Before this CL, the video sink was stuck if a flush was executed:
- after VideoSink.onInputStreamChanged, which is setting
pendingInputStreamBufferPresentationTimeUs and
- before the previous stream was fully rendered.
This is because pendingInputStreamBufferPresentationTimeUs was not reset
to TIME_UNSET when flushing the sink, so that the sink was still waiting
for the last frame of the previous stream to be rendered in
handleInputFrame/Bitmap.
PiperOrigin-RevId: 667924517
Per the javadoc, the method `MediaPeriod.maybeThrowPrepareError` is only allowed to be called before the period has completed preparation. For later errors in loading the streams, `SampleStream.maybeThrowError` will be called instead.
PiperOrigin-RevId: 665831430
It is possible for playback to be stuck when there is failure in loading further data, while the player is required to load more due to the buffered duration being under `DefaultLoadControl.bufferForPlayback`. Therefore, we check if there is any loading error in `isLoadingPossible`, so that the player will allow the playback of the existing data rather than waiting forever for the data that can never be loaded.
Issue: androidx/media#1571
#cherrypick
PiperOrigin-RevId: 665801674
This uses a 'bundled' extractor (`SubtitleExtractor`) because
`MediaParser` doesn't support transcoding subtitles to
`application/x-media3-cues`. This transcoding is required to support
non-legacy subtitle handling where they are parsed during extraction,
instead of during rendering.
PiperOrigin-RevId: 665282298
This is only used as part of legacy subtitle decoding. If any apps are
using it directly (rather than via `DashMediaSource` or `SsMediaSource`)
they probably need to enable legacy decoding in `TextRenderer` to avoid
playback failures.
PiperOrigin-RevId: 663740343
The method currently has two ways of operating with a complicated
contract:
1. outputBuffer == null -> sets a new outputBuffer and starts
draining it as far as possible.
2. outputBuffer != null -> continues draining a previous buffer,
requires that the input parameter is the same as outputBuffer.
This contract can be made cleaner by splitting the logic into
(1) setting a new output buffer and (2) draining the output buffer.
Only one parameter is needed for each method and there is no
requirement to pass in the same argument as before because we already
know what we are draining. This also means the caller of the method
doesn't have to keep track of the actual outputBuffer, allowing
further processing steps in the future that change the drained
instance.
PiperOrigin-RevId: 663316251
Previous to this change, `FrameworkMediaDrm.requiresSecureDecoder`
ignores its `sessionId` parameter on API 31+, and uses only the
`mimeType` parameter. This means the result [assumes the session is
opened at the 'default security
level'](https://developer.android.com/reference/android/media/MediaDrm#requiresSecureDecoder(java.lang.String)):
> The default security level is defined as the highest security level
> supported on the device.
This change is a no-op in all (?) cases, because the `ExoMediaDrm`
interface only exposes the zero-arg `openSession()` method, which in the
framework case **also** assumes the highest security level is preferred:
> By default, sessions are opened at the native security level of the
> device.
However, it seems more obviously correct to only make this
"highest/native security level" assumption in one place
(`openSession()`), and check the session's **actual** security level
everywhere else.
Issue: androidx/media#1603
PiperOrigin-RevId: 662872860
This method was added in API 31 (S) but it's non-functional
(incorrectly, silently, returns `false`) on the Widevine plugin version
(`16.0`) from R (API 30), which some devices up to at least API 34 are
still using.
This results in ExoPlayer incorrectly selecting an insecure decoder for
L1 secure content, and subsequently calling
`MediaCodec.queueInputBuffer` instead of `queueSecureInputBuffer`,
which is not supported and generates the following error:
> Operation not supported in this configuration: ERROR_DRM_CANNOT_HANDLE
Issue: androidx/media#1603
#cherrypick
PiperOrigin-RevId: 662852176
When there is an exception thrown from the `LoadTask`, the `Loader` will call `Loader.Callback.onLoadError`. Some implementations of `onLoadError` method may call `MediaPeriod.onContinueLoadingRequested`, and in the `PreloadMediaSource`, its `PreloadMediaPeriodCallback` will be triggered and then it can further call `continueLoading` if it finds needed. However the above process is currently done synchronously, which will cause problem. By calling `continueLoading`, the `Loader` is set with a `currentTask`, and when that long sync logic in `Loader.Callback.onLoadError` ends, the `Loader` will immediately retry, and then a non-null `currentTask` will cause the `IllegalStateException`.
Issue: androidx/media#1568
#cherrypick
PiperOrigin-RevId: 662550622
Check if the output device supports spatialization for the requested output format. If so, return a stream decoded for 6 channels in a 5.1 layout. Otherwise, return a stream decoded for 2 channels in a binaural layout.
PiperOrigin-RevId: 662546818
This is an additional signal that legacy subtitle support needs to be
explicitly enabled, and is going away at some point.
PiperOrigin-RevId: 661305694
The integration with external libraries like Glide or Coil is
currently extremely complicated. Providing the boilerplate code
to handle the ImageDecoder interface and a better hook for custom
image decoders in DefaultRenderersFactory allows apps to easily
inject their own logic for external image loading libraries.
PiperOrigin-RevId: 659508914
This API allows users to retrieve the sample size, ensuring they can allocate sufficient buffer capacity before utilizing the `readSampleData(ByteBuffer buffer, int offset)` method to read data.
