mpv-player / mpv-player/mpv

ao_avfoundation: tail of every file is dropped at EOF (buffered audio never drained)

Open
#18,461 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
C
Stars
37k
Forks
3.5k
Avg merge
1d 10h
Merged PRs (30d)
22

Description

mpv Information
mpv v0.41.0-1041-gc611c1c6f Copyright © 2000-2026 mpv/MPlayer/mplayer2 projects
libplacebo version: v7.360.1
FFmpeg version: 9.0.1
FFmpeg library versions:
   libavcodec      63.1.101
   libavdevice     63.1.101
   libavfilter     12.1.101
   libavformat     63.1.101
   libavutil       61.1.101
   libswresample   7.1.101
   libswscale      10.1.101
Other Information
- macOS version: macOS 27.0 (26A5425a), Apple Silicon (Mac16,10 / arm64)
- Source of mpv: Homebrew --HEAD build. Also reproduced with the 0.41.0 release bottle.
- Latest known working version: unknown, reproduces on both 0.41.0 release and current git
- Issue started after the following happened: noticed on macOS 27, where #18384 makes
  ao_coreaudio fail and mpv silently falls back to ao_avfoundation
Reproduction Steps
  1. Generate a 30 second test tone:
ffmpeg -f lavfi -i "sine=frequency=440:duration=30" -ac 2 -ar 44100 -c:a pcm_s16le stereo30.wav
  1. mpv --no-config --ao=avfoundation stereo30.wav
  2. Watch the position counter. Playback ends and mpv exits at 00:00:26 / 00:00:30.
  3. Control: mpv --no-config --ao=coreaudio stereo30.wav reaches 00:00:29 / 00:00:30.

The shortfall is a constant amount of time, not a proportion of the file:

file ao final position
16.50 s mono avfoundation 00:00:12
30 s stereo avfoundation 00:00:26
60 s mono avfoundation 00:00:56
30 s stereo coreaudio (control) 00:00:29

4.0 seconds every time, which is exactly what ao_avfoundation reports having buffered:

[ao/avfoundation] device buffer: 88200 samples.
[ao/avfoundation] using soft-buffer of 88200 samples.

88200 + 88200 = 176400 samples at 44100 Hz = 4.000 s.

Measuring total process wall time shows audio really is lost, not just mis-reported:

file duration ao wall time
30.00 s avfoundation 28.61 s
30.00 s coreaudio 30.30 s
16.50 s avfoundation 15.11 s
16.50 s coreaudio 16.74 s

So roughly 1.6 to 1.7 seconds of audio is never played, and the reported position stops
about 4 seconds early. In the log, mpv reaches EOF, logs audio draining and
draining left over audio, waits about 2 seconds, and exits, without waiting for the
AVSampleBufferAudioRenderer to finish playing everything already queued.

Expected Behavior

Playback plays to the end of the file, as ao_coreaudio does.

Actual Behavior

The tail of every file is silently discarded, with no error or warning. On short files this
is severe: a 16 second clip loses about 10 percent of its running time off the end.

Relationship to #18384

These look like the same problem from a user's point of view but are separable, so I am
filing this rather than adding to that thread.

On macOS 27, #18384 makes ao_coreaudio fail with
unable to set the input channel layout on the audio unit ([206][255][255][255]/-50),
mpv falls back to ao_avfoundation, and the user sees truncation. That matches the
"does not play to completion" half of #18384.

But the truncation reproduces with an explicit --ao=avfoundation on files where
ao_coreaudio initialises and plays correctly, including plain stereo files that never
trigger the -50 error at all. It also reproduces on the 0.41.0 release. So it is
independent of #18384 and will still be there once the channel layout bug is fixed.

Log File

Both logs (produced with --log-file, full and untruncated) are here, since the CLI
cannot attach files to an issue:

https://gist.github.com/compupane/aa62a7c7afef8e20597a9c5b71a47c68

  • mpv-avfoundation-truncation.txt is the reproduction
  • mpv-coreaudio-control.txt is the same file with --ao=coreaudio as a control
Sample Files

No sample needed, the ffmpeg one-liner above generates it. Any audio file reproduces it.

I carefully read all instruction and confirm that I did the following:
  • I tested and confirmed that the issue exists with the latest release version or newer.
  • I provided all required information including system and mpv version.
  • I produced the log file with the exact same set of files, parameters, and conditions used in "Reproduction Steps", with the addition of --log-file=output.txt.
  • I produced the log file while the behaviors described in "Actual Behavior" were actively observed.
  • I attached the full, untruncated log file.
  • I attached the backtrace in the case of a crash. (not applicable, no crash)

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start at the ao_avfoundation EOF and audio-draining path, using the provided ffmpeg command and mpv --no-config --ao=avfoundation reproduction. Compare the final drain behavior with ao_coreaudio and the linked full log. Done means buffered audio is played before exit and playback reaches the end of generated files without truncation.

Written by the indexing model from the issue text.

Assessment

Tech stack
c
Domain
audio-video-rtc, backend
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
68/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.