Its only real purpose is to encapsulate a download ID, but it doesn't actually save
any code, and arguably makes it more complicated by having multiple lists of
Downloader instances, indexed by key, and also because it's another (undocumented)
class to understand.
This CL retains the multiple Downloader lists, but they will be flattened in a
child CL.
PiperOrigin-RevId: 308838026
This is equivalent to the method ExoPlayer provides. It's nice for consistency,
and for retrieving the looper from test code.
PiperOrigin-RevId: 308830288
DownloadRunner.getDownloader was creating Downloader instances before the actual
DownloadManager under test would have created them. Tests were then asserting and
manipulating these Downloader instances, which is quite confusing. In particular
FakeDownloader.assertDoesNotStart() is an assertion on a Downloader instance that
only makes sense when called on an instance that DownloadManager would not have
created by itself.
This change replaces FakeDownloader.assertDoesNotStart() with an assertion on
DownloadRunner that no Downloader instance has been created.
PiperOrigin-RevId: 308822398
- Assertion chaining is generally discouraged. For example, because it's harder
to determine which assertion failed given a line number.
- Also removed chaining of the form obj.actionX().assertY(), because it's easy
for someone editing the test to accidentally delete actionX() when deleting
assertY(), where-as actionX() may often be important for subsequent assertions.
PiperOrigin-RevId: 308820503
- Remove assertReleased and replace it with a proper condition variable
that's opened when Downloader.download or Download.remove finish. As
far as I can tell assertReleased was basically implementing "sleep for
10 seconds after the Downloader starts". Note fixing this properly
also makes the tests run much faster!
- Use ConditionVariable instead of CountDownLatch(1).
- Use AtomicInteger instead of volatile int because it's clearer and
allows removal of explanatory comments.
PiperOrigin-RevId: 308819204
- Improve documentation explaining the benefits of ExoPlayer's ConditionVariable
over the one that the platform provides
- Allow Clock injection
- Create TestUtil method for obtaining a ConditionVariable whose block(long)
method times out correctly when used in a Robolectric test
- Add basic unit tests for ConditionVariable
PiperOrigin-RevId: 308812698
A previous change added these getters to CacheDataSource, but it can
also be useful to retrieve these components directly from the factory.
This is useful for tasks where we're going to need to build multiple
CacheDataSource instances (e.g., to make requests in parallel), and
also need to operate directly on the same components. It's a bit more
natural to retrieve them from the factory than from an arbitrary
CacheDataSource in this case, since it can avoid unnatural code where
you create a CacheDataSource instance earlier than you would otherwise
just to use its getters, and/or create one just to use its getters and
then throw it away.
PiperOrigin-RevId: 308606020
- Executor is a superclass of ExecutorService, so this is arguably a little
more flexible.
- It removes the need to use null for direct execution, because Runnable::run
is a direct executor that can be trivially used instead.
- Removing the error-prone "cannot be a direct executor" restriction in the
parallel version of SegmentDownloader requires not relying on the futures
returned from ExecutorService.submit() anyway.
Issue: #5978
PiperOrigin-RevId: 308586620
Currently, this method is only supposed to be called before removing
the listener from the player or when releasing the player.
If called at other times, it will throw an exception later when
a playback session is ended automatically.
issue:#7193
PiperOrigin-RevId: 308254993
The method is used to finish a brief "seeking" state that can be removed entirely
now state changes for seeking are masked.
PiperOrigin-RevId: 308237009
Unmarshal from json to MediaItem instead of Sample. Further the playlist
of MediaItems is converted to Intent extras which are read by the
PlayerActivity.
PiperOrigin-RevId: 308141231
While most ExoPlayer code parsing ByteBuffers is called with buffers in big
endian, in certain situation, buffers in little endian are used too.
MediaCodec produced ByteBuffers are in little endian, while buffers
receive from the sources are in big endian (ByteBuffer's default).
As a result, some code called from AudioSink in passthrough parsed
bytebuffer in little endian. This is not correct because those
format are specified in BigEndian.
Changing the endianness of the ByteBuffer returned from MediaCodec
would impact a lot more code that can currently be tested in the
current COVID lockdown situation.
As a result, this patch instead make the parsing code independent
of the ByteBuffer.order() set. All the code that is called from
DefaultAudioSink now parses the buffer explicitly in Big Endian.
Additionally, the MPEG big endian header data of size 4 bytes was
retrieved with ByteBuffer.get, which only returns one byte.
PiperOrigin-RevId: 308116173
The first session should only be created once we have the media items
and/or called prepare. Otherwise the first session is created with
an EventTime having an empty timeline making it less useful.
issue:#7193
PiperOrigin-RevId: 308100555
Part of what makes these tests hard to deal with (imo) is being
unable to easily run a specific config, or seeing exactly which one
failed (because you always see only the first failure).
I put the parameters as a method on ExtractorAsserts to
reduce the boiler-plate required in each test class.
I'll migrate the extension FlacExtractorTest in a follow-up CL
PiperOrigin-RevId: 307785380
Something that helps a constructor always seemed a bit strange.
It's now possible to use CacheDataSource.Factory directly instead.
PiperOrigin-RevId: 307661930
In the same way it made sense to tie CacheKeyFactory and CacheDataSource
together in
7ea83d7167,
it makes sense to tie the PriorityTaskManager to the CacheDataSource as
well. This prevents error prone scenarios where one can end up being
passed around without (or with the wrong instance of) the other.
This change also internalizes use of PriorityDataSource directly into
CacheDataSource, rather than requiring the caller to chain things
themselves.
PiperOrigin-RevId: 307647290
The sample timestamps are currently rounded to milliseconds, only to
be multiplied by 1000 later. This causes rounding errors where the sample
timestamps don't match the timestamps in the seek table (which are already
in microseconds).
issue:#7086
PiperOrigin-RevId: 307630559
This change generalizes the concept of "reading parameter sets" to
"reading prefix NAL units", ahead of a change that will treat AUD
and suffix SEI NAL units in the same way.
The change also introduces some static isXxxNalUnit methods for
clarity.
Issue: #7113
PiperOrigin-RevId: 307376967
Audio processors are now flushed twice after reconfiguration.
The second flush call cleared the pending trim start bytes so
transitions between tracks were no longer gapless.
Fix this by removing logic to clear pending trim bytes on flush.
As a result we may trim data incorrectly if there is a flush
before any data has been handled for seeking to a non-zero
position, but this edge case will happen rarely and the effect
shouldn't be noticeable.
PiperOrigin-RevId: 307344357
This is the missing attribute to support all features of the Sample with MediaItem. Hence PlayerActivity can use the setMediaItems() method directly without creating actual media sources in the app code.
PiperOrigin-RevId: 307102036
This makes it easier to use the helper methods because the player doesn't
have to be assigned to a SimpleExoPlayer in all cases.
PiperOrigin-RevId: 307039126
Update the comment on a suppression of Field.get(null) which is safe
but blocked by the checker.
These suppressions were introduced in d1e0572448
PiperOrigin-RevId: 307036441
We currently check for shouldContinueLoading, which is the intention to load,
not playbackInfo.isLoading, which is the actual loading state, when detecting
stuck buffering issues.
They only differ if the LoadControl said stop loading (based on nextLoadPosition),
but there is still a load in flight (updating bufferedPosition). This may cause
the exception to be thrown in edge cases that are only temporary.
PiperOrigin-RevId: 307022608
This is required to migrate the PlayerActivity away from Sample to MediaItem. It hence needs adding buildUpon to MediaItem to mix in the customCacheKey and streamKeys before playback.
PiperOrigin-RevId: 306710643
This is a more specific ViewGroup subclass that handles some of the
layout logic automatically. It's designed to work best with a single
child view, as used here.