PiperOrigin-RevId: 658772408
Refactor to replace instances of `RuntimeException` with `IllegalStateException`.
This change ensures exceptions can be handled more specifically without
unintended catches of unrelated exceptions.
PiperOrigin-RevId: 658337459
Added a new data source which acts an adapter to read media data from platform `MediaDataSource`. This enables adding the `setDataSource(MediaDataSource)` API to `MediaExtractorCompat`.
PiperOrigin-RevId: 657564901
Added three `setDataSource` APIs in `MediaExtractorCompat`:
- `setDataSource(Context context, Uri uri, @Nullable Map<String, String> headers)` to set data source with a content URI and optional headers.
- `setDataSource(String path)` to set data source using a file path or HTTP URL.
- `setDataSource(String path, @Nullable Map<String, String> headers)` to set data source using a file path or HTTP URL with optional headers.
PiperOrigin-RevId: 657563973
The builtin speaker is to be supported as a suitable output when that is deliberately selected for the media playback by the user in Wear OS.
PiperOrigin-RevId: 655950824
The DefaultPreloadManagerTest didn't to catch this because we use main looper as the preload looper in the tests. This CL also improves the tests by assigning the preload looper with one that corresponds to a different thread.
PiperOrigin-RevId: 655664189
This is following a renaming of registerInputFrame to handleInputFrame.
- queueInputBitmap is renamed to handleInputBitmap for consistency with
handleInputFrame.
- registerInputStream is renamed to onInputStreamChanged for consistency
with media3 method names.
PiperOrigin-RevId: 655529699
Aligns `MediaExtractorCompat` with platform behavior by using `getDeclaredLength()`
instead of `getLength()` when setting the data source. This ensures compatibility
with asset files where the length is not predefined, even though it may result
in reading across multiple logical files when backed by the same physical file.
PiperOrigin-RevId: 655153390
AudioTrack automatically ramps down volume when pausing. However,
when this happens as part of a AudioSink.flush() operation, we
need to postpone the actual flush() until the ramp down finished
in the audio system. Otherwise audio is just cut off, creating pop
sounds.
Delaying the release is fine now, because DefaultAudioSink starts
creating a new track immediately without waiting for the previous
track to be released.
Also using the opportunity to add more comments about related quirks
of the AudioTrack flush/release handling for more context.
PiperOrigin-RevId: 654794818
We currently wait until a previous AudioTrack from the same
DefaultAudioSink is released on a background thread before attempting
to initialize a new AudioTrack. This is done to avoid issues where
the releasing track blocks some shared audio memory, preventing a new
track from being created.
The current solution has two main shortcomings:
- In most cases, the system can easily handle multiple AudioTracks
and waiting for the release just causes unnecessary delays (e.g.
when seeking).
- It only waits for a previous track from the same DefaultAudioSink,
not accounting for any other tracks that may be in the process of
being released from other players.
To improve on both shortcomings, we can
(1) move the check for "is releasing tracks and thus may block shared
memory" to the static release infrastructure to be shared across all
player instances.
(2) optimistically create a new AudioTrack immediately without waiting
for the previous one to be fully released.
(3) extend the existing retry logic that already retries failed
attempts for 100ms to only start the timer when ongoing releases are
done. This ensures we really waited until we have all shared resources
we can get before giving up completely. This also acts as a replacement
for change (2) to handle situations where creating a second track is
genuinely not possible. Also increase threshold to 200ms as the new
unit test is falky on a real device with just 100ms (highlighting that
the device needed more than 100ms to clean up internal resources).
PiperOrigin-RevId: 654053123
Introduced three `setDataSource` APIs in `MediaExtractorCompat`, enabling the use of `AssetFileDescriptor` and `FileDescriptor` to set the data source.
PiperOrigin-RevId: 653957035
This controller connects the audio output to the MediaCodec so
that it can automatically propagate CTA-2075 loudness metadata.
PiperOrigin-RevId: 653628503
This replaces the existing PlaceholderSurface mode with a more
efficient solution that doesn't require a GL texture or a new
thread.
PiperOrigin-RevId: 653537596
This is a regression test for the bug introduced in bb9ff30c3a
which was manually spotted and fixed in 0d2bf49d6a.
Reverting the fix causes this test to fail.
This test is a bit hacky because we have to munge the stack trace of
the `IllegalStateException` to make it look like it was thrown from
inside `MediaCodec`. We deliberately do this 'badly' (e.g. using
`fakeMethod`) to avoid a future reader being confused by a
fake-but-plausible stack trace.
PiperOrigin-RevId: 652820878
Now the value is guaranteed to be zero (see bb9ff30c3a), we can
remove the rotation handling for it in the UI module. We can also
enforce the documentation more clearly by not even setting the
value to anything other than zero.
PiperOrigin-RevId: 652772091
Added a method to set the timeout for the SNTP request.
Also changed the default timeout to 5s instead of 10s as it seemed quite high.
Issue: androidx/media#1540
PiperOrigin-RevId: 652566008
The change in bb9ff30c3a removed the detection util completely,
but even on API21+, the codec exceptions can be thrown as
IllegalStateException and we should reinstate this check to
correctly classify the exceptions.
PiperOrigin-RevId: 652557091
This removes several workarounds that are no longer needed, including
`codecNeedsMonoChannelCountWorkaround` which has been permanently
disabled since the (incomplete) minSdk 19 clean-up in fb7438378d.
