    return true;
+
+coreaudio_error:
+    talloc_free(ta_ctx);
+    return false;
+}

audio/out/ao_coreaudio_chmap.h

@@ -0,0 +1,25 @@
+/*
+ * 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/>.
+ */
+
+#ifndef MPV_COREAUDIO_CHMAP_H
+#define MPV_COREAUDIO_CHMAP_H
+
+#include <AudioToolbox/AudioToolbox.h>
+
+bool ca_init_chmap(struct ao *ao, AudioDeviceID device);
+
+#endif

audio/out/ao_coreaudio_exclusive.c

@@ -41,297 +41,100 @@
 #include "internal.h"
 #include "audio/format.h"
 #include "osdep/timer.h"
+#include "osdep/atomics.h"
 #include "options/m_option.h"
 #include "common/msg.h"
+#include "audio/out/ao_coreaudio_chmap.h"
 #include "audio/out/ao_coreaudio_properties.h"
 #include "audio/out/ao_coreaudio_utils.h"
 
-static bool ca_format_is_digital(AudioStreamBasicDescription asbd)
-{
-    switch (asbd.mFormatID)
-    case 'IAC3':
-    case 'iac3':
-    case  kAudioFormat60958AC3:
-    case  kAudioFormatAC3:
-        return true;
-    return false;
-}
-
-static bool ca_stream_supports_digital(struct ao *ao, AudioStreamID stream)
-{
-    AudioStreamRangedDescription *formats = NULL;
-    size_t n_formats;
-
-    OSStatus err =
-        CA_GET_ARY(stream, kAudioStreamPropertyAvailablePhysicalFormats,
-                   &formats, &n_formats);
-
-    CHECK_CA_ERROR("Could not get number of stream formats.");
-
-    for (int i = 0; i < n_formats; i++) {
-        AudioStreamBasicDescription asbd = formats[i].mFormat;
-        ca_print_asbd(ao, "supported format:", &(asbd));
-        if (ca_format_is_digital(asbd)) {
-            talloc_free(formats);
-            return true;
-        }
-    }
-
-    talloc_free(formats);
-coreaudio_error:
-    return false;
-}
-
-static bool ca_device_supports_digital(struct ao *ao, AudioDeviceID device)
-{
-    AudioStreamID *streams = NULL;
-    size_t n_streams;
-
-    /* Retrieve all the output streams. */
-    OSStatus err =
-        CA_GET_ARY_O(device, kAudioDevicePropertyStreams, &streams, &n_streams);
-
-    CHECK_CA_ERROR("could not get number of streams.");
-
-    for (int i = 0; i < n_streams; i++) {
-        if (ca_stream_supports_digital(ao, streams[i])) {
-            talloc_free(streams);
-            return true;
-        }
-    }
-
-    talloc_free(streams);
-
-coreaudio_error:
-    return false;
-}
-
-static OSStatus ca_property_listener(
-    AudioObjectPropertySelector selector,
-    AudioObjectID object, uint32_t n_addresses,
-    const AudioObjectPropertyAddress addresses[],
-    void *data)
-{
-    void *talloc_ctx = talloc_new(NULL);
-
-    for (int i = 0; i < n_addresses; i++) {
-        if (addresses[i].mSelector == selector) {
-            if (data) *(volatile int *)data = 1;
-            break;
-        }
-    }
-    talloc_free(talloc_ctx);
-    return noErr;
-}
-
-static OSStatus ca_stream_listener(
-    AudioObjectID object, uint32_t n_addresses,
-    const AudioObjectPropertyAddress addresses[],
-    void *data)
-{
-    return ca_property_listener(kAudioStreamPropertyPhysicalFormat,
-                                object, n_addresses, addresses, data);
-}
-
-static OSStatus ca_device_listener(
-    AudioObjectID object, uint32_t n_addresses,
-    const AudioObjectPropertyAddress addresses[],
-    void *data)
-{
-    return ca_property_listener(kAudioDevicePropertyDeviceHasChanged,
-                                object, n_addresses, addresses, data);
-}
-
-static OSStatus ca_lock_device(AudioDeviceID device, pid_t *pid) {
-    *pid = getpid();
-    OSStatus err = CA_SET(device, kAudioDevicePropertyHogMode, pid);
-    if (err != noErr)
-        *pid = -1;
-
-    return err;
-}
-
-static OSStatus ca_unlock_device(AudioDeviceID device, pid_t *pid) {
-    if (*pid == getpid()) {
-        *pid = -1;
-        return CA_SET(device, kAudioDevicePropertyHogMode, &pid);
-    }
-    return noErr;
-}
-
-static OSStatus ca_change_mixing(struct ao *ao, AudioDeviceID device,
-                                 uint32_t val, bool *changed) {
-    *changed = false;
-
-    AudioObjectPropertyAddress p_addr = (AudioObjectPropertyAddress) {
-        .mSelector = kAudioDevicePropertySupportsMixing,
-        .mScope    = kAudioObjectPropertyScopeGlobal,
-        .mElement  = kAudioObjectPropertyElementMaster,
-    };
-
-    if (AudioObjectHasProperty(device, &p_addr)) {
-        OSStatus err;
-        Boolean writeable = 0;
-        err = CA_SETTABLE(device, kAudioDevicePropertySupportsMixing,
-                          &writeable);
-
-        if (!CHECK_CA_WARN("can't tell if mixing property is settable")) {
-            return err;
-        }
-
-        if (!writeable)
-            return noErr;
-
-        err = CA_SET(device, kAudioDevicePropertySupportsMixing, &val);
-        if (err != noErr)
-            return err;
-
-        if (!CHECK_CA_WARN("can't set mix mode")) {
-            return err;
-        }
-
-        *changed = true;
-    }
-
-    return noErr;
-}
-
-static OSStatus ca_disable_mixing(struct ao *ao,
-                                  AudioDeviceID device, bool *changed) {
-    return ca_change_mixing(ao, device, 0, changed);
-}
-
-static OSStatus ca_enable_mixing(struct ao *ao,
-                                 AudioDeviceID device, bool changed) {
-    if (changed) {
-        bool dont_care = false;
-        return ca_change_mixing(ao, device, 1, &dont_care);
-    }
-
-    return noErr;
-}
-
-static OSStatus ca_change_device_listening(AudioDeviceID device,
-                                           void *flag, bool enabled)
-{
-    AudioObjectPropertyAddress p_addr = (AudioObjectPropertyAddress) {
-        .mSelector = kAudioDevicePropertyDeviceHasChanged,
-        .mScope    = kAudioObjectPropertyScopeGlobal,
-        .mElement  = kAudioObjectPropertyElementMaster,
-    };
-
-    if (enabled) {
-        return AudioObjectAddPropertyListener(
-            device, &p_addr, ca_device_listener, flag);
-    } else {
-        return AudioObjectRemovePropertyListener(
-            device, &p_addr, ca_device_listener, flag);
-    }
-}
-
-static OSStatus ca_enable_device_listener(AudioDeviceID device, void *flag) {
-    return ca_change_device_listening(device, flag, true);
-}
-
-static OSStatus ca_disable_device_listener(AudioDeviceID device, void *flag) {
-    return ca_change_device_listening(device, flag, false);
-}
-
-static bool ca_change_format(struct ao *ao, AudioStreamID stream,
-                             AudioStreamBasicDescription change_format)
-{
-    OSStatus err = noErr;
-    AudioObjectPropertyAddress p_addr;
-    volatile int stream_format_changed = 0;
-
-    ca_print_asbd(ao, "setting stream format:", &change_format);
-
-    /* Install the callback. */
-    p_addr = (AudioObjectPropertyAddress) {
-        .mSelector = kAudioStreamPropertyPhysicalFormat,
-        .mScope    = kAudioObjectPropertyScopeGlobal,
-        .mElement  = kAudioObjectPropertyElementMaster,
-    };
-
-    err = AudioObjectAddPropertyListener(stream, &p_addr, ca_stream_listener,
-                                         (void *)&stream_format_changed);
-    if (!CHECK_CA_WARN("can't add property listener during format change")) {
-        return false;
-    }
-
-    /* Change the format. */
-    err = CA_SET(stream, kAudioStreamPropertyPhysicalFormat, &change_format);
-    if (!CHECK_CA_WARN("error changing physical format")) {
-        return false;
-    }
-
-    /* The AudioStreamSetProperty is not only asynchronious,
-     * it is also not Atomic, in its behaviour.
-     * Therefore we check 5 times before we really give up. */
-    bool format_set = false;
-    for (int i = 0; !format_set && i < 5; i++) {
-        for (int j = 0; !stream_format_changed && j < 50; j++)
-            mp_sleep_us(10000);
-
-        if (stream_format_changed) {
-            stream_format_changed = 0;
-        } else {
-            MP_VERBOSE(ao, "reached timeout\n");
-        }
-
-        AudioStreamBasicDescription actual_format;
-        err = CA_GET(stream, kAudioStreamPropertyPhysicalFormat, &actual_format);
-
-        ca_print_asbd(ao, "actual format in use:", &actual_format);
-        if (actual_format.mSampleRate == change_format.mSampleRate &&
-            actual_format.mFormatID == change_format.mFormatID &&
-            actual_format.mFramesPerPacket == change_format.mFramesPerPacket) {
-            format_set = true;
-        }
-    }
-
-    err = AudioObjectRemovePropertyListener(stream, &p_addr, ca_stream_listener,
-                                            (void *)&stream_format_changed);
-
-    if (!CHECK_CA_WARN("can't remove property listener")) {
-        return false;
-    }
-
-    return format_set;
-}
-
-
 struct priv {
     AudioDeviceID device;   // selected device
 
     bool paused;
 