PiperOrigin-RevId: 306654947
Even though they are annotated as non-null, Java users can pass in null values.
So we should verify the input to fail fast.
PiperOrigin-RevId: 306425487
This is the first CL for the offline qoe reporting
feature. Refer to the design doc for full information.
Design doc: go/exoplayer-offline-qoe
PiperOrigin-RevId: 305459231
Create a Builder that creates SimpleExoPlayer instances with fake
components, suitable for testing.
Basically extracts the Builder from ExoPlayerTestRunner to a standalone
class that can be re-used.
PiperOrigin-RevId: 305458419
This CL is a renaming only. It's mostly about finding a decent naming for the internal Playlist class. The plan is to have a public Playlist class in the converged Player API, so we need to rename the internal one.
PiperOrigin-RevId: 305266196
If we're in the ducked state and updateAudioFocus is called with a
new state for which focus is no longer required, we should restore
the player back to full volume.
Issue: #7182
PiperOrigin-RevId: 305232155
These were introduced in c7164a30a0
In each case I checked that the groups are not optional,
so if they match they must be non-null.
PiperOrigin-RevId: 305213293
This avoids cases where audio focus is never successfully acquired
because another app is holding on to transient audio focus indefinitely.
Issue: #7182
PiperOrigin-RevId: 305108528
When ClippingMediaPeriod first tried to read a buffer, if its end
position was before the end of the stream and it was buffered to its end
position, it would sometimes erroneously signal end-of-stream for
protected content because the sample queue might be waiting for DRM keys
at this point.
Work around the issue temporarily by signaling this specific case back
to ClippingMediaPeriod via the DecoderInputBuffer.
There will likely be a cleaner fix as a result of adding support for
dynamic clip end points in the future, at which point this can be
reverted.
issue:#7188
PiperOrigin-RevId: 305081757
It's interesting WebVTT explicitly handles line & position differently
in horizontal-rl/lr and vertical-lr/rl contexts. position is always
measured from the left of the viewport, even for rtl text, but line in
vertical-rl is measured from the right of the viewport.
We don't have to make Cue match WebVTT (I can go change the WebVTT
decoding instead) but it seems a reasonable 'default' to follow.
PiperOrigin-RevId: 304353900
ExoPlayer needs a codec to decide among WEBVTT and TTML decoder mimeType.
Apple describes IMSC1 in MP4 in
[RFC-8216 Section 3.6](https://tools.ietf.org/html/draft-pantos-hls-rfc8216bis-04#section-3.6).
The DASH manifest specifies the SMPTE-TT captions in the codecs in the manifest
(from W3C [TTML Profiles for Internet Media Subtitles and Captions 1.1](https://www.w3.org/TR/ttml-imsc1.1/#general-0).
DASH just doesn't require the rendition linking, but HLS does.
Apple implies the CODECS attribute of the variant needs to be do this. That is
with SHOULD and MAY language to imply the codec to use for it in the
[Authoring Guidelines](https://developer.apple.com/documentation/http_live_streaming/hls_authoring_specification_for_apple_devices)
This change defaults to WebVTT if no codec is specifed (same as current behavior) otherwise it picks it from the variants
referencing the media.
Without this, URL-encoding is assumed, which means ampersand-codes are
not carried through to the underlying web page correctly.
PiperOrigin-RevId: 304163733
HlsSampleStreamWrapper currently calls HlsMediaChunk.init, which calls back
to HlsSampleStraemWrapper.init. This re-entrancy seems a bit confusing.
PiperOrigin-RevId: 304139462
Selection of a passthrough codec will rarely (if ever) need to be customized, so
remove this capability from MediaCodecInfo.
Applications can still customize whether passthrough is used by overriding
MediaCodecAudioRenderer.usePassthrough, which now also checks for a passthrough
codec.
#exo-offload
PiperOrigin-RevId: 303964682
This is used to generate the initialization vector for encrypting the
cache contents.
Startblock:
<unknown commit> is submitted
and then
3w have passed
PiperOrigin-RevId: 303932151
This avoids massive changes to method signatures to add throws. Also, took suggestion to make it an `IllegalStateException`.
Move the catch outside of the finally that sets `nextLoadPosition` (this allows for possible recovery by reseting the `Extractor` and `TimestampAdjuster`).
Lastly, took the suggestion to make a minimum value for the tolerance (especially usefull for very short i-Frame only segments).
CharSequence is used by Notification builders and allows to set Spannable text.
It's the base interface of the String class, so apps wouldn't be break after
the update.
PiperOrigin-RevId: 303731890
This is required to align google3's stub (currently marked non-null) with Checker Framework's.
More information: go/matcher-nullness-lsc
Tested:
TAP train for global presubmit queue
http://test/OCL:303344687:BASE:303326748:1585344475427:29edc250
PiperOrigin-RevId: 303709053
- They are always listed with their canonical names
- Considering aliases means that blacklisting a decoder can end up
not actually blacklisting it, since it may still be accessible
via an alias. It also means that our decoder fallback logic can
end up falling back to a decoder that we've already tried!
PiperOrigin-RevId: 303348297
This callback will be deprecated, so moving all usages to better callbacks.
Some usages are still remaining that are less straight-forward to update.
PiperOrigin-RevId: 303298834
OnSeekProcessed is documented to be called as soon as all neccessary state changes
as a result of the seek have been made. As we now mask the state changes directly
when calling seekTo, we can also call this callback immediately without changing
the semantics of the method.
Our tests often use this callback as a way to wait for the internal player
to receive all pending commands and then report back. This is a special test
requirement for cases where we want to make sure no further updates happen as
a result of the player handling commands. To facilitate that, a new action is
added with a more descriptive name that achieves the same goal.
PiperOrigin-RevId: 303296210
This moves the playlist API methods to the Player interface. Implementation is moved from ExoPlayerImpl to BasePlayer where possible.
Further the CastPlayer is changed to implement the Player interface. Proper migration of the Playermanager to not use the ConcatenatingMediaSource anymore follows in a separate, future CL.
PiperOrigin-RevId: 302937779
Most of these are no longer needed since aa9eb5abc9
Cleanup change automatically generated by error-prone refactoring
//java/com/google/devtools/staticanalysis/errorprone:UnnecessaryJavacSuppressWarnings_refactoring on targets third_party/java_src/android_libs/exoplayer/v2/... java/com/google/android/libraries/exoplayer/v2/...
PiperOrigin-RevId: 302916092
This is less confusing than having audio processing functionality (e.g., playback
speed adjustment) just "not work" for some pieces of media.
If this change is merged, I will update #6749 to also track making DefaultAudioSink
intelligently enable/disable float output depending on how the audio processors are
configured.
Issue: #7134
PiperOrigin-RevId: 302871568
We currently have multiple places in ExoPlayerImpl that assign PlaybackInfo
instances and then inform listeners of all current changes. This is not ideal
because it causes multiple issues:
1. Some changes may easily be forgotten, e.g. there are clearly some checks
missing to see if isPlaying changed (e.g. in seekTo or setMediaSources)
2. Some callbacks didn't check if the value actually changed before sending
the callback (e.g. for the timeline change in setMediaSources - if the
timeline is still the same, we shouldn't send a onTimelineChanged event).
3. Having multiple callbacks in a single Runnable changes the order of
listener invocations slightly: Currently all events for one listener will
be send first before moving to the next listener. It should however send
a single event to all listeners first before moving to the next event.
All these issues can be solved by always using updatePlaybackInfo and never
assigning playbackInfo directly in another place.
Some tests needed to be updated as well because of issues (2) and (3). Also
added a new test to cover issue (1).
PiperOrigin-RevId: 302844981
It's incorrect to use an AudioFormat returned from AudioProcessor.configure
unless the AudioProcessor is active.