PiperOrigin-RevId: 652495578
This should have no influence on app behavior and other policies
and just allows code to depend on new API 35 platform symbols.
PiperOrigin-RevId: 652414026
We currently use displaySurface == placeholderSurface IF codec != null
as a signal that the codec is configured with a placeholder. In the
future, this placeholder might not be needed and we can decouple this
state a bit better by leaving displaySurface == null in this case and
only using placeholderSurface instead of null when setting a Surface
to the codec.
PiperOrigin-RevId: 652391729
All other AndroidX libraries have already increased their min SDK to
21.
This change renames private symbols to remove `V21` suffixes and
similar, but doesn't change public or protected symbols with similar
names, to avoid needless breakages/churn on users of the library.
Some of the dead code removal is more complex, so I've split it out
into follow-up changes to make it easier to review.
PiperOrigin-RevId: 651776556
MediaCodecRenderer might not be able to determine whether a sample
is last in time. Do not use decoder input skipping optimization
for buffers close to media end duration in order to ensure
last frame is rendered.
PiperOrigin-RevId: 651757814
The integration with `SubtitleTranscodingExtractorOutput` has been
moved inside the relevant `Extractor` implementations instead.
PiperOrigin-RevId: 651213564
VideoSink.registerInputFrame is now called for every input frame (not
only the ones that should be rendered to the input surface) because it's
the VideoSink that decides whether it wants the frame to be rendered.
PiperOrigin-RevId: 651049851
In order for DASH playback to benefit from
FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES, the fMP4 extractor flag
must be set. The smallest API change that allows this is to add an
experimental method to BundledChunkExtractor.
Add a dash end-to-end test to verify that video frames are skipped at
decoder input.
PiperOrigin-RevId: 651046676
The operation almost always acquires resources straight-away,
for example a new thread or file access. This means starting
many metadata retrievals in parallel easily causes resource
contention. This problem can be alleviated by limiting the
maximum number of parallel retrievals.
Local testing showed a 20% speedup for local file retrievals
when limited to 5 in parallel. Any more parallel retrievals
did not show any improvement or started getting slower again
the more operations are allowed.
PiperOrigin-RevId: 650674685
MetadataRetriever is often used in a loop to retrieve data about many
items in a directory for example. In such cases, it's wasteful to
start a new 'playback' thread for each retrieval, in particular since
this thread is doing almost nothing (just triggering loads and handling
events). This change re-uses an existing thread if one exists already.
PiperOrigin-RevId: 650633343
During a seek, or when playing a media with clipped start,
MCVR encounters preroll decode-only buffers that are not rendered.
Use C.BUFFER_FLAG_NO_OTHER_SAMPLE_DEPENDS_ON_THIS to determine
whether a decode-only buffer is unused as reference.
These buffers can be dropped before the decoder.
When this optimization is triggered, increment
decoderCounters.skippedInputBufferCount.
Tested in ExoPlayer demo app on "One hour frame counter (MP4)"
after enabling extractorsFactory.setMp4ExtractorFlags(
FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES);
Observe: "sib" increases on each seek.
PiperOrigin-RevId: 650566216
The idea was to not even write any samples to SampleQueues as
they are not needed during the metadata extraction process.
However, this is not easily possible as long as we use our
existing Extractors and MediaSource/Periods for loading given
how deeply ingrained the extraction of samples is in these
classes. For most common formats like MP4, no samples will be
extracted anyway as they can finish the prepare step without
reading any samples.
PiperOrigin-RevId: 650529371
Upon the call of `PreloadMediaSource.preload`, the source will periodically check the source refresh or period loading error, and trigger `PreloadMediaSource.PreloadControl.onPreloadError`. For now, the `DefaultPreloadManager` will skip the problematic source and continue to preload the next source. The checking of the error will be terminated when the source stops preloading or releases.
PiperOrigin-RevId: 650195817
The percentage should be interpreted as relative to the size of a parent
node.
This change makes this inheritance work correctly for percentages in
both the parent and child. It does not fix the case of a non-percentage
parent size with a percentage child size.
PiperOrigin-RevId: 649631055
Parse LHEVCDecoderConfigurationRecord with the ‘lhvC’ type and set the corresponding sample mime type to video/mv-hevc. With no MV-HEVC decoder available, fallback to single-layer HEVC decoding.
PiperOrigin-RevId: 649119173
The VideoSink has an input and an output surface. Clarify which surface
the Javadoc is referring to. Also document that the VideoSink can be fed
with images, and that multiple renderers can feed the same sink.
PiperOrigin-RevId: 649052551
MCVR crashed because MCVR registers a new input stream to VideoSink on every
`onOutputFormatChanged()`, assuming that `onOutputFormatChanged()` is only
invoked on media item transition. However, it can be called multiple times for
one media item.
PiperOrigin-RevId: 649050576
Lint somehow complains that the integer resulting from the bit-manipulation shouldn't be passed as an @IntDef parameter.
#cherrypick
PiperOrigin-RevId: 648687698
Upon track transition of offloaded playback of gapless tracks, the framework will reset the audiotrack frame position. The `AudioTrackPositionTracker`'s `AudioTimestampPoller` must be made to expect the reset and cache accumulated sum of `AudioTimestamp.framePosition`.
#cherrypick
PiperOrigin-RevId: 647294360
When handling a playback error that originates from a future item in
the playlist, we added support for jumping to that item first,
ensuring the errors 'happen' for the right 'current item'.