-    // digital render callback
+    // audio render callback
     AudioDeviceIOProcID render_cb;
 
     // pid set for hog mode, (-1) means that hog mode on the device was
     // released. hog mode is exclusive access to a device
     pid_t hog_pid;
 
-    // stream selected for digital playback by the detection in init
     AudioStreamID stream;
 
     // stream index in an AudioBufferList
     int stream_idx;
 
-    // format we changed the stream to: for the digital case each application
-    // sets the stream format for a device to what it needs
+    // format we changed the stream to, and the original format to restore
     AudioStreamBasicDescription stream_asbd;
     AudioStreamBasicDescription original_asbd;
 
     bool changed_mixing;
-    int stream_asbd_changed;
-    bool reload_requested;
+
+    atomic_bool reload_requested;
 
     uint32_t hw_latency_us;
 };
 
-static OSStatus render_cb_digital(
+static OSStatus property_listener_cb(
+    AudioObjectID object, uint32_t n_addresses,
+    const AudioObjectPropertyAddress addresses[],
+    void *data)
+{
+    struct ao *ao = data;
+    struct priv *p = ao->priv;
+
+    // Check whether we need to reset the compressed output stream.
+    AudioStreamBasicDescription f;
+    OSErr err = CA_GET(p->stream, kAudioStreamPropertyVirtualFormat, &f);
+    CHECK_CA_WARN("could not get stream format");
+    if (err != noErr || !ca_asbd_equals(&p->stream_asbd, &f)) {
+        if (atomic_compare_exchange_strong(&p->reload_requested,
+                                           &(bool){false}, true))
+        {
+            ao_request_reload(ao);
+            MP_INFO(ao, "Stream format changed! Reloading.\n");
+        }
+    }
+
+    return noErr;
+}
+
+static OSStatus enable_property_listener(struct ao *ao, bool enabled)
+{
+    struct priv *p = ao->priv;
+
+    uint32_t selectors[] = {kAudioDevicePropertyDeviceHasChanged,
+                            kAudioHardwarePropertyDevices};
+    AudioDeviceID devs[] = {p->device,
+                            kAudioObjectSystemObject};
+    assert(MP_ARRAY_SIZE(selectors) == MP_ARRAY_SIZE(devs));
+
+    OSStatus status = noErr;
+    for (int n = 0; n < MP_ARRAY_SIZE(devs); n++) {
+        AudioObjectPropertyAddress addr = {
+            .mScope    = kAudioObjectPropertyScopeGlobal,
+            .mElement  = kAudioObjectPropertyElementMaster,
+            .mSelector = selectors[n],
+        };
+        AudioDeviceID device = devs[n];
+
+        OSStatus status2;
+        if (enabled) {
+            status2 = AudioObjectAddPropertyListener(
+                device, &addr, property_listener_cb, ao);
+        } else {
+            status2 = AudioObjectRemovePropertyListener(
+                device, &addr, property_listener_cb, ao);
+        }
+        if (status == noErr)
+            status = status2;
+    }
+
+    return status;
+}
+
+static OSStatus render_cb_compressed(
         AudioDeviceID device, const AudioTimeStamp *ts,
         const void *in_data, const AudioTimeStamp *in_ts,
         AudioBufferList *out_data, const AudioTimeStamp *out_ts, void *ctx)
@@ -355,58 +158,112 @@ static OSStatus render_cb_digital(
 
     ao_read_data(ao, &buf.mData, pseudo_frames, end);
 
-    // Check whether we need to reset the digital output stream.
-    if (p->stream_asbd_changed) {
-        p->stream_asbd_changed = 0;
-        if (!p->reload_requested && ca_stream_supports_digital(ao, p->stream)) {
-            p->reload_requested = true;
-            ao_request_reload(ao);
-            MP_INFO(ao, "Stream format changed! Reloading.\n");
-        }
-    }
-
     return noErr;
 }
 
-static int init_digital(struct ao *ao, AudioStreamBasicDescription asbd);
+// Apparently, audio devices can have multiple sub-streams. It's not clear to
+// me what devices with multiple streams actually do. So only select the first
+// one that fulfills some minimum requirements.
+// If this is not sufficient, we could duplicate the device list entries for
+// each sub-stream, and make it explicit.
+static int select_stream(struct ao *ao)
+{
+    struct priv *p = ao->priv;
+
+    AudioStreamID *streams;
+    size_t n_streams;
+    OSStatus err;
+
+    /* Get a list of all the streams on this device. */
+    err = CA_GET_ARY_O(p->device, kAudioDevicePropertyStreams,
+                       &streams, &n_streams);
+    CHECK_CA_ERROR("could not get number of streams");
+    for (int i = 0; i < n_streams; i++) {
+        uint32_t direction;
+        err = CA_GET(streams[i], kAudioStreamPropertyDirection, &direction);
+        CHECK_CA_WARN("could not get stream direction");
+        if (err == noErr && direction != 0) {
+            MP_VERBOSE(ao, "Substream %d is not an output stream.\n", i);
+            continue;
+        }
+
+        if (af_fmt_is_pcm(ao->format) || ca_stream_supports_compressed(ao,
+                                                                   streams[i]))
+        {
+            MP_VERBOSE(ao, "Using substream %d/%zd.\n", i, n_streams);
+            p->stream = streams[i];
+            p->stream_idx = i;
+            break;
+        }
+    }
+
+    talloc_free(streams);
+
+    if (p->stream_idx < 0) {
+        MP_ERR(ao, "No useable substream found.\n");
+        goto coreaudio_error;
+    }
+
+    return 0;
+
+coreaudio_error:
+    return -1;
+}
+
+static int find_best_format(struct ao *ao, AudioStreamBasicDescription *out_fmt)
+{
+    struct priv *p = ao->priv;
+
+    // Build ASBD for the input format
+    AudioStreamBasicDescription asbd;
+    ca_fill_asbd(ao, &asbd);
+    ca_print_asbd(ao, "our format:", &asbd);
+
+    *out_fmt = (AudioStreamBasicDescription){0};
+
+    AudioStreamRangedDescription *formats;
+    size_t n_formats;
+    OSStatus err;
+
+    err = CA_GET_ARY(p->stream, kAudioStreamPropertyAvailablePhysicalFormats,
+                     &formats, &n_formats);
+    CHECK_CA_ERROR("could not get number of stream formats");
+
+    for (int j = 0; j < n_formats; j++) {
+        AudioStreamBasicDescription *stream_asbd = &formats[j].mFormat;
+
+        ca_print_asbd(ao, "- ", stream_asbd);
+
+        if (!out_fmt->mFormatID || ca_asbd_is_better(&asbd, out_fmt, stream_asbd))
+            *out_fmt = *stream_asbd;
+    }
+
+    talloc_free(formats);
+
+    if (!out_fmt->mFormatID) {
+        MP_ERR(ao, "no format found\n");
+        return -1;
+    }
+
+    return 0;
+coreaudio_error:
+    return -1;
+}
 
 static int init(struct ao *ao)
 {
     struct priv *p = ao->priv;
 
     OSStatus err = ca_select_device(ao, ao->device, &p->device);
-    CHECK_CA_ERROR("failed to select device");
+    CHECK_CA_ERROR_L(coreaudio_error_nounlock, "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,
-     * or anything else IEC61937-framed. */
-    if (AF_FORMAT_IS_IEC61937(ao->format)) {
-        if (ca_device_supports_digital(ao, p->device))
-            supports_digital = true;
+    if (!af_fmt_is_pcm(ao->format) && !af_fmt_is_spdif(ao->format)) {
+        MP_ERR(ao, "Unsupported format.\n");
+        goto coreaudio_error_nounlock;
     }
 
-    if (!supports_digital) {
-        MP_ERR(ao, "selected device doesn't support digital formats\n");
-        goto coreaudio_error;
-    } // closes if (!supports_digital)
-
-    // Build ASBD for the input format
-    AudioStreamBasicDescription asbd;
-    ca_fill_asbd(ao, &asbd);
-
-    return init_digital(ao, asbd);
-
-coreaudio_error:
-    return CONTROL_ERROR;
-}
-
-static int init_digital(struct ao *ao, AudioStreamBasicDescription asbd)
-{
-    struct priv *p = ao->priv;
-    OSStatus err = noErr;
-
     uint32_t is_alive = 1;
     err = CA_GET(p->device, kAudioDevicePropertyDeviceIsAlive, &is_alive);
     CHECK_CA_WARN("could not check whether device is alive");
@@ -420,107 +277,69 @@ static int init_digital(struct ao *ao, AudioStreamBasicDescription asbd)
     err = ca_disable_mixing(ao, p->device, &p->changed_mixing);
     CHECK_CA_WARN("failed to disable mixing");
 