Issue: #7134
PiperOrigin-RevId: 302674132
This is a no-op for DefaultAudioSink for now, because DefaultAudioSink
currently disables processing anyway if the input uses ENCODING_PCM_FLOAT.
Issue: #7134
PiperOrigin-RevId: 302670534
androidTest of core already has the dependency on its main. Without this
exclude, gradle complains about type duplication when merging dex saying
"Type X is defined multiple times".
PiperOrigin-RevId: 302641585
If the start time of the edit falls within a sample, start from that
sample rather than the next one. This ensures playback can start from
the correct point if the sample is a keyframe, rather than having to
start from the next one.
Issue: #7133
PiperOrigin-RevId: 302639115
- Show renderers with no tracks in EventLogger track logging
- Log renderer names in EventLogger track logging
- Add useful message to ExoPlaybackException instances (including
renderer name for renderer errors)
PiperOrigin-RevId: 302421616
If MaskingMediaSource masks a multi-window media source, it may be that a period
is removed while we are using an initial unprepared masking MediaPeriod. That
means it's not guaranteed that a timeline update still contains our
unpreparedMaskingMediaPeriod and we should ignore timeline updates where the
period is no longer present because the it will be removed anyway.
PiperOrigin-RevId: 302383787
- This change also adds support for VideoFrameMetadataListener in the
AV1 renderer
- This is a preliminary step prior to adding FfmpegVideoDecoder
Issue: #2159
PiperOrigin-RevId: 301702460
This is a necessary step for Decoder implementations to support
audio and video. MediaCodecRenderer.DecoderException is renamed
MediaCodecDecoderException and extends the new DecoderException
Issue: #2159
PiperOrigin-RevId: 301698238
We have two known scenarios where the app could create an OOM error and
we want to handle it gracefully:
1. The app continues to allocate memory but doesn't make any progress
in the buffered position. OOM should be prevented by the default
load control and it should eventually throw an exception.
2. An extractor tries to allocate a large amount of memory on the
Loader thread based on information it read in faulty media files.
In this case we should attempt to play remaining media and then
throw an exception.
Both cases are already handled correctly, but we don't have any tests
ensuring that we don't introduce regressions.
PiperOrigin-RevId: 301585700
This removes a workaround that always continues buffering and instead
detects if the LoadControl returns false even though we don't have
any buffer. If enabled by a flag, this condition throws an exception.
PiperOrigin-RevId: 301584239
The restriction that these classes only work with SimpleDecoders
is unnecessary. An FfmpegVideoRenderer will not be able to use a
SimpleDecoder, because the SimpleDecoder assumption that each input
buffer can be decoded immediately into a corresponding output is
not true for all video codecs that Ffmpeg supports (e.g., H264 does
not have this property). Generalizing SimpleDecoderVideoRenderer to
DecoderVideoRenderer will allow FfmpegVideoRenderer to still use
the base class, without having to use a SimpleDecoder.
This is a preliminary change toward being able to merge a version
of https://github.com/google/ExoPlayer/pull/7079.
Issue: #2159
PiperOrigin-RevId: 301412344
The test is flkay (2/1000 runs) because the decoder initialization
before and after the seek are not perfectly deterministic. I couldn't
find a way to make them deterministic, so slightly chaning the test
setup instead. The test setup change doesn't affect the scenario being
tested.
PiperOrigin-RevId: 301390491
When no tracks are selected (or only tracks of unknown type), the
target buffer size is calculated to be 0. This means the player
won't request to buffer more data, nor can it start playback and
will be stuck forever.
PiperOrigin-RevId: 301374229
The assertion about the expected formats doesn't really belong in a
fake, the assertions should be closer to the test method.
This gets in the way when I try and write a new test in
AnalyticsCollectorTest that doesn't use the expected, constant Format
(because i want to specify drmInitData) - but changing the expected
Format is tricky because it's hard-coded into the FakeVideoRenderer
inner class.
I replaced the assertion in FakeRenderer with assertions in test
methods that used to assert on the format count.
PiperOrigin-RevId: 301353072
Previously calling removeListener would remove all instances of that
listener. Now it only removes a single instance.
This probably should have been part of introducing the Multiset:
2bd4d61b9b
PiperOrigin-RevId: 301191940
Without this change there's confusing behaviour if you pass e.g.
AnalyticsCollector (which implements both DrmSessionEventListener and
MediaSourceEventListener) to MediaSource.addEventListener: It will
receive DRM events too, even though you never passed it to
MediaSource.addDrmEventListener.
Also add some tests for MediaSourceEventDispatcher.
PiperOrigin-RevId: 301169915
This prevents OOM errors for high bitrate streams that attempt to fill
the buffer regardless of the memory usage.
Also change the max buffer sizes to ensure this is a no-op for video
streams < 20Mbps and audio streams < 260kbps.
Issue:#6647
Issue:#7006
PiperOrigin-RevId: 300720299
After this change users of ExoPlayerImpl and SimpleExoPlayer can use the media item base playlist API which converts the media item to a media source.
It adds the media item based methods to the ExoPlayer instead of the Player interface. This avoids a big change in the CastPlayer which requires migrating the cast extension to the MediaItem of the core module (follow up CLs).
PiperOrigin-RevId: 300575567
This CL removes the prefixes to the tests added after a6d0caaa3c.
This CL is generated by following command
$ find -name '*Test.java' | xargs -I{} sed -i 's/^\ \ public\ void\ test\([A-Z]\)\(.*\)$/ public void \L\1\E\2/' {}
PiperOrigin-RevId: 300547504
This avoids duplicate events being dispatched to object foo if
addListener(foo) is called more than once.
Part of issue:#6765
PiperOrigin-RevId: 300529733
We currently have the following logic to update renderers during
period transitions:
1. Wait for the currently reading period to finish reading all its
streams.
a. Advance reading period.
b. Set all streams that can't be replaced to final.
c. If streams can be replaced, replace them now.
2. Wait until playback position reaches the transition point
a. Disable all unneeded renderers (or those that need
re-enabling).
b. Advance playing period.
c. Enable all new renderers (i.e. all except the ones where
we replaced streams directly in step 1c.
This logic causes delays because steps 2a and 2c can easily happen
before 2b. Doing this allows a smooth transition for cases where
renderers change or where they need to be re-enabled.
The new order after this change is:
1. Wait for currently reading period to finish reading.
a. Advance reading period.
b. Set all streams that can't be replaced to final.
2. Update reading renderers iteratively.
a. If streams can be replaced, replace them asap.
b. If renderes need to be disabled, do so as soon as the
respective renderer ended.
c. Once step b is fully finished, enable or re-enable all new
renderers.
3. Wait unril playback position reaches the transition point AND
all tasks in step 2 are done (i.e. all renderers are set up for the
playing period).
a. Advance playing period.
As a nice side effect, decoder enabled and disabled events are now
always reported for the reading period, which is more consistent with
other renderer callbacks.
PiperOrigin-RevId: 300526983
This list was meant to simplify some usages where we only want to
make operations on enabled renderers. However, keeping this list
up-to-date is slightly error-prone (because renderers aren't added
and removed from this list next to the corresponding enable and disable
calls) and it makes it hard to do more fine-grained changes that only
enable or disabled a single renderer at a time.
PiperOrigin-RevId: 300513788
After this change MediaClocks and ExoPlayerImplInternal don't depend on the deprecated PlaybackParameter anymore but only know about playbackSpeed.
PiperOrigin-RevId: 300428791
This change deprecates the PlaybackParameters and remove the skipSilenceField from the PlaybackParameters. This implies that enabling and disabling skipping silences needs to be done on the Player.AudioComponent after this change.