See 79b688ef30.
However, when we add this new position discontinuity to the
playback state, there may already be other position discontinuities
pending from other parts of the code that executed before the
error. As we can't control that in this case (because it's part
of a generic try/catch block), we need to send any pending
updates first before handling the new change.
Issue: androidx/media#1483
#cherrypick
PiperOrigin-RevId: 646968309
Setting a `null` value doesn't remove the key as expected per the `MediaFormat` API documentation, using the `removeKey` method instead which is only available starting API level 29.
PiperOrigin-RevId: 646462402
ExoPlayer used to call `stop()` before `release()`. This was removed in
<unknown commit>.
A framework bug introduced in Android 11 (API 30) resulted in some
DRM -> clear transitions failing during `MediaCodec.configure()`. An
investigation in Issue: google/ExoPlayer#8696 and b/191966399 identified that this was
due to `release()` returning 'too early' and the subsequent
`configure()` call was then trying to re-use a `Surface` that hadn't
been fully detached from the previous codec. This was fixed in
Android 13 (API 33) with http://r.android.com/2094347.
ExoPlayer worked around the framework bug by adding an arbitrary 50ms
sleep after a failed codec initialization, followed by retrying. This
was enough to resolve the problem in the test scenario on a OnePlus
AC2003.
Issue: androidx/media#1497 points out that 50ms might not be the appropriate delay
for all devices, so it's an incomplete fix. They suggested re-adding the
`MediaCodec.stop()` call instead. This also reliably resolves the issue
on the OnePlus AC2003 (with neither workaround in place, the problem
repros almost immediately).
PiperOrigin-RevId: 646461943
Extractors should not report additional tracks once they called
ExtractorOutput.endTracks. This causes thread safety issues in
ProgressiveMediaPeriod where the array of sample queues is
extended while the playback thread accesses the arrays.
Detecting this problem early is beneficial to avoid unexplained
exceptions later one. In most cases where this may happen (namely
TS extractors finding new tracks), it's better to ignore the new
tracks instead of failing completely. So this change adds a
warning log message and assigns a placeholder output.
Note: The same workaround already exists in HlsSampleStreamWrapper
and MediaExtractorCompat.
Issue: androidx/media#1476
#cherrypick
PiperOrigin-RevId: 646427213
The two arrays need to have the same length and the selection
must match in their nullness (unless for TYPE_NONE
renderers). Clarify this more clearly in the docs and add
new asssertions for it. This avoids that the player is failing
in obscure ways much later.
Issue: androidx/media#1473
#cherrypick
PiperOrigin-RevId: 646086833
The test is flaky because the decoding process in the renderer
depends on some timing from MediaCodec beyond our control and
the new keyframe added in the test is sometimes 'dropped' when
it arrives too late.
We can fix this by controlling the test progress a bit more
tightly: first rendering with the same current time until the
key frame is processed and then start increasing the time
until we've reached the end.
#cherrypick
PiperOrigin-RevId: 646064352
Summary:
This change aims to add a generic `CustomData` field to the `Format` class.
The intent is to allow ExoPlayer customers to add extra data to the Format class without forcing
specific data to be included, impacting customers that do not need it and would allow for the data
to be changed without requiring changes to the `Media3` codebase.
Some cases are not handled correctly at the moment:
- Pausing during suppressed playback should not clear the
suppression state.
- Transient focus loss while paused should be reported as
a playback suppression.
Issue: androidx/media#1436
#cherrypick
PiperOrigin-RevId: 644971218
Sometimes the reason for the current state may change. If we don't
report this again, users have no way of knowing that the reason
changed.
Also adjust ExoPlayerImpl and MediaControllerImplBase accordingly.
SimpleBasePlayer already adheres to this logic.
#cherrypick
PiperOrigin-RevId: 644970236
This helps to keep the reason always together with the state it
is referring to, avoiding any side channels and making sure there
are no accidental inconsistencies.
#cherrypick
PiperOrigin-RevId: 644969317
The 'player commands' returned to ExoPlayerImpl instruct the
player on how to treat the current audio focus state.
The current return value when playWhenReady==false is misleading
because it implies we are definitely not allowed to play as if
we've lost focus. Instead, we should return the actual player
command corresponding to the focus state we are in.
This has no practical effect in ExoPlayerImpl as we already
ignore the 'player command' completely when playWhenReady=false.
To facilitate this change, we also introduce a new internal
state for FOCUS_NOT_REQUESTED to distinguish it from the state
in which we lost focus.
#cherrypick
PiperOrigin-RevId: 644416586
There are a lot of tests for AudioFocusManager in isolation,
but almost none for the handling in ExoPlayer.
Add test coverage for all the common cases, including some
currently broken behavior that is indicated by TODOs.
PiperOrigin-RevId: 644319251
This method allows customizing the maximum position when using `Player.seekToPrevious()`.
This commit also adds two new methods to `TestExoPlayerBuilder`:
- `setMaxSeekToPreviousPosition(long)`
- `getMaxSeekToPreviousPosition()`
This change introduces a new class in Media3 `MediaExtractorCompat`, designed to be a drop-in replacement for platform `MediaExtractor`. While not all APIs are currently supported, the core functionality for the most common use cases of `MediaExtractor` is now available. Full API compatibility will be achieved in the future.