-    AudioStreamID *streams;
-    size_t n_streams;
-
-    /* Get a list of all the streams on this device. */
-    err = CA_GET_ARY_O(p->device, kAudioDevicePropertyStreams,
-                       &streams, &n_streams);
-
-    CHECK_CA_ERROR("could not get number of streams");
-
-    for (int i = 0; i < n_streams && p->stream_idx < 0; i++) {
-        bool digital = ca_stream_supports_digital(ao, streams[i]);
-
-        if (digital) {
-            err = CA_GET(streams[i], kAudioStreamPropertyPhysicalFormat,
-                         &p->original_asbd);
-            if (!CHECK_CA_WARN("could not get stream's physical format to "
-                               "revert to, getting the next one"))
-                continue;
-
-            AudioStreamRangedDescription *formats;
-            size_t n_formats;
-
-            err = CA_GET_ARY(streams[i],
-                             kAudioStreamPropertyAvailablePhysicalFormats,
-                             &formats, &n_formats);
-
-            if (!CHECK_CA_WARN("could not get number of stream formats"))
-                continue; // try next one
-
-            int req_rate_format = -1;
-            int max_rate_format = -1;
-
-            p->stream = streams[i];
-            p->stream_idx = i;
-
-            for (int j = 0; j < n_formats; j++)
-                if (ca_format_is_digital(formats[j].mFormat)) {
-                    // select the digital format that has exactly the same
-                    // samplerate. If an exact match cannot be found, select
-                    // the format with highest samplerate as backup.
-                    if (formats[j].mFormat.mSampleRate == asbd.mSampleRate) {
-                        req_rate_format = j;
-                        break;
-                    } else if (max_rate_format < 0 ||
-                        formats[j].mFormat.mSampleRate >
-                        formats[max_rate_format].mFormat.mSampleRate)
-                        max_rate_format = j;
-                }
-
-            if (req_rate_format >= 0)
-                p->stream_asbd = formats[req_rate_format].mFormat;
-            else
-                p->stream_asbd = formats[max_rate_format].mFormat;
-
-            talloc_free(formats);
-        }
-    }
-
-    talloc_free(streams);
-
-    if (p->stream_idx < 0) {
-        MP_WARN(ao , "can't find any digital output stream format\n");
-        goto coreaudio_error;
-    }
-
-    if (!ca_change_format(ao, p->stream, p->stream_asbd))
+    if (select_stream(ao) < 0)
         goto coreaudio_error;
 
-    void *changed = (void *) &(p->stream_asbd_changed);
-    err = ca_enable_device_listener(p->device, changed);
-    CHECK_CA_ERROR("cannot install format change listener during init");
+    AudioStreamBasicDescription hwfmt;
+    if (find_best_format(ao, &hwfmt) < 0)
+        goto coreaudio_error;
 
-    if (p->stream_asbd.mFormatFlags & kAudioFormatFlagIsBigEndian)
-        MP_WARN(ao, "stream has non-native byte order, output may fail\n");
+    err = CA_GET(p->stream, kAudioStreamPropertyPhysicalFormat,
+                 &p->original_asbd);
+    CHECK_CA_ERROR("could not get stream's original physical format");
+
+    // Even if changing the physical format fails, we can try using the current
+    // virtual format.
+    ca_change_physical_format_sync(ao, p->stream, hwfmt);
+
+    if (!ca_init_chmap(ao, p->device))
+        goto coreaudio_error;
+
+    err = CA_GET(p->stream, kAudioStreamPropertyVirtualFormat, &p->stream_asbd);
+    CHECK_CA_ERROR("could not get stream's virtual format");
+
+    ca_print_asbd(ao, "virtual format", &p->stream_asbd);
+
+    int new_format = ca_asbd_to_mp_format(&p->stream_asbd);
+
+    // If both old and new formats are spdif, avoid changing it due to the
+    // imperfect mapping between mp and CA formats.
+    if (!(af_fmt_is_spdif(ao->format) && af_fmt_is_spdif(new_format)))
+        ao->format = new_format;
+
+    if (!ao->format || af_fmt_is_planar(ao->format)) {
+        MP_ERR(ao, "hardware format not supported\n");
+        goto coreaudio_error;
+    }
 
     ao->samplerate = p->stream_asbd.mSampleRate;
-    ao->bps = ao->samplerate *
-                  (p->stream_asbd.mBytesPerPacket /
-                   p->stream_asbd.mFramesPerPacket);
 
-    uint32_t latency_frames = 0;
-    err = CA_GET_O(p->device, kAudioDevicePropertyLatency, &latency_frames);
-    if (err != noErr) {
-        CHECK_CA_WARN("cannot get device latency");
-        latency_frames = 0;
+    if (ao->channels.num != p->stream_asbd.mChannelsPerFrame) {
+        // We really expect that ca_init_chmap() fixes the layout to the HW's.
+        MP_ERR(ao, "number of channels changed, and unknown channel layout!\n");
+        goto coreaudio_error;
     }
 
-    p->hw_latency_us = ca_frames_to_us(ao, latency_frames);
+    p->hw_latency_us = ca_get_device_latency_us(ao, p->device);
     MP_VERBOSE(ao, "base latency: %d microseconds\n", (int)p->hw_latency_us);
 
+    err = enable_property_listener(ao, true);
+    CHECK_CA_ERROR("cannot install format change listener during init");
+
     err = AudioDeviceCreateIOProcID(p->device,
-                                    (AudioDeviceIOProc)render_cb_digital,
+                                    (AudioDeviceIOProc)render_cb_compressed,
                                     (void *)ao,
                                     &p->render_cb);
-    CHECK_CA_ERROR("failed to register digital render callback");
+    CHECK_CA_ERROR("failed to register audio render callback");
 
     return CONTROL_TRUE;
 
 coreaudio_error:
+    err = enable_property_listener(ao, false);
+    CHECK_CA_WARN("can't remove format change listener");
     err = ca_unlock_device(p->device, &p->hog_pid);
     CHECK_CA_WARN("can't release hog mode");
+coreaudio_error_nounlock:
     return CONTROL_ERROR;
 }
 
@@ -529,8 +348,7 @@ static void uninit(struct ao *ao)
     struct priv *p = ao->priv;
     OSStatus err = noErr;
 
-    void *changed = (void *) &(p->stream_asbd_changed);
-    err = ca_disable_device_listener(p->device, changed);
+    err = enable_property_listener(ao, false);
     CHECK_CA_WARN("can't remove device listener, this may cause a crash");
 
     err = AudioDeviceStop(p->device, p->render_cb);
@@ -539,7 +357,7 @@ static void uninit(struct ao *ao)
     err = AudioDeviceDestroyIOProcID(p->device, p->render_cb);
     CHECK_CA_WARN("failed to remove device render callback");
 
-    if (!ca_change_format(ao, p->stream, p->original_asbd))
+    if (!ca_change_physical_format_sync(ao, p->stream, p->original_asbd))
         MP_WARN(ao, "can't revert to original device format");
 
     err = ca_enable_mixing(ao, p->device, p->changed_mixing);
@@ -554,7 +372,7 @@ static void audio_pause(struct ao *ao)
     struct priv *p = ao->priv;
 
     OSStatus err = AudioDeviceStop(p->device, p->render_cb);
-    CHECK_CA_WARN("can't stop digital device");
+    CHECK_CA_WARN("can't stop audio device");
 }
 
 static void audio_resume(struct ao *ao)
@@ -562,7 +380,7 @@ static void audio_resume(struct ao *ao)
     struct priv *p = ao->priv;
 
     OSStatus err = AudioDeviceStart(p->device, p->render_cb);
-    CHECK_CA_WARN("can't start digital device");
+    CHECK_CA_WARN("can't start audio device");
 }
 
 #define OPT_BASE_STRUCT struct priv
@@ -577,7 +395,6 @@ const struct ao_driver audio_out_coreaudio_exclusive = {
     .list_devs = ca_get_device_list,
     .priv_size = sizeof(struct priv),
     .priv_defaults = &(const struct priv){
-        .stream_asbd_changed = 0,
         .hog_pid = -1,
         .stream = 0,
         .stream_idx = -1,

audio/out/ao_coreaudio_utils.c

@@ -28,6 +28,7 @@
 #include "audio/out/ao_coreaudio_properties.h"
 #include "osdep/timer.h"
 #include "osdep/endian.h"
+#include "osdep/semaphore.h"
 #include "audio/format.h"
 
 CFStringRef cfstr_from_cstr(char *str)
@@ -88,7 +89,7 @@ OSStatus ca_select_device(struct ao *ao, char* name, AudioDeviceID *device)
     OSStatus err = noErr;
     *device = kAudioObjectUnknown;
 
-    if (name) {
+    if (name && name[0]) {
         CFStringRef uid = cfstr_from_cstr(name);
         AudioValueTranslation v = (AudioValueTranslation) {
             .mInputData = &uid,
@@ -116,9 +117,8 @@ OSStatus ca_select_device(struct ao *ao, char* name, AudioDeviceID *device)
 
     if (mp_msg_test(ao->log, MSGL_V)) {
         char *desc;
-        err = CA_GET_STR(*device, kAudioObjectPropertyName, &desc);
-        CHECK_CA_WARN("could not get selected audio device name");
-        if (err == noErr) {
+        OSStatus err2 = CA_GET_STR(*device, kAudioObjectPropertyName, &desc);
+        if (err2 == noErr) {
             MP_VERBOSE(ao, "selected audio output device: %s (%" PRIu32 ")\n",
                            desc, *device);
             talloc_free(desc);
@@ -129,7 +129,7 @@ coreaudio_error:
     return err;
 }
 
-char *fourcc_repr(void *talloc_ctx, uint32_t code)
+char *fourcc_repr_buf(char *buf, size_t buf_size, uint32_t code)
 {
     // Extract FourCC letters from the uint32_t and finde out if it's a valid
     // code that is made of letters.
@@ -146,64 +146,119 @@ char *fourcc_repr(void *talloc_ctx, uint32_t code)
             valid_fourcc = false;
     }
 
-    char *repr;
     if (valid_fourcc)
-        repr = talloc_asprintf(talloc_ctx, "'%c%c%c%c'",
-                               fcc[0], fcc[1], fcc[2], fcc[3]);
+        snprintf(buf, buf_size, "'%c%c%c%c'", fcc[0], fcc[1], fcc[2], fcc[3]);
     else
-        repr = talloc_asprintf(NULL, "%u", (unsigned int)code);
+        snprintf(buf, buf_size, "%u", (unsigned int)code);
 
