SDK / exoplayer

Also queue textures from a different thread in TextureAssetLoader, to
have a behaviour closer to reality.

PiperOrigin-RevId: 538473089
(cherry picked from commit 01dbda1f768d0511c85fdcdb4a6972e85a477426)

PiperOrigin-RevId: 538469993
(cherry picked from commit bc94a0e189b92310cc28c644aa11771d534987d5)

#minor-release

PiperOrigin-RevId: 538466615
(cherry picked from commit ac18e2ea5f05cb23af1f9d12c102feffb8e9c1ff)

Confirms that multiple textures can be output, and that timestamps and pixels
are as expected.

PiperOrigin-RevId: 538459296
(cherry picked from commit cfcc53b3829f5dedbd4a2e96af87f3b05dec7715)

PiperOrigin-RevId: 538437142
(cherry picked from commit 136f323f97bc048d9cc9a6b703b72073aa6132de)

PiperOrigin-RevId: 538209925
(cherry picked from commit 824960bae924141a5036744226c5fe79b37d3c12)

PiperOrigin-RevId: 538175466
(cherry picked from commit 6dc8e060c958a2f36dd8d7d1b2f67585082a2a22)

PiperOrigin-RevId: 538173603
(cherry picked from commit 2c7d929db17806176a9b592238292ca2fe436c92)

The sample timestamp carried by the emsg box can have a significant delta when comparing to the earliest presentation timestamp of the segment. Using this timestamp to intialize the timestamp offset in TimestampAdjuster will cause the media sample to have a wrong adjusted timestamp. So we should defer adjusting the metadata sample timestamp until the TimestampAdjuster is initialized with a real media sample.

PiperOrigin-RevId: 538172841
(cherry picked from commit 08c189e768d77e9e2a9665b23fa384e776f9678b)

PiperOrigin-RevId: 538129792
(cherry picked from commit 13df52b97e64cf54071e130faa2c46001d6b7029)

When initiated by MediaController, it should be possible for `MediaSession` to pass `MediaItems` to the `Player` if they have `LocalConfiguration`. In such case, it is not required to override `MediaSession.Callback.onAddMediaItems`, because the new current default implementation will handle it.

However, in other cases, MediaItem.toBundle() will continue to strip the LocalConfiguration information.

Issue: androidx/media#282

#minor-release

PiperOrigin-RevId: 537993460
(cherry picked from commit bcddaf27654ed342ce70fc7a270d478953c2fb80)

PiperOrigin-RevId: 537908595
(cherry picked from commit f8be8c087de550d2d7a2d1c1b7dc9591804c93f8)

`outputSurfaceInfoChanged` is not reset when `defaultShaderProgram` is null.
That is, on the first time `ensureConfigured()` is called with output size
changed, `outputSurfaceInfoChanged` is not set to false after creating the
`defaultShaderProgram`, and `defaultShaderProgram` will be created again on the
second time `ensureConfigured()` is called.

PiperOrigin-RevId: 537870404
(cherry picked from commit 3d8a2f8b8a81f022d71a12abf3deaabd426ca8ca)

https://github.com/google/ExoPlayer/commit/fedd99b4911e61f18e68e57021ac8b55db7e5f44 introduced using `DefaultDecoderFactory.getDecoderInfo(format) != null` caused certain tests not to be skipped when they were expected to be, creating more mh failures.

PiperOrigin-RevId: 537820370
(cherry picked from commit c78151b52d9ee53a08681b2a7ab3be352bd43572)

PiperOrigin-RevId: 537814319
(cherry picked from commit 68939399b4bb6f6b72376d3f8cf0441c1285e230)

ExoPlayer extractors (backing `MetadataRetriever`) now parse the color format
from the bitstream so using `MetadataRetriever` should be an equivalent but
more lightweight way to verify the color info.

Also remove try/catch blocks in test code calling into these methods, and add
skipping based on decoder capabilities in the cases where it was missing.

PiperOrigin-RevId: 537789483
(cherry picked from commit 88db01116afa54e52f8516533256d447de45e8a4)

Previously, we would apply a general effect to signal wanting to transcode.

PiperOrigin-RevId: 537034455
(cherry picked from commit fb032f0c1feb42878f4ebdfa7f673e2cdd13949f)

PiperOrigin-RevId: 537014587
(cherry picked from commit 8166b3590fe7d629f55c1e8af27091f8309d4cb7)

PiperOrigin-RevId: 536770380
(cherry picked from commit 11c52d2de1e07b5c497438ee86651e0bd399b3d2)

PiperOrigin-RevId: 536727079
(cherry picked from commit a8baf21cae4c47de0bdcb8e3f9474d4914662d36)

#minor-release

PiperOrigin-RevId: 536724725
(cherry picked from commit 792aa7295827471902f6641cf83f4c4c12effedc)

In some cases the codec selected for decoding has a different MIME type than
the media. In thoses cases Transformer continued to use the media's MIME type
and that caused codec configuration failures.