PiperOrigin-RevId: 643045429
When deselecting the single sample track and later re-selecting this
track, the current shortcuts in ProgressiveMediaPeriod don't handle
this case correctly and cause assertion failures.
In particular, this change fixes 3 issues:
1. When re-selecting the single sample track, we have cleared the
SampleQueue and need to reload the sample. The existing shortcut
should only be applied to avoid the reload when starting from a
non-zero position.
2. When de-selecting the track, ProgressiveMediaPeriod is left in
an inconsistent state where the sample queues are empty but
loadingFinished is still true. Fix this by resetting
loadingFinished to false.
3. When seeking, we avoid reloading the stream if we can keep
inside the existing samples. This logic assumes that all
remaining samples will continue to be loaded in the queue.
This condition isn't true though for single sample tracks
that have been de-selected. They appear to support the seek
inside the queue (=no seek necessary, always supported), but
still require a new load if there is no ongoing one to load
the sample. Fix this by checking this implicit assumption
(still loading, or loading finished).
PiperOrigin-RevId: 642650248
A fatal `PlaybackException` is mapped to a legacy playback state
in state `STATE_ERROR` with error code, message and extras. A
non-fatal error sent to controllers with `MediaSession.sendError`
is synced to the legacy session by setting error code and message
and merging the extras while preserving the rest of the state in
sync with the session player.
Vice versa, a `MediaController` connected to a legacy session receives
fatal errors through `Player.onPlayerErrorChanged()` and non-fatal errors
through `MediaController.Listener.onError()`.
Error codes are mapped in `LegacyConversions`. Values of error codes
in `@SessionError.ErrorCode` come from `@PlaybackExceptino.ErrorCode`
with the exception of `@SessionError.ERROR_IO` and
`@SessionError.ERROR_UNKNOWN`. These already exist in
`@PlaybackException.ErrorCode` and are mapped accordingly to avoid
semantic duplicates.
PiperOrigin-RevId: 642595517
- Video release should check for buffer timestamp (which is renderer-offsetted), rather than the frame timestamp
- ImageRenderer should report ended after all of it's outputs are released, rather than when finished consuming its input.
Add tests for timestamp handling
PiperOrigin-RevId: 642587290
It's currently not possible to remove a previously set image output
on ExoPlayer, although the underlying renderer already supports
receiving null to clear the output. Marking the parameter as
nullable allows apps to clear it as well.
PiperOrigin-RevId: 642569081
In order to do that, make the VideoSink nullable in MCVR.
We want to avoid calling VideoFrameReleaseControl.setClock directly
from MCVR when the sink is enabled. The goal is to handle all the
communication with the release control from the sink/sink provider.
PiperOrigin-RevId: 642542063
These are no longer needed now that the `Bundleable` interface has been
removed. Public methods are deprecated, package-private ones are
removed. Callers are migrated in both cases (except where tests
explicitly exist for the deprecated method).
PiperOrigin-RevId: 642294451
This interface is not used in the library. Callers can use the
`Bundle toBundle()` and `static Foo fromBundle(Bundle)` methods
defined directly on each type instead.
PiperOrigin-RevId: 642271609
The previous wording suggested that `setVideoEffects()` may **only** be
called before `prepare()`, i.e. the effect cannot be changed during
playback. The intent is instead that `setVideoEffects()` must be called
once before playback in order to configure the effects pipeline, but
the effect can then be changed during playback by further calls to
`setVideoEffects()`.
Issue: androidx/media#1393
PiperOrigin-RevId: 641853629
Before, the listener was set in onReadyToInitializeCodec, which means
that it was reset every time a new codec was used.
We need to set the listener every time MCVR is enabled (not only the
first time), because it might have been set by another renderer.
PiperOrigin-RevId: 641825717
Before this change, the timestamps output from composition playback is offset
with the renderer offset. After this change, the offset is removed and the
timestamp behaviour converges with Transformer, that is, the timestamps of
video/images frames will follow that of the composition. For example, with a
composition of two 10-s items, clipping the first with 2s at the start, the
timestamp of the first frame in the second item, will be 8s.
PiperOrigin-RevId: 641121358
When the frame release control invalidates a buffer and returns that
the buffer must be ignored, we need to exit early before performing
additional checks that may result in method calls using the invalid
buffer.
PiperOrigin-RevId: 640555688
Usages removed in this CL are:
- onProcessedStreamChange, which was already called from the VideoSink
(via VideoFrameRenderControl)
- setOutputSurface, which was also already called from the VideoSink
- setFrameRate, which this CL now sets in the VideoSink
PiperOrigin-RevId: 640530903
Before this CL, we were checking whether the video sink was initialized
to determine whether it should be used. In the meantime,
shouldUseVideoSink has been introduced. Use this boolean to check
whether to use the video sink as it's a clearer signal.
PiperOrigin-RevId: 640499147
This avoids having to add AnalyticsListener and catching the `onPlayerReleased` callback.
The `Exoplayer.release()` method is blocking and one can be sure that the player is released if the call returned. However, the method is useful for UI testing and asserting that the player is released after a certain UI action, e.g. closing the activity or detaching a window.
PiperOrigin-RevId: 640114416
The CMCD data was incorrectly added to the `dataSpec` of the media segment instead of the init segment.
Also relaxed the condition for playbackRate to be C.RATE_UNSET when creating an instance of CmcdData.Factory as there was nothing enforcing this check.