-    return repr;
+    return buf;
 }
 
 bool check_ca_st(struct ao *ao, int level, OSStatus code, const char *message)
 {
     if (code == noErr) return true;
 
-    char *error_string = fourcc_repr(NULL, code);
-    mp_msg(ao->log, level, "%s (%s)\n", message, error_string);
-    talloc_free(error_string);
+    mp_msg(ao->log, level, "%s (%s)\n", message, fourcc_repr(code));
 
     return false;
 }
 
-void ca_fill_asbd(struct ao *ao, AudioStreamBasicDescription *asbd)
+static void ca_fill_asbd_raw(AudioStreamBasicDescription *asbd, int mp_format,
+                             int samplerate, int num_channels)
 {
-    asbd->mSampleRate       = ao->samplerate;
+    asbd->mSampleRate       = samplerate;
     // Set "AC3" for other spdif formats too - unknown if that works.
-    asbd->mFormatID         = AF_FORMAT_IS_IEC61937(ao->format) ?
+    asbd->mFormatID         = af_fmt_is_spdif(mp_format) ?
                               kAudioFormat60958AC3 :
                               kAudioFormatLinearPCM;
-    asbd->mChannelsPerFrame = ao->channels.num;
-    asbd->mBitsPerChannel   = af_fmt2bits(ao->format);
+    asbd->mChannelsPerFrame = num_channels;
+    asbd->mBitsPerChannel   = af_fmt_to_bytes(mp_format) * 8;
     asbd->mFormatFlags      = kAudioFormatFlagIsPacked;
 
-    if ((ao->format & AF_FORMAT_TYPE_MASK) == AF_FORMAT_F)
-        asbd->mFormatFlags |= kAudioFormatFlagIsFloat;
+    int channels_per_buffer = num_channels;
+    if (af_fmt_is_planar(mp_format)) {
+        asbd->mFormatFlags |= kAudioFormatFlagIsNonInterleaved;
+        channels_per_buffer = 1;
+    }
 
-    if ((ao->format & AF_FORMAT_SIGN_MASK) == AF_FORMAT_SI)
+    if (af_fmt_is_float(mp_format)) {
+        asbd->mFormatFlags |= kAudioFormatFlagIsFloat;
+    } else if (!af_fmt_is_unsigned(mp_format)) {
         asbd->mFormatFlags |= kAudioFormatFlagIsSignedInteger;
+    }
 
     if (BYTE_ORDER == BIG_ENDIAN)
         asbd->mFormatFlags |= kAudioFormatFlagIsBigEndian;
 
     asbd->mFramesPerPacket = 1;
     asbd->mBytesPerPacket = asbd->mBytesPerFrame =
-        asbd->mFramesPerPacket * asbd->mChannelsPerFrame *
+        asbd->mFramesPerPacket * channels_per_buffer *
         (asbd->mBitsPerChannel / 8);
 }
 
+void ca_fill_asbd(struct ao *ao, AudioStreamBasicDescription *asbd)
+{
+    ca_fill_asbd_raw(asbd, ao->format, ao->samplerate, ao->channels.num);
+}
+
+bool ca_formatid_is_compressed(uint32_t formatid)
+{
+    switch (formatid)
+    case 'IAC3':
+    case 'iac3':
+    case  kAudioFormat60958AC3:
+    case  kAudioFormatAC3:
+        return true;
+    return false;
+}
+
+// This might be wrong, but for now it's sufficient for us.
+static uint32_t ca_normalize_formatid(uint32_t formatID)
+{
+    return ca_formatid_is_compressed(formatID) ? kAudioFormat60958AC3 : formatID;
+}
+
+bool ca_asbd_equals(const AudioStreamBasicDescription *a,
+                    const AudioStreamBasicDescription *b)
+{
+    int flags = kAudioFormatFlagIsPacked | kAudioFormatFlagIsFloat |
+                kAudioFormatFlagIsSignedInteger | kAudioFormatFlagIsBigEndian;
+    bool spdif = ca_formatid_is_compressed(a->mFormatID) &&
+                 ca_formatid_is_compressed(b->mFormatID);
+
+    return (a->mFormatFlags & flags) == (b->mFormatFlags & flags) &&
+           a->mBitsPerChannel == b->mBitsPerChannel &&
+           ca_normalize_formatid(a->mFormatID) ==
+                ca_normalize_formatid(b->mFormatID) &&
+           (spdif || a->mBytesPerPacket == b->mBytesPerPacket) &&
+           (spdif || a->mChannelsPerFrame == b->mChannelsPerFrame) &&
+           a->mSampleRate == b->mSampleRate;
+}
+
+// Return the AF_FORMAT_* (AF_FORMAT_S16 etc.) corresponding to the asbd.
+int ca_asbd_to_mp_format(const AudioStreamBasicDescription *asbd)
+{
+    for (int fmt = 1; fmt < AF_FORMAT_COUNT; fmt++) {
+        AudioStreamBasicDescription mp_asbd = {0};
+        ca_fill_asbd_raw(&mp_asbd, fmt, asbd->mSampleRate, asbd->mChannelsPerFrame);
+        if (ca_asbd_equals(&mp_asbd, asbd))
+            return af_fmt_is_spdif(fmt) ? AF_FORMAT_S_AC3 : fmt;
+    }
+    return 0;
+}
+
 void ca_print_asbd(struct ao *ao, const char *description,
                    const AudioStreamBasicDescription *asbd)
 {
     uint32_t flags  = asbd->mFormatFlags;
-    char *format    = fourcc_repr(NULL, asbd->mFormatID);
+    char *format    = fourcc_repr(asbd->mFormatID);
+    int mpfmt       = ca_asbd_to_mp_format(asbd);
 
     MP_VERBOSE(ao,
-       "%s %7.1fHz %" PRIu32 "bit [%s]"
-       "[%" PRIu32 "][%" PRIu32 "][%" PRIu32 "]"
-       "[%" PRIu32 "][%" PRIu32 "] "
-       "%s %s %s%s%s%s\n",
+       "%s %7.1fHz %" PRIu32 "bit %s "
+       "[%" PRIu32 "][%" PRIu32 "bpp][%" PRIu32 "fbp]"
+       "[%" PRIu32 "bpf][%" PRIu32 "ch] "
+       "%s %s %s%s%s%s (%s)\n",
        description, asbd->mSampleRate, asbd->mBitsPerChannel, format,
        asbd->mFormatFlags, asbd->mBytesPerPacket, asbd->mFramesPerPacket,
        asbd->mBytesPerFrame, asbd->mChannelsPerFrame,
@@ -212,9 +267,49 @@ void ca_print_asbd(struct ao *ao, const char *description,
        (flags & kAudioFormatFlagIsSignedInteger) ? "S" : "U",
        (flags & kAudioFormatFlagIsPacked) ? " packed" : "",
        (flags & kAudioFormatFlagIsAlignedHigh) ? " aligned" : "",
-       (flags & kAudioFormatFlagIsNonInterleaved) ? " P" : "");
+       (flags & kAudioFormatFlagIsNonInterleaved) ? " P" : "",
+       mpfmt ? af_fmt_to_str(mpfmt) : "-");
+}
 
-    talloc_free(format);
+// Return whether new is an improvement over old. Assume a higher value means
+// better quality, and we always prefer the value closest to the requested one,
+// which is still larger than the requested one.
+// Equal values prefer the new one (so ca_asbd_is_better() checks other params).
+static bool value_is_better(double req, double old, double new)
+{
+    if (new >= req) {
+        return old < req || new <= old;
+    } else {
+        return old < req && new >= old;
+    }
+}
+
+// Return whether new is an improvement over old (req is the requested format).
+bool ca_asbd_is_better(AudioStreamBasicDescription *req,
+                       AudioStreamBasicDescription *old,
+                       AudioStreamBasicDescription *new)
+{
+    if (new->mChannelsPerFrame > MP_NUM_CHANNELS)
+        return false;
+    if (old->mChannelsPerFrame > MP_NUM_CHANNELS)
+        return true;
+    if (req->mFormatID != new->mFormatID)
+        return false;
+    if (req->mFormatID != old->mFormatID)
+        return true;
+
+    if (!value_is_better(req->mBitsPerChannel, old->mBitsPerChannel,
+                         new->mBitsPerChannel))
+        return false;
+
+    if (!value_is_better(req->mSampleRate, old->mSampleRate, new->mSampleRate))
+        return false;
+
+    if (!value_is_better(req->mChannelsPerFrame, old->mChannelsPerFrame,
+                         new->mChannelsPerFrame))
+        return false;
+
+    return true;
 }
 
 int64_t ca_frames_to_us(struct ao *ao, uint32_t frames)
@@ -232,3 +327,229 @@ int64_t ca_get_latency(const AudioTimeStamp *ts)
 