Removed `EncoderUtil.findCodecForFormat()` as we stopped using the method it
uses for finding a codec. Plus, the method is only used in the test.

See also `MediaCodecUtil.getALternativeCodecMimeType()`.

PiperOrigin-RevId: 536683663
(cherry picked from commit fedd99b4911e61f18e68e57021ac8b55db7e5f44)

Otherwise, texture output errors out if video decoding decodes faster than audio,
hitting the end of the file, while audio is still in the middle of the file.

PiperOrigin-RevId: 536679568
(cherry picked from commit 2ff8b446e648883144687c4c020b2d23a0b4e1d3)

This future.get() duplicates the wait done in
singleThreadExecutorService.awaitTermination(). If awaitTermination times out, this future.get() would also result in unnecessary blocking.

PiperOrigin-RevId: 536442153
(cherry picked from commit d7f4fea0b9d1be3f5d0211dc50b0f1fbaec05564)

PiperOrigin-RevId: 536351494
(cherry picked from commit ba8b9d13d0f1e6cce16a4ff5ea9271f7dd5a9e36)

Deprecated field `MediaItem.playbackProperties` remains for backwards compatibility, but its type is changed from `MediaItem.PlaybackProperties` to `MediaItem.LocalConfiguration`. The private `MediaItem` constructor will now also take in a `LocalConfiguration` argument instead.

PiperOrigin-RevId: 535648420
(cherry picked from commit b89cc652e5d466591617bfb5c9183a033cdf74b9)

With this change we will now support loading bitmaps from all the formats documented [here](https://developer.android.com/guide/topics/media/media-formats#image-formats) except for gifs (because they are animated). Java doc is added to express this.

PiperOrigin-RevId: 535610152
(cherry picked from commit 7740ccc9382c55e246f5b7e5ea4e120d99b662bc)

PiperOrigin-RevId: 535554628
(cherry picked from commit 8ba6b12ebc28a5fefe5f0b1d9a565794ebaf9ea6)

*** Original commit ***

ExoPlayer: Add setVideoFrameProcessorFactory().

This allows apps to use a custom VideoFrameProcessor implementation for video
playback. This may be useful, for example, when outputting to a texture.
***

PiperOrigin-RevId: 536391597
(cherry picked from commit dbbb0919075a9c95f45d51307b434eda4c4744bd)

#minor-release

PiperOrigin-RevId: 535268825
(cherry picked from commit e7d1c57e80b7416a8093c098ccc9b927d361fe16)

PiperOrigin-RevId: 535255453
(cherry picked from commit 6a9167097e78341b3bd3ad8309ee79ae35ecfd13)

PiperOrigin-RevId: 535233266
(cherry picked from commit 02698a8a4c169546c17c20c8336a05470b957da7)

PiperOrigin-RevId: 535183521
(cherry picked from commit c9ca510a8382abc601faa3d91e95fde845ea952a)

This change moves the default logic into the actual Player
implementations, but does not introduce any behavior changes compared
to addMediaItems+removeMediaItems except to make the updates "atomic"
in ExoPlayerImpl, SimpleBasePlayer and MediaController. It also
provides backwards compatbility for cases where Players don't support
the operation.

Issue: google/ExoPlayer#8046

#minor-release

PiperOrigin-RevId: 534945089
(cherry picked from commit 6309b11792b05306d004747af35d67f32a352782)

PiperOrigin-RevId: 534927167
(cherry picked from commit d8e30c351e9e790730942c8d336b1e4689f571c2)

Issue: google/ExoPlayer#11160

#minor-release

PiperOrigin-RevId: 534896789
(cherry picked from commit b72ef3e9b7ce14eaf98de5b2cc32be6d25a61ed8)

Also, document that texture output disables manual frame release.

In the past, texture output would lead to surface output methods throwing. Now,
they're simply no-ops instead.

PiperOrigin-RevId: 534894168
(cherry picked from commit 21ee2e2243c2d0f2875ae7506e90217fe6044bc7)

PiperOrigin-RevId: 534869452
(cherry picked from commit a785d7d01643cc6c07a741a33e588c8c43feb31c)

When the video renderer is disabled, the video size is set to 0/0
and sent to listeners. The `PlayerView` potentially still has the last frame
displayed when the player is stopped or an error occurs. This may have the
effect that the frame is displayed distorted.

Not changing the aspect ratio when the video size arrives when the player is IDLE
avoids the problem. In the case when playback starts again and the renderes is
enabled, another video size is sent to the listener.

#minor-release

PiperOrigin-RevId: 534860889
(cherry picked from commit 3e3aaf8816aa4a425343ac4aaf2299cd6e9f8085)