#minor-release
PiperOrigin-RevId: 639046080
When running in asynchronous mode, MediaCodec will be running the CPU to signal input and output buffers being made available for use by the player. With ExoPlayer.experimentalSetDynamicSchedulingEnabled set to true, ExoPlayer will wakeup to make rendering progress when MediaCodec raises these signals. In this way, ExoPlayer work will align more closely with CPU wake-cycles.
PiperOrigin-RevId: 638962108
Currently ExoPlayer schedules its main work loop on a 10 ms interval. When renderers cannot make any more progress (ex: hardware buffers are fully written with audio data), ExoPlayer should be able to schedule the next work task further than 10ms out into the future.
Through `experimentalSetDynamicSchedulingEnabled` and these changes to `MediaCodecAudioRenderer`, ExoPlayer can use the data provided by the audio renderer to dynamically schedule its work tasks based on when it is expected that progress can be made.
PiperOrigin-RevId: 638677454
Currently ExoPlayer schedules its main work loop on a 10 ms interval. When renderers cannot make any more progress(ex: hardware buffers are fully written with audio data), ExoPlayer should be able to schedule the next work task further than 10Ms out.
Through `experimentalSetDynamicSchedulingEnabled`, ExoPlayer will dynamically schedule its work tasks based on when renderers are expected to be able to make progress.
PiperOrigin-RevId: 638676318
If subtitle-parsing-during-extraction is enabled (now defaults to on),
the 'outer' extractor class name is often
`SubtitleTranscodingExtractor`, leading to some slightly useless error
messages like:
`None of the available extractors (FragmentedMp4Extractor, Mp4Extractor, SubtitleTranscodingExtractor, SubtitleTranscodingExtractor, SubtitleTranscodingExtractor, SubtitleTranscodingExtractor, SubtitleTranscodingExtractor, SubtitleTranscodingExtractor, TsExtractor, MatroskaExtractor, SubtitleTranscodingExtractor, SubtitleTranscodingExtractor, SubtitleTranscodingExtractor, SubtitleTranscodingExtractor, AviExtractor, SubtitleTranscodingExtractor, SubtitleTranscodingExtractor, SubtitleTranscodingExtractor, SubtitleTranscodingExtractor, SubtitleTranscodingExtractor, SubtitleTranscodingExtractor)`
PiperOrigin-RevId: 636834354
If MediaCodec allocates passes an image buffer with a cropped region,
SurfaceTexture.getTransformMatrix will cut off 2 pixels from each dimensions.
The resulting videos will appear a little stretched.
This patch inspects the SurfaceTexture transform matrix, and guesses what the
unscaled transform matrix should be.
Behind experimentalAdjustSurfaceTextureTransformationMatrix flag
PiperOrigin-RevId: 635721267
This case is most likely to happen when re-preparing a multi-period
live stream after an error. The live timeline can easily move on to
new periods in the meantime, creating this type of update.
The behavior before this change has two bugs:
- The player resolves the new start position to a subsequent period
that existed in the old timeline, or ends playback if that cannot
be found. The more useful behavior is to restart playback in the
same live item if it still exists.
- MaskingMediaSource creates a pending MaskingMediaPeriod using the
old timeline and then attempts to create the real period from the
updated source. This fails because MediaSource.createPeriod is
called with a periodUid that does no longer exist at this point.
We already have logic to not override the start position and need
to extend this to also not prepare the real source.
Issue: androidx/media#1329
PiperOrigin-RevId: 634833030
The two affected tests where playing until a specific
position to enable the player to read ahead. The method
pauses at exactly the target position, but then has
temporarily undetermined behavior because the playback
thread uses player.getClock().onThreadBlocked() that lets
the playback thread make progress in parallel to the test
thread. The tests were flaky because they sometimes made
so much progress that they ended playback before we could
query the updated renderer state.
This can be fixed by using
run(player).untilBackgroundThreadCondition instead, which
is guaranteed to be fully deterministic, but may not be able
to stop at exactly the desired position (which we don't
really need anyway for this test).
PiperOrigin-RevId: 634699752
The method is only allowed to be called on prepared items.
This check was currently missing and also causing the
corresponding test to be flaky in ExoPlayerTest.
PiperOrigin-RevId: 634694077
Before this change, if a playback error is thrown the test fails with a
timeout and no additional info:
```
java.util.concurrent.TimeoutException
at androidx.media3.exoplayer.e2etest.WebvttPlaybackTest.stallPlayerUntilCondition(WebvttPlaybackTest.java:361)
```
After this change, the test failure includes a much more useful stack
trace, e.g. from 0352db9a37:
```
Caused by: java.lang.IllegalStateException: Legacy decoding is disabled, can't handle text/vtt samples (expected application/x-media3-cues).
at androidx.media3.common.util.Assertions.checkState(Assertions.java:100)
at androidx.media3.exoplayer.text.TextRenderer.assertLegacyDecodingEnabledIfRequired(TextRenderer.java:587)
at androidx.media3.exoplayer.text.TextRenderer.onStreamChanged(TextRenderer.java:210)
```
PiperOrigin-RevId: 634672138
To override this change, and go back to parsing during rendering,
apps must make two method calls:
1. `MediaSource.Factory.experimentalParseSubtitlesDuringExtraction(false)`
2. `TextRenderer.experimentalSetLegacyDecodingEnabled(true)`
PiperOrigin-RevId: 634262798
The test currently resets the time too far in the past and then has
to run through ~30000 additional iterations of doSomeWork to reach
the end, sometimes triggering the test timeout.