After submission of the change, all Player API changes are done which are required to bring playbackSpeed and skipSilenceEnabled in the converged Player API state.
PiperOrigin-RevId: 300420843
In the HLS Spec (https://tools.ietf.org/html/draft-pantos-hls-rfc8216bis-04#section-4.4.3.6), specifically this or condition of this statement:
"... defined by the EXT-X-MAP tag, or is located between the start of the resource
and the offset of the first I-frame segment in that resource."
Change adds code to add this "implicit" Media Initialization Segment if no EXT-X-MAP defines one explicitly.
Pull in change to default to not use i-Frames for base AdaptiveTrackSelection.
The expectation is a subclases, that support transitioning to iFrame, will select these tracks.
This was a change suggested by @ojw28 to allow adpative track selection to work with IFrame only `Variant` as any other adaptive seleted `Track` in the `TrackGroup`. This actually works quite well with the exceptions that:
1. IFrame only tracks do not contain the same Rendition Group references as other non-iframe only varaints (so track selection should disable these tracks even in other non-video renderers.
2. Adapting to iFrame tracks only based on bandwidth might not be so good (as often larger spacial resolution iFrame track might have higher bitrate then lower spacial resolution non-iframe only track).
This change includes one proposed workaround in the AdaptiveTrackSelection class (only use iFrame for higher playback speeds).
Backout making the original `FormatAdjustingSampleQueue` an outer class and combining the new timestamp checking logic. This way the diff from `HlsSampleStreamWrapper` to dev-v2 are easier to see.
Update with code review suggestions:
1. rename to HlsCheckedSampleQueue to HlsSampleQueue and combine with format adjusting class.
2. Copywrite in added classes
3. Capture additional items useful for recovery and reporting in the exception class
4. Remove extraneous logging
5. eliminate magic number (50 seconds) and use percentage of duration
WIP:
- eliminate null check for chunk (chunkless prepare starts load before sampleQueue are created)
- work out recovery strategy
Add a SampleQueue subclass that checks the timestamp range of media samples queued to it and reports an exception on load if the timestamp is outside of spec bounds.
(Smashed to a single commit prior to rebase)
*** Original commit ***
Rollback of 949bbcfb2e
*** Original commit ***
Add masking for playWhenReady.
Masking is needed as soon as updates to a value can happen both in EPI
and EPII. PlayWhenReady is currently not masked because all updates
happen in EPI only. As soon as we allow pausing at certain times
(e.g. end of a stream), playWhenReady c...
***
PiperOrigin-RevId: 300330307
*** Original commit ***
Add masking for playWhenReady.
Masking is needed as soon as updates to a value can happen both in EPI
and EPII. PlayWhenReady is currently not masked because all updates
happen in EPI only. As soon as we allow pausing at certain times
(e.g. end of a stream), playWhenReady changes may be triggered by EPII
as well and that's why we need masking.
To know when the value actually changed, we also need to update the
internal state to include whether playback is supppressed.
***
PiperOrigin-RevId: 300303307
Masking is needed as soon as updates to a value can happen both in EPI
and EPII. PlayWhenReady is currently not masked because all updates
happen in EPI only. As soon as we allow pausing at certain times
(e.g. end of a stream), playWhenReady changes may be triggered by EPII
as well and that's why we need masking.
To know when the value actually changed, we also need to update the
internal state to include whether playback is supppressed.
PiperOrigin-RevId: 300284613
Exceeding the period duration may mean that that playback transitions
to another item even if the player is currently paused.
PiperOrigin-RevId: 300133655
There is no need for that, updating them is often forgotten and
takes up too much line space. Replace by a single TAG per test class.
PiperOrigin-RevId: 300113072
This change uses mime types in a functionally equivalent way to how we used the extension hint so far.
Using a mimeType instead of the extension has some advantages. Most importantly mimeTypes are used by the cast SDK with which we want to achieve interoperability in the cast extension.
Using a mimeType instead of the extension hint further appears to be a bit more clear (which might be opinionated). Further mime types are a well known and widely used concept to identify file type on the internet and it provides asterix based generalizations (audio/*, */*) which could express the media type OTHER that ExoPlayer is using internally (no usage of asterix required so far though).
PiperOrigin-RevId: 300058945
handlePrepare/Stop/SetPlayReady can be merged together as they all
handle changes to the desires state of the player.
Also, simplify parts of the control flow by not mixing code that
determines if audio focus needs to be handled with code that actually
acquires or abandons the focus.
PiperOrigin-RevId: 299824857
Without this option it's impossible to merge periods covering
different timestamps (at least not without playback issues).
Issue:issue:#6103
PiperOrigin-RevId: 299817540
The new version fixes some warnings in Gradle builds. Also
add missing indirect compileOnly dependencies to fix some more warnings
Issue:issue:#7007
PiperOrigin-RevId: 298855510
This option marks streams as final such that renderers play them out
completely, then waits until this happened, and then sets the player
to paused. After that, the player can continue to read the next period
to quickly resume playback if needed.
PiperOrigin-RevId: 298824745
Update it to allow any listener class to be registered (and thus
require the caller of dispatch() to provide the type of listener
to call).
Maintain MediaSourceEventListener.EventDispatcher as a sub-class
for now so that all existing references continue to work. This
avoids creating a huge diff in one CL. The intention is to in-line
these incrementally.
This is pre-work for issue:#6765
PiperOrigin-RevId: 298818198
The default Robolectric Looper simulation mode is changing to PAUSED from LEGACY.
The following tests fail in this new mode, and are thus being defaulted to LEGACY.
For more details see go/robolectric-legacy-looper-mode-lsc
Cleanup change automatically generated by error-prone refactoring
//third_party/android/androidx_test/tools/errorprone/java/androidx/test/tools/errorprone:LegacyLooperModeConverter_refactoring
Tested:
TAP train for global presubmit queue
http://test/OCL:297627974:BASE:298600828:1583273401491:7d94dbaa
PiperOrigin-RevId: 298809656
DashMediaSource is created with Manifest data and streamKeys can be empty given that streamKeys are only created out of MediaItem inside DashMediaSource, and this particular implementation don't extract them out of MediaItem properties.
Tested=Verified playback works.
PiperOrigin-RevId: 298550021
This change adds the createMediaSource(MediaItem mediaItem) method to the MediaSourceFactory interface. It doesn't deprecate createMediaSource(Uri uri) to keep the cl smaller. Deprecation and removing calls to the deprecated method from within the library and extension follow in a separate CL.
PiperOrigin-RevId: 298352442
The test had two problems:
1. It posts messages using a Handler and we need to idle the main looper
to actually deliver this message.
2. SimpleDecoder uses a background thread that is not within our control
from the test. Ensure the decoding happens after we queue input buffers
by using a lock.
PiperOrigin-RevId: 298300175
This is one step toward following the google3's test naming convention.
See go/java-testing/getting_started#basic-test-template for details
why prefix test isn't necessary.
This CL is generated by following command
$ find -name '*Test.java' | xargs -I{} sed -i 's/^\ \ public\ void\ test\([A-Z]\)\(.*\)$/ public void \L\1\E\2/' {}
and then manually modified following tests where test method names conflict with test target.
- VorbisUtilTest
- VorbisReaderTest
- UtilTest
- DownloadManagerDashTest
- DefaultOggSeekerTest
- OggPageHeaderTest
- HlsMasterPlaylistParserTest
PiperOrigin-RevId: 298074653
We should only ignore seek to the current position if we are
currently READY or BUFFERING. Also, pending initial seek positions
should only be saved while we have an empty timeline, independent of
the player state.
Issue:#6886
PiperOrigin-RevId: 297813854
This is more accurate since it's just a placeholder and none of the
values is provided by the media.
