"ParserException: searched too many bytes" with specific MP3 files

[thibseisel]

Issue description

When playing MP3 files located on the device's storage with SimpleExoPlayer, some will trigger a source error when played to completion:

2018-10-14 11:52:06.440 E/ExoPlayerImplInternal: Source error.
    com.google.android.exoplayer2.ParserException: Searched too many bytes.
        at com.google.android.exoplayer2.extractor.mp3.Mp3Extractor.synchronize(Mp3Extractor.java:282)
        at com.google.android.exoplayer2.extractor.mp3.Mp3Extractor.read(Mp3Extractor.java:169)
        at com.google.android.exoplayer2.source.ExtractorMediaPeriod$ExtractingLoadable.load(ExtractorMediaPeriod.java:898)
        at com.google.android.exoplayer2.upstream.Loader$LoadTask.run(Loader.java:379)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1162)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:636)
        at java.lang.Thread.run(Thread.java:764)

At the time of writing, I reproduced the bug only with specific files (all tracks from the same album).
Are my files corrupted, or is this a real ExoPlayer issue ? It doesn't happen with ExoPlayer 2.8.4.

Reproduction steps

1. Create an instance of SimpleExoPlayer with the following code

SimpleExoPlayer exoPlayer = ExoPlayerFactory.newSimpleInstance(context, new DefaultRenderersFactory(context), new DefaultTrackSelector());
AudioAttributes musicAttributes = new AudioAttributes.Builder()
            .setContentType(C.CONTENT_TYPE_MUSIC)
            .setUsage(C.USAGE_MEDIA)
            .build();
exoPlayer.setAudioAttributes(musicAttributes, true);

2. Prepare the player with the media to play in a ConcatenatingMediaSource

ExtractorMediaSource uriSource = ExtractorMediaSource.Factory(new DefaultDataSourceFactory(...))
                    .setExtractorsFactory(new DefaultExtractorsFactory())
                    .setTag(mediaDescriptionCompat)
                    .createMediaSource(theMp3FileUri);
ConcatenatingMediaSource concatSource = new ConcatenatingMediaSource(new MediaSource[] { uriSource });
exoPlayer.prepare(concatSource);
exoPlayer.setPlayWhenReady(true);

3. Wait until the song is played to completion. You can also seek near the end of the song and wait for it to complete.

4. The player moves to the ERROR state and a "source error" is printed to the logcat.

Link to test content

This happens for every track from the album "Breakin' Outta Hell" of Airbourne, but might happen with other media.
Since this media is protected by law I can't share it here, but maybe I could share only the few last seconds of the song in order to reproduce the bug. Tell me if you need it.

Version of ExoPlayer being used

2.9.0 (doesn't happen with 2.8.4)

Device(s) and version(s) of Android being used

Google Nexus 5X - Android API 27
The issue happens every time with the specified files.

A full bug report captured from the device

Tell me if that's needed.

[andrewlewis]

Given that this seems to be a regression since 2.8.4 it might be a bug, but to investigate we need to be able to reproduce the issue. If you can find some minimal input that reproduces the issue and can be shared please send it to [email protected].

[ojw28]

@andrewlewis - content is received to look at this now (in case you didn't notice :)).

[andrewlewis]

It looks like the file has about 150 KB of non-MP3 data at the end. On 2.8.4 we handled the ParserException resulting from reaching it as non-fatal, then retried synchronization and hit the end of the stream, which was not treated as an error. On 2.9.0 the behavior changed to treat ParserException as a fatal exception, hence the error.

I'm not sure how common it is to have a lot of non-MP3 data after the audio stream like this, and I haven't seen it before. If it is common we should probably add some special handling so it's not treated as an error. If your app needs to handle these particular streams you could handle ParserException and suppress the error.

[ojw28]

I think we can tell from the XingSeeker where the end of the MP3 data is, in which case it's a bug that we're trying to read the 150 KB on the end. Ditto for any other Seekers that give us this information. I fixed a similar issue for WavExtractor not too long ago.