Fix it by resetting the time to the intended start position when
transitioning items.
PiperOrigin-RevId: 633918706
This annotation is only needed on public classes.
This change also removes the `/* package */` comment from some `public`
classes.
PiperOrigin-RevId: 633864544
This means the content source is 'prepared' instantly with a
placeholder, enabling all further preparation steps (e.g. loading
preroll ads) while the actual content is still preparing. This
improvement can speed up the start time for prerolls in manifest-based
content that doesn't have a zero-time preparation step like progressive
media.
Issue: androidx/media#1358
PiperOrigin-RevId: 633640746
This allows apps to better detect when the platform
reclaims a codec. This requires adding the error code
to MediaCodecDecoderException.
PiperOrigin-RevId: 633588914
The class currently tracks the input format itself, updating it too
early in onConfigure() instead of onFlush(). This causes issues when
the format changes and the new values are applied to the silence
skipping logic of the old format. The fix is to use the base class
input format handling instead.
Issue: androidx/media#1352
PiperOrigin-RevId: 633232368
Due to loading a MediaSourceFactory via reflection:
Before this change
* the content type was logged as an integer, rather than a human-readable string
* `ClassNotFoundException` was swallowed silently by `maybeLoadSupplier` without telling the user what module they were missing
After:
* ClassNotFoundException is swallowed silently ONLY when determining supported types
* ClassNotFoundException is bubbled up when we are trying to play media without the corresponding module properly loaded
PiperOrigin-RevId: 632568989
The same doSomeWork iteration that triggers the silence skipping
discontinuity may already have another discontinuities (like
AUTO_TRANSITION), which should take precedence over the silence
skipping.
PiperOrigin-RevId: 632432851
When the period has loaded to the end of the source, the `period.getBufferedPositionUs` will be set to `C.TIME_END_OF_SOURCE`, which is a negative value. Thus, the original `continueLoadingPredicate` will never turn to `false`, as the `bufferedPositionUs` is definitely less than the target preload position that is expected to be positive.
In this change, we added `PreloadMediaSource.PreloadControl.onLoadedToTheEndOfSource(PreloadMediaSource)` to indicate that the source has loaded to the end. This allows the `DefaultPreloadManager` and the custom `PreloadMediaSource.PreloadControl` implementations to preload the next source or take other actions.
This bug was not revealed by the the `DefaultPreloadManagerTest` because the related tests were all using the `FakeMediaSource` and only setting the preload target to `STAGE_TIMELINE_REFRESHED`. Thus, the tests for testing the `invalidate()` behaviors were modified to use the real progressive media whenever possible, unless we have to use `FakeMediaSource` to squeeze a chance to do more operations between the preloading of sources to test some special scenarios.
PiperOrigin-RevId: 631776442
For offloaded playback, reset the tracking field for stream completion in `DefaultAudioSink` prior to calling `AudioTrack.stop()` so that `AudioTrack.StreamEventCallback#onPresentationEnded` correctly identifies when all pending data has been played.
#minor-release
PiperOrigin-RevId: 631744805
When the PlaybackParameters are set to their DEFAULT value, we
currently bypass the audio processor chain when determining the
output media position, under the assumption that no timestamp
change happens in the audio processors. This assumption may not
be true as the audio processors can change playout durations on
their own accord independent of the provided PlaybackParameters.
To correctly reflect any updated playout duration, we can just
always check the audio processor chain. The default implementation
will continue to assume that only the SonicAudioProcessor changes
the playout duration.
PiperOrigin-RevId: 631726112
Without it, the DAC doesn't render the full sentence. And the link is not actually linking to the proper constructors.
#minor-release
PiperOrigin-RevId: 630395271
Set `VideoSink`'s offset during seeking in `MCVR.onPositionReset()`
This one is necessary in some cases, where `onProcessedStreamChange()` is not
invoked during a seek. For example, when seeking when playback has ended.
PiperOrigin-RevId: 630056723
To find the column of an index in a matrix the formula "column = index % width" should be used, not "column = index % height"
If inputFormat.tileCountVertical was equal to 1 then it would not throw an error, but instead result in the first tile of the bitmap always being returned. If inputFormat.tileCountVertical was larger than 1 then Bitmap.createBitmap() would throw an error as it would attempt to go outside the bounds of outputBitmap
ImageRenderTest has been updated to test for 2x3 images so that tileCountVertical != tileCountHorizontal. These tests passed previously because they were equal, so using tileCountVertical produced the same results as tileCountHorizontal
The last rebuffer time was being updated erroneously, even in the absence of rebuffering events, resulting in incorrect `bs` (buffer starvation) key in CMCD.
Issue: androidx/media#1124
PiperOrigin-RevId: 629731796
These subtitles were skipped because they are marked as shouldBeSkipped
based on their timestamps. The fix removes this flag entirely in
SimpleSubtitleDecoder because TextRenderer handles potential skipping
if needed.
PiperOrigin-RevId: 629717970
`updateRebufferingState` is invoked immediately preceding this `if-else`, with no alteration of state occurring in between, making this invocation unnecessary.
PiperOrigin-RevId: 629694531
This lets apps update the task manager priority and send the
priority message to all renderers so that they can adjust their
resources if needed.