     return (out - now) * 1e-3;
 }
+
+bool ca_stream_supports_compressed(struct ao *ao, AudioStreamID stream)
+{
+    AudioStreamRangedDescription *formats = NULL;
+    size_t n_formats;
+
+    OSStatus err =
+        CA_GET_ARY(stream, kAudioStreamPropertyAvailablePhysicalFormats,
+                   &formats, &n_formats);
+
+    CHECK_CA_ERROR("Could not get number of stream formats.");
+
+    for (int i = 0; i < n_formats; i++) {
+        AudioStreamBasicDescription asbd = formats[i].mFormat;
+        if (ca_formatid_is_compressed(asbd.mFormatID)) {
+            talloc_free(formats);
+            return true;
+        }
+    }
+
+    talloc_free(formats);
+coreaudio_error:
+    return false;
+}
+
+bool ca_device_supports_compressed(struct ao *ao, AudioDeviceID device)
+{
+    AudioStreamID *streams = NULL;
+    size_t n_streams;
+
+    /* Retrieve all the output streams. */
+    OSStatus err =
+        CA_GET_ARY_O(device, kAudioDevicePropertyStreams, &streams, &n_streams);
+
+    CHECK_CA_ERROR("could not get number of streams.");
+
+    for (int i = 0; i < n_streams; i++) {
+        if (ca_stream_supports_compressed(ao, streams[i])) {
+            talloc_free(streams);
+            return true;
+        }
+    }
+
+    talloc_free(streams);
+
+coreaudio_error:
+    return false;
+}
+
+OSStatus ca_lock_device(AudioDeviceID device, pid_t *pid)
+{
+    *pid = getpid();
+    OSStatus err = CA_SET(device, kAudioDevicePropertyHogMode, pid);
+    if (err != noErr)
+        *pid = -1;
+
+    return err;
+}
+
+OSStatus ca_unlock_device(AudioDeviceID device, pid_t *pid)
+{
+    if (*pid == getpid()) {
+        *pid = -1;
+        return CA_SET(device, kAudioDevicePropertyHogMode, &pid);
+    }
+    return noErr;
+}
+
+static OSStatus ca_change_mixing(struct ao *ao, AudioDeviceID device,
+                                 uint32_t val, bool *changed)
+{
+    *changed = false;
+
+    AudioObjectPropertyAddress p_addr = (AudioObjectPropertyAddress) {
+        .mSelector = kAudioDevicePropertySupportsMixing,
+        .mScope    = kAudioObjectPropertyScopeGlobal,
+        .mElement  = kAudioObjectPropertyElementMaster,
+    };
+
+    if (AudioObjectHasProperty(device, &p_addr)) {
+        OSStatus err;
+        Boolean writeable = 0;
+        err = CA_SETTABLE(device, kAudioDevicePropertySupportsMixing,
+                          &writeable);
+
+        if (!CHECK_CA_WARN("can't tell if mixing property is settable")) {
+            return err;
+        }
+
+        if (!writeable)
+            return noErr;
+
+        err = CA_SET(device, kAudioDevicePropertySupportsMixing, &val);
+        if (err != noErr)
+            return err;
+
+        if (!CHECK_CA_WARN("can't set mix mode")) {
+            return err;
+        }
+
+        *changed = true;
+    }
+
+    return noErr;
+}
+
+OSStatus ca_disable_mixing(struct ao *ao, AudioDeviceID device, bool *changed)
+{
+    return ca_change_mixing(ao, device, 0, changed);
+}
+
+OSStatus ca_enable_mixing(struct ao *ao, AudioDeviceID device, bool changed)
+{
+    if (changed) {
+        bool dont_care = false;
+        return ca_change_mixing(ao, device, 1, &dont_care);
+    }
+
+    return noErr;
+}
+
+int64_t ca_get_device_latency_us(struct ao *ao, AudioDeviceID device)
+{
+    uint32_t latency_frames = 0;
+    uint32_t latency_properties[] = {
+        kAudioDevicePropertyLatency,
+        kAudioDevicePropertyBufferFrameSize,
+        kAudioDevicePropertySafetyOffset,
+    };
+    for (int n = 0; n < MP_ARRAY_SIZE(latency_properties); n++) {
+        uint32_t temp;
+        OSStatus err = CA_GET_O(device, latency_properties[n], &temp);
+        CHECK_CA_WARN("cannot get device latency");
+        if (err == noErr) {
+            latency_frames += temp;
+            MP_VERBOSE(ao, "Latency property %s: %d frames\n",
+                       fourcc_repr(latency_properties[n]), (int)temp);
+        }
+    }
+
+    return ca_frames_to_us(ao, latency_frames);
+}
+
+static OSStatus ca_change_format_listener(
+    AudioObjectID object, uint32_t n_addresses,
+    const AudioObjectPropertyAddress addresses[],
+    void *data)
+{
+    sem_t *sem = data;
+    sem_post(sem);
+    return noErr;
+}
+
+bool ca_change_physical_format_sync(struct ao *ao, AudioStreamID stream,
+                                    AudioStreamBasicDescription change_format)
+{
+    OSStatus err = noErr;
+    bool format_set = false;
+
+    ca_print_asbd(ao, "setting stream physical format:", &change_format);
+
+    sem_t wakeup;
+    if (sem_init(&wakeup, 0, 0)) {
+        MP_WARN(ao, "OOM\n");
+        return false;
+    }
+
+    AudioStreamBasicDescription prev_format;
+    err = CA_GET(stream, kAudioStreamPropertyPhysicalFormat, &prev_format);
+    CHECK_CA_ERROR("can't get current physical format");
+
+    /* Install the callback. */
+    AudioObjectPropertyAddress p_addr = {
+        .mSelector = kAudioStreamPropertyPhysicalFormat,
+        .mScope    = kAudioObjectPropertyScopeGlobal,
+        .mElement  = kAudioObjectPropertyElementMaster,
+    };
+
+    err = AudioObjectAddPropertyListener(stream, &p_addr,
+                                         ca_change_format_listener,
+                                         &wakeup);
+    CHECK_CA_ERROR("can't add property listener during format change");
+
+    /* Change the format. */
+    err = CA_SET(stream, kAudioStreamPropertyPhysicalFormat, &change_format);
+    CHECK_CA_WARN("error changing physical format");
+
+    /* The AudioStreamSetProperty is not only asynchronous,
+     * it is also not Atomic, in its behaviour. */
+    struct timespec timeout = mp_rel_time_to_timespec(0.5);
+    AudioStreamBasicDescription actual_format = {0};
+    while (1) {
+        err = CA_GET(stream, kAudioStreamPropertyPhysicalFormat, &actual_format);
+        if (!CHECK_CA_WARN("could not retrieve physical format"))
+            break;
+
+        format_set = ca_asbd_equals(&change_format, &actual_format);
+        if (format_set)
+            break;
+
+        if (sem_timedwait(&wakeup, &timeout)) {
+            MP_VERBOSE(ao, "reached timeout\n");
+            break;
+        }
+    }
+
+    ca_print_asbd(ao, "actual format in use:", &actual_format);
+
+    if (!format_set) {
+        MP_WARN(ao, "changing physical format failed\n");
+        // Some drivers just fuck up and get into a broken state. Restore the
+        // old format in this case.
+        err = CA_SET(stream, kAudioStreamPropertyPhysicalFormat, &prev_format);
+        CHECK_CA_WARN("error restoring physical format");
+    }
+
+    err = AudioObjectRemovePropertyListener(stream, &p_addr,
+                                            ca_change_format_listener,
+                                            &wakeup);
+    CHECK_CA_ERROR("can't remove property listener");
+
+coreaudio_error:
+    sem_destroy(&wakeup);
+    return format_set;
+}
+

audio/out/ao_coreaudio_utils.h

@@ -31,7 +31,9 @@
 CFStringRef cfstr_from_cstr(char *str);
 char *cfstr_get_cstr(CFStringRef cfstr);
 
-char *fourcc_repr(void *talloc_ctx, uint32_t code);
+char *fourcc_repr_buf(char *buf, size_t buf_size, uint32_t code);
+#define fourcc_repr(code) fourcc_repr_buf((char[40]){0}, 40, code)
+
 bool check_ca_st(struct ao *ao, int level, OSStatus code, const char *message);
 
 #define CHECK_CA_ERROR_L(label, message) \
@@ -52,11 +54,28 @@ bool check_ca_st(struct ao *ao, int level, OSStatus code, const char *message);
 void ca_get_device_list(struct ao *ao, struct ao_device_list *list);
 OSStatus ca_select_device(struct ao *ao, char* name, AudioDeviceID *device);
 
+bool ca_formatid_is_compressed(uint32_t formatid);
 void ca_fill_asbd(struct ao *ao, AudioStreamBasicDescription *asbd);
 void ca_print_asbd(struct ao *ao, const char *description,
                    const AudioStreamBasicDescription *asbd);
+bool ca_asbd_equals(const AudioStreamBasicDescription *a,
+                    const AudioStreamBasicDescription *b);
+int ca_asbd_to_mp_format(const AudioStreamBasicDescription *asbd);
+bool ca_asbd_is_better(AudioStreamBasicDescription *req,
+                       AudioStreamBasicDescription *old,
+                       AudioStreamBasicDescription *new);
 
 int64_t ca_frames_to_us(struct ao *ao, uint32_t frames);
 int64_t ca_get_latency(const AudioTimeStamp *ts);
 
+bool ca_device_supports_compressed(struct ao *ao, AudioDeviceID device);
+bool ca_stream_supports_compressed(struct ao *ao, AudioStreamID stream);
+OSStatus ca_lock_device(AudioDeviceID device, pid_t *pid);
+OSStatus ca_unlock_device(AudioDeviceID device, pid_t *pid);
+OSStatus ca_disable_mixing(struct ao *ao, AudioDeviceID device, bool *changed);
+OSStatus ca_enable_mixing(struct ao *ao, AudioDeviceID device, bool changed);
+int64_t ca_get_device_latency_us(struct ao *ao, AudioDeviceID device);
+bool ca_change_physical_format_sync(struct ao *ao, AudioStreamID stream,
+                                    AudioStreamBasicDescription change_format);
+
 #endif /* MPV_COREAUDIO_UTILS_H */

audio/out/ao_dsound.c

@@ -288,7 +288,7 @@ static int InitDirectSound(struct ao *ao)
     /* Set DirectSound Cooperative level, ie what control we want over Windows
      * sound device. In our case, DSSCL_EXCLUSIVE means that we can modify the
      * settings of the primary buffer, but also that only the sound of our
-     * application will be hearable when it will have the focus.
+     * application will be audible when it will have the focus.
      * !!! (this is not really working as intended yet because to set the
      * cooperative level you need the window handle of your application, and
      * I don't know of any easy way to get it. Especially since we might play
@@ -444,7 +444,7 @@ static int init(struct ao *ao)
     int format = af_fmt_from_planar(ao->format);
     int rate = ao->samplerate;
 