It also allows to fix a problem in ClippingMediaSource where we
couldn't detect a clipping error because we didn't know if the
timeline is a placeholder or not.
Issue:#5924
PiperOrigin-RevId: 297813606
When a new Timeline arrives in the Player, we check whether we can keep
existing MediaPeriods. This check currently involves a condition that
checks if the MediaPeriod is already prepared. The only reason we do
that is to avoid calling MediaPeriod.seekToUs, which is not allowed
on an unprepared MediaPeriod.
It's better to keep the MediaPeriod to prevent restarting the
preparation process. The prepration check can move further down to the
place right before we would call seekToUs.
PiperOrigin-RevId: 297812584
AnalyticsCollector keeps a list of existing MediaPeriodInfo that need
to be updated to new Timelines when they arrive. This already
happens in all cases except that the playingMediaPeriod wasn't updated
when it didn't change during the timeline update.
PiperOrigin-RevId: 297812038
The positions were interchangeably used with window and period
positions. This change more clearly ensures that all positions in the
AdPlaybackState are based on periods and that we use the right adjustments
for all usages.
PiperOrigin-RevId: 297811633
This removes noisy data from the dump files and allows updating less
dump files if a format field is added, removed or updated.
PiperOrigin-RevId: 297617138
Parallel asynchronous calls to MediaCodec.queueSecureInputBuffer() may
produce garbled video on some platforms. This workaround synchronizes
calls to MediaCodec.queueSecureInputBuffer() so that only one call is
in flight.
PiperOrigin-RevId: 297601037
This change adds the callback onSkipSilenceEnabledChange to the AudioListener and calls it when changed by the user by calling setPlaybackParameters, or when changed internally by the DefaultAudioSink if the parameters are not applicable. It needs to be plumped through AudioSink, AudioRenderer to SEP which eventually calls the AudioListener.
No changes to the Player interface so far. The getter of skipSilenceEnabled is added to SimpleExoPlayer for completeness, but not yet to the Player interface. The setter is not yet exposed, but implemented as a private method for implementation reasons.
PiperOrigin-RevId: 297590291
DefaultLoadControl applies the same min buffer duration to audio
and video. By default, min buffer is set equal to max buffer (50 seconds).
PiperOrigin-RevId: 297324489
Note: The dump files will need updating again when the extractors
are modified to only set the appropriate bitrate. Enhancing the
test first is nice, because it means that in subsequent CLs the
dump file updates can be used to quickly see what's changed in
the output.
PiperOrigin-RevId: 297188367
- Avoid having ExtractorHolder expose the underlying extractor.
- Make ProgressiveMP inject a DataSource instead of a DefaultExtractor.
This CL should introduce no functional changes.
PiperOrigin-RevId: 296944788
This change deprecates PlaybackParameter in AudioSink and splits it into playbackSpeed and skipSilenceEnabled. These properties are set separately in a future CL. The playback speed will be set through the MediaClock, while skipSilenceEnabled will be set by sending a message to the audio renderer.
PiperOrigin-RevId: 296457043
- Deprecate old Format.createXXX methods
- Deprecate most Format.copyXXX methods
- Stop using deprecated Format.copyXXX methods in the library
Note: Replacing library usages of Format.createXXX method
will be done in follow up CLs. These changes aren't purely
mechanical because we need to decide which out of peakBitrate
and averageBitrate to set in each case where currently a
single bitrate is provided.
Issue: #2863
PiperOrigin-RevId: 296450935
It's not clear why we're currently outputting the format in both init()
and consume() - it seems likely that this was accidentally introduced
in <unknown commit>
when we started outputting the format in consume() but didn't remove it
from init().
The SCTE-35 TsExtractorTest doesn't pass with the current code because
when it seeks back to the beginning of the file init() is not called
again so the second pass through only sees one sample (and
ExtractorAsserts checks that seeking back to 0 produces the same output
as reading the file from the beginning).
I generated the SCTE-35 sample using TSDuck's tsp command:
$ tsp --add-input-stuffing 1/10 \
-I file testdata/src/test/assets/ts/sample.ts \
-P pmt --service 1 --add-programinfo-id 0x43554549 --add-pid 600/0x86 \
-P spliceinject --service 1 --files scte_35.xml --wait-first-batch \
-P filter --negate --pid 0x1FFF \
-O file testdata/src/test/assets/ts/sample_scte35.ts
This was adapted from the instructions in section 5.2.24.1 of the TSDuck
user manual: https://tsduck.io/download/docs/tsduck.pdf
PiperOrigin-RevId: 296217084
Not important, but when overriding a method that can return null,
it seems preferable to put @Override first, followed by what it's
overriding (which includes the @Nullable).
Also remove explicit @NonNull use in the core library. @NonNull is
propagated by default, so this is redundant.
PiperOrigin-RevId: 296188379
This is one step toward following the google3's test naming convention.
See go/java-testing/getting_started#basic-test-template for details
why prefix test isn't necessary.
This CL is generated by following command
$ find -name '*Test.java' | xargs -I{} sed -i 's/^\ \ public\ void\ test\([A-Z]\)\(.*\)$/ public void \L\1\E\2/' {}
PiperOrigin-RevId: 296169886
When copying CryptoInfo for asynchronous input buffer queueing,
re-use the destination CryptoInfo's arrays to avoid re-allocating
memory.
PiperOrigin-RevId: 295706771
Note on UnknownNull: Where there exists a generically typed
base class and both null and non-null types are permitted,
we need to clear the default non-null that's otherwise
propagated everywhere. This then lets the nullness of the
type work properly.
PiperOrigin-RevId: 295582444
CryptoInfo.iv length is always 16. When the actual initialization vector
is shorter, zero out the trailing bytes.
Issue: #6982
PiperOrigin-RevId: 295575845
Previously, the input buffer accessUnit count
was passed as the output of the audio processors
which did not make sense as the processors can
split buffers.
This patch passes the access unit count through a
member variable.
PiperOrigin-RevId: 295572577
This is closer to how our actual renderers look like and allows tests
to use a similar logic to show/suppress the first frame in follow-up changes.
PiperOrigin-RevId: 295557548
We will eventually remove the workaround from ExoPlayerImplInternal
added in
b84bde0252
and replace it by an error that gets thrown if the LoadControl behaves
badly.
PiperOrigin-RevId: 295556131
Previously that was implicit, it was the passthrough dance:
- MimeTypes.getEncoding(inputFormat) did not
implement RAW MIME, so it was returning INVALID_ENCODING for PCM
- and allowPassthrough was returning
MimeTypes.getEncoding() != INVALID_ENCODING
- which was setting codecInfo.passthrough, and passthroughEnabled.
Thus Util.isEncodingPcm(encoding) is the opposite of
(MimeTypes.getEncoding(inputFormat) != C.ENCODING_INVALID)
This was rather implicit and brittle as anyone
could add support for audio/RAW in
MimeTypes.getEncoding and break PCM playback.
As a result make it explicit that allowPassthrough
must always return false in PCM.
PiperOrigin-RevId: 295144013
Modify TrackOutput.sampleData() to accept SampleDataReader instead of ExtractorInput. SampleDataReader supports only read and skip calls, which all sampleData() implementations already restrict themselves to.
PiperOrigin-RevId: 294905155
Currently only one access unit can be written per
buffer write. This has been found to be power
inefficient in an offload situation.
#exo-offload
PiperOrigin-RevId: 294886188
This information isn't easily available to the player at the moment (or it
would need to revaluate this every time), so add it to MediaPeriodInfo similar
to the existing isLastInTimelinePeriod.
The player needs to know whether a MediaPeriod is the last in its Timeline
window if we want to add an option to pause at the end of a window.
PiperOrigin-RevId: 294877628
Add fields in DecoderCounters for computing the average video frame
processing offset.