PiperOrigin-RevId: 629426058
This change adds a 'free-form' text genre to the existing
`bear-id3.mp3` sample, and adds a new sample with a 'numeric' genre ([9
is Metal](https://mutagen-specs.readthedocs.io/en/latest/id3/id3v2.4.0-frames.html#appendix-a-genre-list-from-id3v1)).
The samples were modified with:
```shell
$ id3edit --set-genre "Gorpcore" bear-id3.mp3
$ id3edit --set-genre "9" bear-id3-numeric-genre.mp3
```
Reading the numeric genre with `exiftool` shows the mapping back to
'Metal':
```
$ exiftool bear-id3-numeric-genre.mp3 | grep Genre
Genre : Metal
```
The playback dumps don't contain the genre because it's not yet
propagated to `MediaMetadata.genre`. That is done in a follow-up
change.
Issue: androidx/media#1305
PiperOrigin-RevId: 629043506
This removes a window of inconsistency between the timeline
known to the player and the most recent timeline in the
`MediaSourceList` by removing the `MSG_DO_SOME_WORK` when
sending `MSG_PLAYLIST_UPDATE_REQUESTED`. `MSG_DO_SOME_WORK`
is then sent again when the playlist update is requested.
PiperOrigin-RevId: 629021752
Some test cases are still using `ConcatenatingMediaSource`
even if they do not test specific features of the concatenating
source. Apparently these test have a slightly different timing
behavior when it comes to updating the `MediaPeriodQueue` and
emitting change events with analytics.
Using the playlist API ensures testing the future-proof code path.
PiperOrigin-RevId: 628413460
The first has a string genre, and various other values set, generated
from `sample.mp4` with the command below [1].
The second has a numeric genre, to test `gnre` atom parsing. This
parsing is currently broken, the fix is in a follow-up change. This
file was also generated from `sample.mp4` with the command below [2].
This change also includes `CommentFrame.text` in its `toString`
representation, otherwise there's no difference between e.g. different
levels of `ITUNESADVISORY` in the extractor dump files.
Issue: androidx/media#1305
-----
[1]
```shell
$ AP_PADDING="DEFAULT_PAD=0" \
AtomicParsley sample.mp4 \
--artist "Test Artist" \
--album "Test Album" \
--tracknum 2/12 \
--disk 2/3 \
--year 2024 \
--genre "Gorpcore" \
--bpm 120 \
--compilation true \
--advisory clean \
--gapless true \
--sortOrder artist "Sorting Artist" \
--sortOrder album "Sorting Album" \
--preventOptimizing \
-o sample_with_metadata.mp4
```
[2]
```shell
$ AP_PADDING="DEFAULT_PAD=0" \
AtomicParsley sample.mp4 \
--genre "Metal" \
--preventOptimizing \
-o sample_with_numeric_genre.mp4
```
PiperOrigin-RevId: 628345458
Before this CL, the video sink was not reconfigured for the second video
in a sequence with video-image-video. For example, the stream offset
and listener were not set for the second video.
PiperOrigin-RevId: 628065991
Instead of initializing the video sink outside the renderer with an
empty format for composition preview, we initialize it in the renderer
with the input format for video.
PiperOrigin-RevId: 627313708
When playing a Composition, the color transfer of an image is incorrectly
passed down to be SMPTE170M, but it should be SRGB.
PiperOrigin-RevId: 626425396
For audio with more than 12 channels, no channel mask was determined, which
meant that the code to check spatializability would throw because of creating
an invalid audio format.
Return early if the channel mask was invalid instead (and assume spatialization
isn't possible).
PiperOrigin-RevId: 625618683
Before this CL, the FrameMetadataListener was set in
onReadyToInitializeCodec while it had already been set when handling
message MSG_SET_VIDEO_FRAME_METADATA_LISTENER.
PiperOrigin-RevId: 624940824
Some seek operations are currently filtered in the base classes if
the target index is not explicitly specified and the implicitly
assumed operation doesn't have an obvious target index. (Example:
calling seekToNext() at the last item in a playlist)
This is too opinionated because the actual player implementation
using this base class may be able to handle this seek request (e.g.
by adding new items on the fly or using other logic to select
the most suitable next item).
This can be solved by forwarding all seek requests to the respective
handler methods even if they are a presumed no-op. Also clarify the
Javadoc that the provided index is just an assumption if it wasn't
provided explicitly in the method call.
PiperOrigin-RevId: 624887116
Before this CL, the offset was set:
- in onPositionReset for composition preview
- in onReadyToInitializeCodec for ExoPlayer
The code flow is easier to reason about if the path for both use cases
is as shared as possible.
PiperOrigin-RevId: 624870150
This removes the TODOs without updating the links, because the
DAC-hosted images are not co-located with the hosted javadoc (unlike
when these images were referenced on exoplayer.dev before 10342507f7),
and therefore we would need to include the full path anyway, at which
point it seems clearer and more robust to keep using a fully-qualified
URL with the domain too.
PiperOrigin-RevId: 623452217
Compared to `release`, the `reset` method doesn't release the preload manager instance. This applies to the use case that an app wants to discard all the sources but keep the preload manager active for later usage.
Also rename the `releaseSourceInternal` to `removeSourceInternal`, as the latter sounds more generic for different preload manager implementations.
PiperOrigin-RevId: 623148723