-    if (!AF_FORMAT_IS_IEC61937(format)) {
+    if (!af_fmt_is_spdif(format)) {
         struct mp_chmap_sel sel = {0};
         mp_chmap_sel_add_waveext(&sel);
         if (!ao_chmap_sel_adjust(ao, &sel, &ao->channels))
@@ -456,7 +456,7 @@ static int init(struct ao *ao)
     case AF_FORMAT_U8:
         break;
     default:
-        if (AF_FORMAT_IS_IEC61937(format))
+        if (af_fmt_is_spdif(format))
             break;
         MP_VERBOSE(ao, "format %s not supported defaulting to Signed 16-bit Little-Endian\n",
                    af_fmt_to_str(format));
@@ -465,7 +465,7 @@ static int init(struct ao *ao)
     //set our audio parameters
     ao->samplerate = rate;
     ao->format = format;
-    ao->bps = ao->channels.num * rate * af_fmt2bps(format);
+    ao->bps = ao->channels.num * rate * af_fmt_to_bytes(format);
     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));
@@ -478,7 +478,7 @@ 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_IEC61937(format)) {
+    if (af_fmt_is_spdif(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;
@@ -486,7 +486,7 @@ static int init(struct ao *ao)
     } else {
         wformat.Format.wFormatTag = (ao->channels.num > 2)
                                     ? WAVE_FORMAT_EXTENSIBLE : WAVE_FORMAT_PCM;
-        int bps = af_fmt2bps(format);
+        int bps = af_fmt_to_bytes(format);
         wformat.Format.wBitsPerSample = bps * 8;
         wformat.Format.nBlockAlign = wformat.Format.nChannels * bps;
     }
@@ -616,9 +616,9 @@ static int check_free_buffer_size(struct ao *ao)
     space = p->buffer_size - (p->write_offset - play_offset);
     // |              | <-- const --> |                |                 |
     // buffer start   play_cursor     write_cursor     p->write_offset   buffer end
-    // play_cursor is the actual postion of the play cursor
+    // play_cursor is the actual position of the play cursor
     // write_cursor is the position after which it is assumed to be save to write data
-    // p->write_offset is the postion where we actually write the data to
+    // p->write_offset is the position where we actually write the data to
     if (space > p->buffer_size)
         space -= p->buffer_size;                        // p->write_offset < play_offset
     // Check for buffer underruns. An underrun happens if DirectSound

audio/out/ao_lavc.c

@@ -25,7 +25,6 @@
 #include <limits.h>
 
 #include <libavutil/common.h>
-#include <libavutil/audioconvert.h>
 
 #include "config.h"
 #include "options/options.h"
@@ -136,7 +135,7 @@ static int init(struct ao *ao)
 
     select_format(ao, codec);
 
-    ac->sample_size = af_fmt2bps(ao->format);
+    ac->sample_size = af_fmt_to_bytes(ao->format);
     ac->stream->codec->sample_fmt = af_to_avformat(ao->format);
     ac->stream->codec->bits_per_raw_sample = ac->sample_size * 8;
 
@@ -241,7 +240,7 @@ static int encode(struct ao *ao, double apts, void **data)
         frame->format = af_to_avformat(ao->format);
         frame->nb_samples = ac->aframesize;
 
-        size_t num_planes = AF_FORMAT_IS_PLANAR(ao->format) ? ao->channels.num : 1;
+        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];
@@ -350,7 +349,7 @@ static int play(struct ao *ao, void **data, int samples, int flags)
     double pts = ectx->last_audio_in_pts;
     pts += ectx->samples_since_last_pts / (double)ao->samplerate;
 
-    size_t num_planes = AF_FORMAT_IS_PLANAR(ao->format) ? ao->channels.num : 1;
+    size_t num_planes = af_fmt_is_planar(ao->format) ? ao->channels.num : 1;
 
     void *tempdata = NULL;
     void *padded[MP_NUM_CHANNELS];

audio/out/ao_null.c

@@ -56,6 +56,8 @@ struct priv {
     // called with sizes not aligned to this, a rounded size will be returned.
     // (This is not needed by the AO API, but many AOs behave this way.)
     int outburst;       // samples
+
+    char **channel_layouts;
 };
 
 static void drain(struct ao *ao)
@@ -88,8 +90,19 @@ static int init(struct ao *ao)
 
     ao->untimed = priv->untimed;
 
-    struct mp_chmap_sel sel = {0};
-    mp_chmap_sel_add_any(&sel);
+    struct mp_chmap_sel sel = {.tmp = ao};
+    if (priv->channel_layouts) {
+        for (int n = 0; priv->channel_layouts[n]; n++) {
+            struct mp_chmap map = {0};
+            if (!mp_chmap_from_str(&map, bstr0(priv->channel_layouts[n]))) {
+                MP_FATAL(ao, "Invalid channel map in option.\n");
+                return -1;
+            }
+            mp_chmap_sel_add_map(&sel, &map);
+        }
+    } else {
+        mp_chmap_sel_add_any(&sel);
+    }
     if (!ao_chmap_sel_adjust(ao, &sel, &ao->channels))
         mp_chmap_from_channels(&ao->channels, 2);
 
@@ -231,6 +244,7 @@ const struct ao_driver audio_out_null = {
         OPT_FLOATRANGE("latency", latency_sec, 0, 0, 100),
         OPT_FLAG("broken-eof", broken_eof, 0),
         OPT_FLAG("broken-delay", broken_delay, 0),
+        OPT_STRINGLIST("channel-layouts", channel_layouts, 0),
         {0}
     },
 };

audio/out/ao_oss.c

@@ -94,40 +94,20 @@ static const struct mp_chmap oss_layouts[MP_NUM_CHANNELS + 1] = {
 #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_U32_NE 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_S32_NE 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_NE,       AF_FORMAT_U16},
     {AFMT_S16_NE,       AF_FORMAT_S16},
-#ifdef AFMT_U24_NE
-    {AFMT_U24_NE,       AF_FORMAT_U24},
-#endif
 #ifdef AFMT_S24_NE
     {AFMT_S24_NE,       AF_FORMAT_S24},
 #endif
-#ifdef AFMT_U32_NE
-    {AFMT_U32_NE,       AF_FORMAT_U32},
-#endif
 #ifdef AFMT_S32_NE
     {AFMT_S32_NE,       AF_FORMAT_S32},
 #endif
@@ -204,7 +184,7 @@ static int control(struct ao *ao, enum aocontrol cmd, void *arg)
             return CONTROL_OK;
 #endif
 
-        if (AF_FORMAT_IS_SPECIAL(ao->format))
+        if (!af_fmt_is_pcm(ao->format))
             return CONTROL_TRUE;
 
         if ((fd = open(p->oss_mixer_device, O_RDONLY)) != -1) {
@@ -267,7 +247,7 @@ 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))
+    if (oss_format == -1 && af_fmt_is_spdif(*format))
         oss_format = AFMT_AC3;
 
     if (oss_format == -1) {
@@ -323,7 +303,7 @@ static int reopen_device(struct ao *ao, bool allow_format_changes)
     fcntl(p->audio_fd, F_SETFD, FD_CLOEXEC);
 #endif
 
-    if (AF_FORMAT_IS_IEC61937(format)) {
+    if (af_fmt_is_spdif(format)) {
         if (ioctl(p->audio_fd, SNDCTL_DSP_SPEED, &samplerate) == -1)
             goto fail;
         // Probably could be fixed by setting number of channels; needs testing.
@@ -347,7 +327,7 @@ static int reopen_device(struct ao *ao, bool allow_format_changes)
 
     MP_VERBOSE(ao, "sample format: %s\n", af_fmt_to_str(format));
 
-    if (!AF_FORMAT_IS_IEC61937(format)) {
+    if (!af_fmt_is_spdif(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]);
@@ -412,7 +392,7 @@ static int reopen_device(struct ao *ao, bool allow_format_changes)
         }
     }
 
-    p->outburst -= p->outburst % (channels.num * af_fmt2bps(format)); // round down
+    p->outburst -= p->outburst % (channels.num * af_fmt_to_bytes(format)); // round down
 
     return 0;
 

audio/out/ao_pcm.c

@@ -73,7 +73,7 @@ static void fput32le(uint32_t val, FILE *fp)
 static void write_wave_header(struct ao *ao, FILE *fp, uint64_t data_length)
 {
     uint16_t fmt = ao->format == AF_FORMAT_FLOAT ? WAV_ID_FLOAT_PCM : WAV_ID_PCM;
-    int bits = af_fmt2bits(ao->format);
+    int bits = af_fmt_to_bytes(ao->format) * 8;
 