The MediaCodecVideoRenderer reports the video frame processing offset
and the demo app presents it on the debug information.
PiperOrigin-RevId: 294677878
The transition in EPII happens based on the duration of the periods, not
the windows. So use this duration in case they are not the same.
Also make sure to send the message at duration-1 because the playback position
at the duration is technically already the first position in the next item.
PiperOrigin-RevId: 294669335
This simplifies sending messages to the end of a stream. This was
already possible, but users needed to wait until the duration is known
before sending the message. This duration resolution can happen as part of the message position resolution.
PiperOrigin-RevId: 294626984
Issue #6853
Previously the body of an HTTP response was not included in an
InvalidResponseCodeException. This change adds a field to store
the response body, and updates DefaultHttpDataSource to populate
this value.
Language is already applicable to image formats as well.
It can also apply to video formats. For example when the
video contains burnt in subtitles. Also, some animation
studios produce multiple variants of animations, in which
text within the scene itself (e.g., a signpost) is
replaced to be language specific.
PiperOrigin-RevId: 294500227
- Replace tests specific to HTTP request headers with tests that
more generally test constructor correctness.
- Add test specifically around the constructor that infers the
HTTP method from postBody, since that's a non-trivial bit of
constructor logic.
Note: In general, I'm not sure this type of test (i.e., testing a
bit of trivial copying through a constructor) is worth the cost
of writing and maintaining them. That said, checking the new
uriPositionOffset is calculated correctly is non-trivial, as is
the HTTP method inference, so keeping them for now.
PiperOrigin-RevId: 294488684
This is a preliminary step toward adding a DataSpec.Builder,
which is needed for sanity when adding DataSpec.customData.
The existing absoluteStreamPosition field is very error prone,
because anyone using a Builder to adjust the request position
will need to remember to adjust two values:
dataSpec
.buildUpon()
.setAbsoluteStreamPosition(x)
.setPosition(x)
.build();
Furthermore, the difference between position and
absoluteStreamPosition is irrelevant in nearly all cases. In
the core library, the difference is only relevant when initializing
AES encryption/decryption to write/read cache files.
Replacing absoluteStreamPosition with uriPositionOffset will
simplify the code block above to:
dataSpec
.buildUpon()
.setPosition(x)
.build();
PiperOrigin-RevId: 294485644
Currently, the subset of audio tracks for adaptation is selected
purely based on the size of the subset in the track group, completely
ignoring the previously selected best individual track.
This change ignores all tracks with a different configuration than the
previously selected best track.
PiperOrigin-RevId: 294231806
While the window in which a message or a seek position is resolved is
still a placeholder, we need to re-resolve the position using the
user intent (i.e. the window position) when the timeline updates.
PiperOrigin-RevId: 293346360
The period in this method is most likely prepopulated correctly, but
this assumption is very error-prone and we should populate it locally.
PiperOrigin-RevId: 293345873
This allows to simulate samples in a stream more accurately
particularly when streams are prepared at a non-zero position and
issuing a sample with position=0 is not expected.
Also makes seek more realistic by also issuing one sample again.
PiperOrigin-RevId: 293344081
Once we receive an update from a masked source, we first start the
preparation of an already pending period, and only then notify the
player of the new timeline. If the period prepares immediately inline,
the MediaPeriod.onPrepared callback arrives before the
onPlaylistUpdateRequested call in the Player. THis is the wrong order
and causes issues when the player tries to lookup information in the
timeline that doesn't exist yet.
This change fixes preroll playbacks before live streams.
PiperOrigin-RevId: 293340031
The functional change is to set the bitrate fields to Format.NO_VALUE
in the case that they're non-positive in the header data. Else it's
very easy to to take the fields and copy them directly into Format as
incorrect values.
Issue #2863
PiperOrigin-RevId: 292920141
This change deprecates Player.onPlayerStateChanged(boolean pwr, int state). It removes deprecation for trivial cases. I'll remove other deprecated usages (mostly in ui module) in follow-up CLs to not bloat this CL.
PiperOrigin-RevId: 292917872
This condition is trying to detect when it might be necessary
to switch from a non-secure to a secure codec. This is not
possible if the DRM session is unchanged, unless a different
codec is required for some other reason (e.g., H264 -> H265),
which is anyway handled by canKeepCodec below.
PiperOrigin-RevId: 292909126
*** Original commit ***
Rollback of 3c56b113e4
*** Original commit ***
Rollback of d48dc4c159
*** Original commit ***
Move getting-stuck-prevention into DefaultLoadControl.
We recently added code that prevents getting stuck if the buffer is low and
the LoadControl refuses to continue loading (b84bde0252).
Move this logic...
***
PiperOrigin-RevId: 292457964
Passing EXT-X-KEY DrmInitData through the FragmentedMp4Extractor
doesn't work for streams with key rotation, because an extractor
instance is used for multiple segments, but is only passed the
EXT-X-KEY DrmInitData corresponding to the first segment.
This change removes passing DrmInitData through the extractor,
and instead passes it via FormatAdjustingSampleQueue. This is
in-line with how manifest DrmInitData is handled during DASH
playbacks.
Issue: #6903
PiperOrigin-RevId: 292323429
This is a nice-regardless improvement to SampleQueue, which will
likely to used to fix the referenced issue. It makes it possible
for SampleQueue subclasses to support dynamic changes to format
adjustment in a non-hacky way.
Issue: #6903
PiperOrigin-RevId: 292314720
Fix outline style subtitle
When a ForegroundColorSpan changes the foreground color, it is also applied
to the outline painter. In order to keep the correct color, one needs to
filter out theses span. We do this with a new cue that is our text source
for the specific edgeLayout.
Take care to only apply background color once
Test: Manual check - Subtitle view can show custom color subtitles from specific Subtitle
Renderer and outline is shown correctly using user defined color.
Calling setViewType with the same view type as the
subtitleView is using would throw InvalidArgumentException
instead of being a noop.
PiperOrigin-RevId: 291937202
We currently have 3 states, but the NOT_SET state isn't needed
anymore. We can therefore replace the IntDef by a simple boolean.
PiperOrigin-RevId: 291926049
We currently apply new parameter checkpoints from an absolute media
time and then substract the current media start time again to retrieve
the media time offset for this playback parameter checkpoint.
However, the media start time may change when unexpected discontinuities
happen (the start time doesn't actually change, but we change it to
correct for this discontinuity). This then invalidates the absolute
media time in the playback parameter checkpoints (which should have been
corrected as well).
Avoid this problem by also only applying the new start position
from the checkpoint. We don't have to save the start position anymore
because it will cancel itself out.
Also add some documentation and code clarification for improved
readability.
PiperOrigin-RevId: 291923069
SubtitleView now becomes a ViewGroup that owns a SubtitleTextView. It
handles some common styling defaults, but delegates the underlying
values down into SubtitleTextView through the SubtitleView.Output
interface.
When I add a SubtitleWebView this will also be a ViewGroup containing
a WebView and will implement SubtitleView.Output and convert Cue styling
into HTML & CSS.
PiperOrigin-RevId: 291903294
It doesn't seem worth keeping the cap, since the device
will presumably stop receiving major version updates at
some point anyway.
Issue: #6899
PiperOrigin-RevId: 291899439
This seeker is only seeking to frames that have already been read by the
extractor for playback. Seeking to not-yet-read frames will be
implemented in another change.
Issue: #6787
PiperOrigin-RevId: 291888899
We currently only keep the requested next content start position while
we are playing ads. However, we should also keep at least before a content
period is fully prepared to not loose the information about the user intent.
PiperOrigin-RevId: 291705752
There are existing bugs that need this flag to know whether the
current information in the window is still a placeholder or can
already be relied on for further calculation.
