Ian Baker

Thanks for pointing out this can likely be fixed by using `TrueHdSampleRechunker` in `FragmentedMp4Extractor`. I'm afraid we are unlikely to work on this soon, but we would consider a high...

I played around a bit with this, and added a new test file with **three** SeekHead elements (https://github.com/androidx/media/pull/2268/commits/f5a6ed0f3f5e3ad827e886de088bd93ecbc2178e). When I run the test with this new file **without** the `MatroskaExtractor`...

Thanks for the follow-up fixes. I've re-run `MatroskaExtractorTest.mkvSample_threeSeekHeads()` with `DumpFileAsserts.DUMP_FILE_ACTION = WRITE_TO_LOCAL` to update the dump files (and uploaded this commit). Re-running the test with `DumpFileAsserts.DUMP_FILE_ACTION = COMPARE_WITH_EXISTING` results in...

The [`Extractor` javadoc](https://developer.android.com/reference/androidx/media3/extractor/Extractor#read(androidx.media3.extractor.ExtractorInput,androidx.media3.extractor.PositionHolder)) describes what happens when an IO error is encountered: > When this method throws an `IOException`, extraction may continue by providing an `ExtractorInput` with an unchanged read...

It looks like the 'restart from the beginning' behaviour is happening when `retryFromStartIfLive = true` [1], and this stream is determined to be 'live' (`!isOnDemand`) before the `Cues` element is...

With some logging added locally [1], in the `retryFromStartIfLive` case, I see `MatroskaExtractor.seek(long position, long timeUs)` called with `position=0` before the `read(...)`, to indicate that the subsequent `read(...)` will **not**...

`mediainfo` shows that the provided video file is encoded with the H.264 `[email protected]` profile and level. Android devices are only required (by the CDD) to support the Baseline and Main...

On the devices where you observe the issue, what do you see in logcat [with an `EventLogger` attached](https://developer.android.com/media/media3/exoplayer/debug-logging)? Specifically, what value do you see for `supported=` on the video track...

In the Pixel 4 XL log it looks like decoding failed completely - do you have some logic to automatically retry on a player error like that? Please can you...

I'm afraid the BR only **just** captures the `MediaCodec$CodecException` stacktrace, with no logging included from before it - so it's not that useful unfortunately. It looks like the exception was...