     // Master RIFF chunk
     fput32le(WAV_ID_RIFF, fp);
@@ -135,7 +135,7 @@ static int init(struct ao *ao)
         case AF_FORMAT_FLOAT:
              break;
         default:
-            if (!AF_FORMAT_IS_IEC61937(ao->format))
+            if (!af_fmt_is_spdif(ao->format))
                 ao->format = AF_FORMAT_S16;
             break;
         }
@@ -146,7 +146,7 @@ static int init(struct ao *ao)
     if (!ao_chmap_sel_adjust(ao, &sel, &ao->channels))
         return -1;
 
-    ao->bps = ao->channels.num * ao->samplerate * af_fmt2bps(ao->format);
+    ao->bps = ao->channels.num * ao->samplerate * af_fmt_to_bytes(ao->format);
 
     MP_INFO(ao, "File: %s (%s)\nPCM: Samplerate: %d Hz Channels: %d Format: %s\n",
             priv->outputfilename,

audio/out/ao_pulse.c

@@ -210,7 +210,7 @@ static pa_encoding_t map_digital_format(int format)
     case AF_FORMAT_S_AAC:   return PA_ENCODING_MPEG2_AAC_IEC61937;
 #endif
     default:
-        if (AF_FORMAT_IS_IEC61937(format))
+        if (af_fmt_is_spdif(format))
             return PA_ENCODING_ANY;
         return PA_ENCODING_PCM;
     }

audio/out/ao_rsound.c

@@ -48,26 +48,9 @@ static int set_format(struct ao *ao)
     case AF_FORMAT_U8:
         rsd_format = RSD_U8;
         break;
-    case AF_FORMAT_S8:
-        rsd_format = RSD_S8;
-        break;
-    case AF_FORMAT_S16:
-        rsd_format = RSD_S16_NE;
-        break;
-    case AF_FORMAT_U16:
-        rsd_format = RSD_U16_NE;
-        break;
-    case AF_FORMAT_S24:
-    case AF_FORMAT_U24:
-        rsd_format = RSD_S32_NE;
-        ao->format = AF_FORMAT_S32;
-        break;
     case AF_FORMAT_S32:
         rsd_format = RSD_S32_NE;
         break;
-    case AF_FORMAT_U32:
-        rsd_format = RSD_U32_NE;
-        break;
     default:
         rsd_format = RSD_S16_NE;
         ao->format = AF_FORMAT_S16;

audio/out/ao_sdl.c

@@ -39,8 +39,6 @@ struct priv
 
 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},

audio/out/ao_sndio.c

@@ -98,7 +98,8 @@ static const struct mp_chmap sndio_layouts[MP_NUM_CHANNELS + 1] = {
     {0},                                        // 6.1
     MP_CHMAP8(FL, FR, BL, BR, FC, LFE, SL, SR), // 7.1
     /* above is the fixed channel assignment for sndio, since we need to fill
-       all channels and cannot insert silence, not all layouts are supported. */
+       all channels and cannot insert silence, not all layouts are supported.
+       NOTE: MP_SPEAKER_ID_NA could be used to add padding channels. */
 };
 
 /*
@@ -114,12 +115,8 @@ static int init(struct ao *ao)
     };
     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;
@@ -176,14 +173,14 @@ static int init(struct ao *ao)
         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 ? AF_FORMAT_S16 : AF_FORMAT_U16;
-    } else if ((p->par.bits == 24 || p->par.msb) && p->par.bps == 3) {
-        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 ? AF_FORMAT_S32 : AF_FORMAT_U32;
+    if (p->par.bits == 8 && p->par.bps == 1 && !p->par.sig) {
+        ao->format = AF_FORMAT_U8;
+    } else if (p->par.bits == 16 && p->par.bps == 2 && p->par.sig) {
+        ao->format = AF_FORMAT_S16;
+    } else if ((p->par.bits == 24 || p->par.msb) && p->par.bps == 3 && p->par.sig) {
+        ao->format = AF_FORMAT_S24;
+    } else if ((p->par.bits == 32 || p->par.msb) && p->par.bps == 4 && p->par.sig) {
+        ao->format = AF_FORMAT_S32;
     } else {
         MP_ERR(ao, "couldn't set format\n");
         goto error;

audio/out/ao_wasapi.c

@@ -390,6 +390,7 @@ static int hotplug_init(struct ao *ao)
 {
     MP_DBG(ao, "Hotplug init\n");
     struct wasapi_state *state = ao->priv;
+    state->log = ao->log;
     CoInitializeEx(NULL, COINIT_APARTMENTTHREADED);
     HRESULT hr = CoCreateInstance(&CLSID_MMDeviceEnumerator, NULL, CLSCTX_ALL,
                                   &IID_IMMDeviceEnumerator, (void **)&state->pEnumerator);

audio/out/ao_wasapi_utils.c

@@ -85,13 +85,13 @@ const struct wasapi_fmt_mapping wasapi_fmt_table[] = {
 
 static const GUID *format_to_subtype(int format)
 {
-    if (AF_FORMAT_IS_SPECIAL(format)) {
+    if (af_fmt_is_spdif(format)) {
         for (int i = 0; wasapi_fmt_table[i].format; i++) {
             if (wasapi_fmt_table[i].format == format)
                 return wasapi_fmt_table[i].subtype;
         }
         return &KSDATAFORMAT_SPECIFIER_NONE;
-    } else if (AF_FORMAT_IS_FLOAT(format)) {
+    } else if (af_fmt_is_float(format)) {
         return &KSDATAFORMAT_SUBTYPE_IEEE_FLOAT;
     }
     return &KSDATAFORMAT_SUBTYPE_PCM;
@@ -229,7 +229,7 @@ static void set_waveformat(WAVEFORMATEXTENSIBLE *wformat,
     wformat->Format.wFormatTag     = WAVE_FORMAT_EXTENSIBLE;
     wformat->Format.nChannels      = channels->num;
     wformat->Format.nSamplesPerSec = samplerate;
-    wformat->Format.wBitsPerSample = af_fmt2bits(format);
+    wformat->Format.wBitsPerSample = af_fmt_to_bytes(format) * 8;
     wformat->Format.cbSize = sizeof(WAVEFORMATEXTENSIBLE) - sizeof(WAVEFORMATEX);
 
     wformat->SubFormat                   = *format_to_subtype(format);
@@ -306,9 +306,11 @@ static int format_from_waveformat(WAVEFORMATEX *wf)
     // Since mpv doesn't have the notion of "valid bits", we just specify a
     // format with the container size. The least significant, "invalid"
     // bits will be excess precision ignored by wasapi.
-    // The change_bits operations should be a no-op for properly
+    // The change_bytes operations should be a no-op for properly
     // configured "special" formats, otherwise it will return 0.
-    return af_fmt_change_bits(format, wf->wBitsPerSample);
+    if (wf->wBitsPerSample % 8)
+        return 0;
+    return af_fmt_change_bytes(format, wf->wBitsPerSample / 8);
 }
 
 static bool chmap_from_waveformat(struct mp_chmap *channels, const WAVEFORMATEX *wf)
@@ -365,7 +367,7 @@ static bool set_ao_format(struct ao *ao, WAVEFORMATEX *wf, AUDCLNT_SHAREMODE sha
     }
 
     // Do not touch the ao for passthrough, just assume that we set WAVEFORMATEX correctly.
-    if (!AF_FORMAT_IS_SPECIAL(format)) {
+    if (af_fmt_is_pcm(format)) {
         struct mp_chmap channels;
         if (!chmap_from_waveformat(&channels, wf)) {
             MP_ERR(ao, "Unable to construct channel map from WAVEFORMAT %s\n",
@@ -575,7 +577,7 @@ static bool find_formats(struct ao *ao)
         // might be passthrough). If that fails, do a pcm format
         // search.
         return find_formats_exclusive(ao, true);
-    } else if (AF_FORMAT_IS_SPECIAL(ao->format)) {
+    } else if (af_fmt_is_spdif(ao->format)) {
         // If a passthrough format is requested, but exclusive mode
         // was not explicitly set, try only the requested passthrough
         // format in exclusive mode. Fall back on shared mode if that

audio/out/internal.h

@@ -59,7 +59,7 @@ struct ao {
     char *redirect;
 
     // Internal events (use ao_request_reload(), ao_hotplug_event())
-    atomic_bool request_reload, request_hotplug;
+    atomic_int events_;
 
     int buffer;
     double def_buffer;

audio/out/pull.c

@@ -155,7 +155,7 @@ end:
 
     // 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);
+        af_fill_silence((char *)data[n] + bytes, full_bytes - bytes, ao->format);
 
     return bytes / ao->sstride;
 }
@@ -206,15 +206,6 @@ static void resume(struct ao *ao)
     ao->driver->resume(ao);
 }
 
-static void drain(struct ao *ao)
-{
-    struct ao_pull_state *p = ao->api_priv;
-    int state = atomic_load(&p->state);
-    if (IS_PLAYING(state))
-        mp_sleep_us(get_delay(ao) * 1000000);
-    reset(ao);
-}
-
 static bool get_eof(struct ao *ao)
 {
     struct ao_pull_state *p = ao->api_priv;
@@ -223,6 +214,24 @@ static bool get_eof(struct ao *ao)
     return mp_ring_buffered(p->buffers[0]) == 0;
 }
 