This flag will probably only ever be set in DummyTimeline, so it's
not added to the window.set method to avoid updating all clients.
Issue:#5924
PiperOrigin-RevId: 291705637
It could either live in .util or .extractor, but since it's not needed
outside the extractor package in core (and the FLAC extension), and
FlacStreamMetadataTest uses a FLAC asset, it seems preferable to move it
into the extractor package.
PiperOrigin-RevId: 291372032
CeaUtil depends on TrackOutput so should live in the extractors package.
To avoid having Cea708Decoder depend on extractors, this change also
moves the initialization data building/parsing to CodecSpecificDataUtil.
PiperOrigin-RevId: 291348317
Using @Test(expected=...) is discouraged, since the test will pass if
any statement in the test method throws the expected exception.
PiperOrigin-RevId: 291338224
Since we have the Playlist layer on top, it's always guaranteed that a
new playlist item has a new uid. So we can just keep the old one in all
cases and don't have to be careful to delete it. The deletion was necessary
previously in case multiple MediaSources use the same uids.
PiperOrigin-RevId: 291152349
Move Cea708InitializationData functionality into CeaUtil and delete the
class.
Move CeaUtil to the util package as it needs to be used both in
extraction and decoding.
PiperOrigin-RevId: 291128220
The only remaining use is to provide the initial position for the first media
period in the queue. At this point, the value is always equivalent to
PlaybackInfo.positionUs because they are set together to the same value in
PlaybackInfo.copyWithNewPosition.
PiperOrigin-RevId: 290749294
We have a flag that gets turned on if the current playback position has been
reset, so that messages from this reset position can be triggered. This
mechanism isn't actually depending on the startPositionUs field because it
should always be triggered in the first iteration with the new position.
PiperOrigin-RevId: 290749201
There are a couple of problems with how positions in PlaybackInfo are set at
the moment:
1. PositionUs isn't allowed to be C.TIME_UNSET. This is prevented by always
resolving to the current default position if needed.
2. In some places a window position was used as a period position. Use correct
position resolution procedure.
3. When creating a placeholder position to restart playback, we used the first
period in a window, not the one where the default position is.
4. The start position for ads was in some cases set to 0 without checking
the ad resume position.
PiperOrigin-RevId: 290749042
I needed to use Cue.Builder instead of just SpannableStringBuilder for
the regionOutput values, so I could attach the vertical info where
appropriate (since this is a property of the Cue, not a span).
PiperOrigin-RevId: 290709294
If an exception is thrown between updating the timeline and updating
the position in playbackInfo, the state may be inconsistent.
Exceptions are expected to be thrown while updating the player state
and we should handle such cases in a consistent way.
Similar to how we handle the same situation in seekToInternal,
the state is updated in a final block such that it gets updated to the
latest state even if an error occurs. Moving both the timeline and
position update together also ensures they always stay consistent.
PiperOrigin-RevId: 290624020
This restructure moves all the position resolving code to a static method and
removes the dependency of the MediaPeriodQueue on having an up-to-date timeline.
Both steps allow simplified reasoning about the code as the static method can't change
the state of the player, and there is no risk the queue can use the wrong timeline.
These propoerties allow to fix a bug causing inconsistent states in a follow-up step.
PiperOrigin-RevId: 290616395
It's never assigned or accessed from outside the class.
It was added in
<unknown commit>
then the accessor was removed in
<unknown commit>
PiperOrigin-RevId: 290601998
This makes MPEG audio utilities similar to utilities we have for WAV,
AC-3 etc., and moves them out of the extractor package so that an
extractor module can be split out without needing to have a class in the
extractor package in the common library.
PiperOrigin-RevId: 290595089
*** Original commit ***
Rollback of ff89170b00
*** Original commit ***
Fix some logic in AnalyticsCollector.
All events issued from ExoPlayerImpl (i.e. Player.EventListener events) currently
try to use the media period data from the playing media period as set in the
playback thread queue. This is only correct as long as there no pending masking
operations in ExoPlayerImpl. That's why we currently disable this whenever the timeline
is...
***
PiperOrigin-RevId: 290593700
*** Original commit ***
Fix some logic in AnalyticsCollector.
All events issued from ExoPlayerImpl (i.e. Player.EventListener events) currently
try to use the media period data from the playing media period as set in the
playback thread queue. This is only correct as long as there no pending masking
operations in ExoPlayerImpl. That's why we currently disable this whenever the timeline
is empty or a seek is pending. Since adding all the playlist API methods to the player,
this is no longer the right choice. Moreover,...
***
PiperOrigin-RevId: 290312118
Currently both are updated by separate setters. If an exception is thrown between
the two setters, the state may not be consistent.
Avoid this problem by always setting them together.
PiperOrigin-RevId: 290293600
All events issued from ExoPlayerImpl (i.e. Player.EventListener events) currently
try to use the media period data from the playing media period as set in the
playback thread queue. This is only correct as long as there no pending masking
operations in ExoPlayerImpl. That's why we currently disable this whenever the timeline
is empty or a seek is pending. Since adding all the playlist API methods to the player,
this is no longer the right choice. Moreover, we don't have a definite API that tells
AnalyticsCollector when a playlist API call has been handled (and we don't want to
have one).
We can fix this by always using the current Player position information as the source
of truth (instead of the media period queue). This is definitely more correct and also
works while a masking operation is pending. To fill in the additional information from
the media period queue, we can look up a matching media period. This may not be the
first one in the list if an operation is pending.
The new methods are similar to the previous tryResolveWindowIndex method, but:
1. They are always used (i.e. the current Player state is the main source of truth)
2. They also check the correct ad playback state, that was just ignored previously.
PiperOrigin-RevId: 290284916
Currently seeks are basically ignored. However, it's more realistic to re-queue the
single sample if the seek is to position 0.
PiperOrigin-RevId: 290273564
*** Original commit ***
Rollback of d48dc4c159
*** Original commit ***
Move getting-stuck-prevention into DefaultLoadControl.
We recently added code that prevents getting stuck if the buffer is low and
the LoadControl refuses to continue loading (b84bde0252).
Move this logic into DefaultLoadControl to keep the workaround, and also apply the
maximum buffer size check in bytes if enabled. ExoPlayerImplInternal will now
throw if a...
***
PiperOrigin-RevId: 290273295
- Add additional Listener methods to DownloadManager, to inform of
changes to whether the downloads are paused or waiting for requirements.
- Only schedule the Scheduler if we really are waiting for requirements.
- Only restart the service if we're no longer waiting for requirements,
and if there are queued downloads that will now be restarted.
Previously the service would be restarted whenever the requirements
were met, regardless of whether there was any work to do.
- Restart service if it might be stopping, as well as if it's already
stopped. Also restart service if there's a download state change to a
state for which the service should be started, if.
Issue: #6798
PiperOrigin-RevId: 290270547
First cl towards DefaultMediaSourceFactory (<unknown commit>) does not change the MediaSourceFactory interface except adding @Nullable anotations to setters.
PiperOrigin-RevId: 290269640
Doing so prevent Codec which is a big object with
JNI to be garbage collected.
As the codec can not be hold in the listener, there
is no way to call release, as it must be called on
the same codec.
As a result the release method is also removed.
The downside is that at runtime some callbacks may be
dropped but it should be a short transitive state.
This also simplifies lifecycle of the listener as
the client does not have to know if release needs
to be called or not.
An alternative would have been to hold a weak ref,
but I deemed it too complicated for the
runtime gain.
PiperOrigin-RevId: 290231659
This fixes an issue where a DownloadService implementation
that allows foreground but doesn't provide a scheduler would
not be restarted in the case that it was still in memory but
classed as idle by the platform.
It also speeds up service restart in the case that a
scheduler is provided.