+static void drain(struct ao *ao)
+{
+    struct ao_pull_state *p = ao->api_priv;
+    int state = atomic_load(&p->state);
+    if (IS_PLAYING(state)) {
+        // Wait for lower bound.
+        mp_sleep_us(mp_ring_buffered(p->buffers[0]) / (double)ao->bps * 1e6);
+        // And then poll for actual end. (Unfortunately, this code considers
+        // audio APIs which do not want you to use mutexes in the audio
+        // callback, and an extra semaphore would require slightly more effort.)
+        // Limit to arbitrary ~250ms max. waiting for robustness.
+        int64_t max = mp_time_us() + 250000;
+        while (mp_time_us() < max && !get_eof(ao))
+            mp_sleep_us(1);
+    }
+    reset(ao);
+}
+
 static void uninit(struct ao *ao)
 {
     ao->driver->uninit(ao);

audio/out/push.c

@@ -340,7 +340,8 @@ static void *playthread(void *arg)
                 pthread_cond_signal(&p->wakeup); // for draining
 
                 if (p->still_playing && timeout > 0) {
-                    mpthread_cond_timedwait_rel(&p->wakeup, &p->lock, timeout);
+                    struct timespec ts = mp_rel_time_to_timespec(timeout);
+                    pthread_cond_timedwait(&p->wakeup, &p->lock, &ts);
                 } else {
                     pthread_cond_wait(&p->wakeup, &p->lock);
                 }
@@ -352,8 +353,10 @@ static void *playthread(void *arg)
                     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);
+                    if (!p->need_wakeup) {
+                        struct timespec ts = mp_rel_time_to_timespec(timeout);
+                        pthread_cond_timedwait(&p->wakeup, &p->lock, &ts);
+                    }
                 }
             }
             MP_STATS(ao, "end audio wait");
@@ -435,7 +438,7 @@ const struct ao_driver ao_api_push = {
 int ao_play_silence(struct ao *ao, int samples)
 {
     assert(ao->api == &ao_api_push);
-    if (samples <= 0 || AF_FORMAT_IS_SPECIAL(ao->format) || !ao->driver->play)
+    if (samples <= 0 || !af_fmt_is_pcm(ao->format) || !ao->driver->play)
         return 0;
     char *p = talloc_size(NULL, samples * ao->sstride);
     af_fill_silence(p, samples * ao->sstride, ao->format);

bootstrap.py

@@ -5,13 +5,13 @@
 from __future__ import print_function
 import os, sys, stat, hashlib, subprocess
 
-WAFRELEASE = "waf-1.8.4"
-WAFURLS    = ["http://ftp.waf.io/pub/release/" + WAFRELEASE,
+WAFRELEASE = "waf-1.8.12"
+WAFURLS    = ["https://waf.io/" + WAFRELEASE,
               "http://www.freehackers.org/~tnagy/release/" + WAFRELEASE]
-SHA256HASH = "f02035fa5d8814f33f19b2b20d43822ddef6bb39b955ca196c2a247a1f9ffaa8"
+SHA256HASH = "01bf2beab2106d1558800c8709bc2c8e496d3da4a2ca343fe091f22fca60c98b"
 
 if os.path.exists("waf"):
-    wafver = subprocess.check_output(['./waf', '--version']).decode()
+    wafver = subprocess.check_output([sys.executable, './waf', '--version']).decode()
     if WAFRELEASE.split('-')[1] == wafver.split(' ')[1]:
         print("Found 'waf', skipping download.")
         sys.exit(0)

common/av_common.c

@@ -59,15 +59,12 @@ void mp_copy_lav_codec_headers(AVCodecContext *avctx, AVCodecContext *st)
     avctx->width                    = st->width;
     avctx->height                   = st->height;
     avctx->pix_fmt                  = st->pix_fmt;
-    avctx->sample_aspect_ratio      = st->sample_aspect_ratio;
     avctx->chroma_sample_location   = st->chroma_sample_location;
     avctx->sample_rate              = st->sample_rate;
     avctx->channels                 = st->channels;
     avctx->block_align              = st->block_align;
     avctx->channel_layout           = st->channel_layout;
     avctx->bits_per_coded_sample    = st->bits_per_coded_sample;
-    // Required in FFmpeg 2.5.x / Libav 11, deprecated afterwards.
-    avctx->stream_codec_tag         = st->stream_codec_tag;
 }
 
 // We merely pass-through our PTS/DTS as an int64_t; libavcodec won't use it.

common/av_log.c

@@ -134,8 +134,9 @@ static void mp_msg_av_log_callback(void *ptr, int level, const char *fmt,
     struct mp_log *log = get_av_log(ptr);
 
     if (mp_msg_test(log, mp_level)) {
-        if (log_print_prefix)
-            mp_msg(log, mp_level, "%s: ", avc ? avc->item_name(ptr) : "?");
+        const char *prefix = avc ? avc->item_name(ptr) : NULL;
+        if (log_print_prefix && prefix)
+            mp_msg(log, mp_level, "%s: ", prefix);
         log_print_prefix = fmt[strlen(fmt) - 1] == '\n';
 
         mp_msg_va(log, mp_level, fmt, vl);
@@ -173,6 +174,7 @@ void uninit_libav(struct mpv_global *global)
 {
     pthread_mutex_lock(&log_lock);
     if (log_mpv_instance == global) {
+        av_log_set_callback(av_log_default_callback);
         log_mpv_instance = NULL;
         talloc_free(log_root);
     }
@@ -207,36 +209,17 @@ void print_libav_versions(struct mp_log *log, int v)
 
     mp_msg(log, v, "%s library versions:\n", LIB_PREFIX);
 
-    bool mismatch = false;
-    bool broken = false;
     for (int n = 0; n < MP_ARRAY_SIZE(libs); n++) {
         const struct lib *l = &libs[n];
         mp_msg(log, v, "   %-15s %d.%d.%d", l->name, V(l->buildv));
-        if (l->buildv != l->runv) {
+        if (l->buildv != l->runv)
             mp_msg(log, v, " (runtime %d.%d.%d)", V(l->runv));
-            mismatch = true;
-            broken |= ((l->buildv & 255) >= 100) != ((l->runv & 255) >= 100);
-        }
         mp_msg(log, v, "\n");
     }
-    // This just won't work. It's 100% broken.
-    if (broken) {
-        mp_fatal(log, "mpv was compiled and linked against a mixture of Libav "
-                 "and FFmpeg versions. This won't work and will most likely "
-                 "crash at some point. Aborting.\n");
-        abort();
-    }
-    // We don't "really" support mismatched libraries, but if you like to
-    // suffer, you're free to enjoy the terrible aspects of dynamic linking.
-    // In particular, we don't use all these crazy accessors ffmpeg wants us
-    // to use in order to be ABI compatible after Libav merges - because that
-    // would make our code incompatible to Libav. It's madness.
-    if (mismatch) {
-        mp_warn(log, "Warning: mpv was compiled against a different version of "
-                "%s than the shared\nlibrary it is linked against. This can "
-                "expose subtle ABI compatibility issues\nand can lead to "
-                "misbehavior and crashes.\n", LIB_PREFIX);
-    }
+
+#if HAVE_AV_VERSION_INFO
+    mp_msg(log, v, "%s version: %s\n", LIB_PREFIX, av_version_info());
+#endif
 }
 
 #undef V

common/codecs.c

@@ -130,6 +130,28 @@ struct mp_decoder_list *mp_select_decoders(struct mp_decoder_list *all,
     return list;
 }
 
+// selection is a ","-separated list of decoders, all in the given family.
+struct mp_decoder_list *mp_select_decoder_list(struct mp_decoder_list *all,
+                                               const char *codec,
+                                               const char *family,
+                                               const char *selection)
+{
+    struct mp_decoder_list *list = talloc_zero(NULL, struct mp_decoder_list);
+    bstr sel = bstr0(selection);
+    while (sel.len) {
+        bstr decoder;
+        bstr_split_tok(sel, ",", &decoder, &sel);
+        add_new(list, find_decoder(all, bstr0(family), decoder), codec);
+    }
+    return list;
+}
+
+void mp_append_decoders(struct mp_decoder_list *list, struct mp_decoder_list *a)
+{
+    for (int n = 0; n < a->num_entries; n++)
+        add_new(list, &a->entries[n], NULL);
+}
+
 void mp_print_decoders(struct mp_log *log, int msgl, const char *header,
                        struct mp_decoder_list *list)
 {

common/codecs.h

@@ -37,6 +37,13 @@ struct mp_decoder_list *mp_select_decoders(struct mp_decoder_list *all,
                                            const char *codec,
                                            const char *selection);
 
+struct mp_decoder_list *mp_select_decoder_list(struct mp_decoder_list *all,
+                                               const char *codec,
+                                               const char *family,
+                                               const char *selection);
+
+void mp_append_decoders(struct mp_decoder_list *list, struct mp_decoder_list *a);
+
 struct mp_log;
 void mp_print_decoders(struct mp_log *log, int msgl, const char *header,
                        struct mp_decoder_list *list);

common/common.c

@@ -95,6 +95,12 @@ void mp_rect_union(struct mp_rect *rc, const struct mp_rect *rc2)
     rc->y1 = FFMAX(rc->y1, rc2->y1);
 }
 
+// Returns whether or not a point is contained by rc
+bool mp_rect_contains(struct mp_rect *rc, int x, int y)
+{
+    return rc->x0 <= x && x < rc->x1 && rc->y0 <= y && y < rc->y1;
+}
+
 // Set rc to the intersection of rc and src.
 // Return false if the result is empty.
 bool mp_rect_intersection(struct mp_rect *rc, const struct mp_rect *rc2)

common/common.h

@@ -75,6 +75,7 @@ struct mp_rect {
 
 void mp_rect_union(struct mp_rect *rc, const struct mp_rect *src);
 bool mp_rect_intersection(struct mp_rect *rc, const struct mp_rect *rc2);
+bool mp_rect_contains(struct mp_rect *rc, int x, int y);
 
 int mp_snprintf_cat(char *str, size_t size, const char *format, ...)
     PRINTF_ATTRIBUTE(3, 4);