Issue: #6798
PiperOrigin-RevId: 290068960
This method has two use cases:
1. Seeking. Calls are immediately preceded by a call to rewind(), and
the returned value isn't important unless it's ADVANCED_FAILED (i.e.
the caller is only interested in success and failure).
2. Advancing. The return value is important unless it's ADVANCED_FAILED,
in which case the caller wants to treat it as 0.
This change creates separate methods for each use case. The new seekTo
methods automatically rewind and return a boolean. The updated advanceTo
method returns 0 directly in cases where ADVANCED_FAILED was returned.
Arguments that were always hard-coded to true by callers have also been
removed.
This change is a step toward one possible solution for #6155. How we'll
solve that issue is still up for discussion, but this change seems like
one we should make regardless!
Issue: #6155
PiperOrigin-RevId: 290053743
- DownloadManagerHelper now passes all downloads to the
DownloadService when the service is attached (and once
the downloads are known). The service then starts the
foreground notification updater if necessary. This fixes
the ref'd issue.
- Don't call getScheduler() if the service is background
only. This was already documented to be the case on the
DownloadService constructor.
- If the service is started in the foreground on SDK level
26 and higher, satisfy the condition to move the service
to the foreground in onStartCommand rather than in stop().
It's much more obviously correct, and should produce the
same end result.
Issue: #6798
PiperOrigin-RevId: 290050024
Currently the following sequence of events happens at automatic period transitions:
1. Update queue (=release old playing period)
2. Disable unused renderers
3. Enable newly needed renderers
This order requires difficult to follow workarounds in AnalyticsCollector for all
events related to step 2 (disable renderers). The current media period has already
been advanced so can't be used. The current workaround saves the last known playing
media period that was published as part of a PlaybackInfo update in ExoPlayerImpl.
This works in most cases, but is inherently wrong because the published state in
ExoPlayerImpl may be completely unrelated to the updates happening on the playback
thread (e.g. if many other operations are pending).
Simplify and fix this problem by changing the order of the events above:
1. Disable unused renderers.
2. Update queue
3. Enable newly needed renderers.
This way the current playing media period can be used for both renderer disable
and renderer enable events, thus it's correct in all cases and the workaround
in AnalyticsCollector can be removed.
PiperOrigin-RevId: 290037225
Whenever we set the maskingWindowIndex, we also need to set the masking period index
and the masking position. Otherwise it may cause exception when the period index
is used together with the masking timeline.
PiperOrigin-RevId: 290036973
This optimization allows a ChunkSampleStream to output track
formats as soon as they're parsed during an InitializationChunk
load, rather than waiting until after the InitializationChunk
load is completed.
In DASH VOD, a single InitializationChunk typically loads the
moov and sidx atoms in that order. Hence for long form content
where the sidx is a non-trivial size, this may result in the
track formats being output a non-negligible period of time
sooner than was previously the case. This allows downstream
renderers to start codec initialization sooner, potentially
decreasing startup latency.
For a single test stream on a fast & stable network, this
pretty consistently reduced elapsed time until both audio and
video codecs have been initialized from ~0.5s to ~0.3 seconds
on a Galaxy S8. For 5 test runs without and with these patches,
the eventTime logged by EventLogger for the second decoder
init were:
Without (secs): 0.47 0.47 0.45 0.48 0.46
With (secs) : 0.32 0.33 0.34 0.31 0.40
PiperOrigin-RevId: 289845089
The current order of operations means that the Format is only passed
to the chunk's output after the DataSource has been opened. This
means that establishing the network connection and the downstream
renderers initializing their codecs are effectively serialized to
occur one after the other.
In the new order, the Format is passed to the chunk's output before
the DataSource has been opened. This allows the downstream renderers
to initialize their codecs in parallel with the network connection
being established, and hence latency at the start of playback is
reduced.
PiperOrigin-RevId: 289841854
We currently have a currentMediaPeriodId and an activeSessionId that are more
or less tracking the same thing unless the current media period isn't "active" yet.
Simplify this logic by using a single currentSessionId field and the respective
isActive flag of this session. Also move all session creation and activation code in
the same method to make it easier to reason about the code.
This change also fixes a subtle bug where events after a seek to a new window
are not ignored as they should.
PiperOrigin-RevId: 289432181
As discovered whilst investigating #6798, there are cases
where these methods are not correctly. They were added as
convenience methods that could be overridden by concrete
DownloadService implementations, but since they don't work
properly it's preferable to require application code to
listen to their DownloadManager directly instead.
Notes:
- The original proposal to fix#6798 stored the state change
events in memory until they could be delivered. This approach
is not ideal because the events still end up being delivered
later than they should be. We also want to fix the root cause
in a different way that does not require doing this.
- This change does not fix#6798. It's a preparatory step.
Issue: #6798
PiperOrigin-RevId: 289418555
This change makes it clear the SampleQueue doesn't outlive
the wrapping PlayerTrackEmsgHandler, and releases it from
PlayerTrackEmsgHandler.release().
This change is a no-op because calling release() is the
same as reset() in this case (the behavior only differs if
a non-dummy DrmSessionManager is being used).
PiperOrigin-RevId: 289416622
As a result, onMediaChunkLoadStarted gets invoked on the loading thread, and
init on the playback thread, matching the thread access comments.
Issue:#6321
PiperOrigin-RevId: 289167981
Dash live streams require that the client has an accurate wall clock
time and in absence of a UTCTiming element, this is assumed to be the
NTP time.
This change adds NTP time offset resolution for DASH live streams
without such timing elements.
PiperOrigin-RevId: 289098796
Unfortunately devices such as the MI 8 do not provide an alternative
working decoder. Some Vivo devices do provide one though.
PiperOrigin-RevId: 288911897
This improves readability by making clearer that reading no bytes from
the input in readFrames() is an expected case if the buffer is full.
PiperOrigin-RevId: 288711841
Currently, seeks are only tracked if both the start and the end of the seek
is within the same window. This means no seeking state is reported if the
playback switches to a new window during the seek.
This problem is fixed by tracking seek start and end events in all cases,
but only report seeking state once the window becomes the foreground window.
PiperOrigin-RevId: 288706674
These make the interesting bits of each assertion harder to follow imo.
Also remove all the assertWithMessage() calls at the same time, Olly
convinced me these are rarely useful since you can click from the stack
trace to the failing line in the IDE.
PiperOrigin-RevId: 288470704
- Simulate IO exceptions in the test using FlacBinarySearchSeeker for
seeking in FlacExtractorTests. This makes the test slower but covers
more test cases.
PiperOrigin-RevId: 288285057
There's currently no rendering support for ruby text in SubtitleView
or SubtitlePainter, but this does have a visible impact with the
current implementation by stripping the ruby text from Cue.text
meaning it doesn't show up at all under the 'naive' rendering.
This is an improvement over the current behaviour of including
the ruby text in-line with the base text (no rubies is better than
wrongly rendered rubies).
PiperOrigin-RevId: 288280416
This typo was introduced in ddb70d96ad
when migrating a static method with parameter `durationUs` to an
instance method where the correct field to use was `blockDurationUs`
(but `durationUs` also exists).
The test that catches this was only added in 45013ece1e (and
therefore configured with the wrong expected output data).
issue:#6833
PiperOrigin-RevId: 288274197
It's not used. I was trying to work out how to correctly cascade my
text-combine-upright styling, but deleting the method seemed easier...
PiperOrigin-RevId: 287989480
I decided the flags bit was a bit unclear so I played around with this
It's also needed for more 'complex' assertions like colors - I didn't
want to just chuck in a fourth int parameter to create:
hasForegroundColorSpan(int start, int end, int flags, int color)
PiperOrigin-RevId: 287989424
Relying on the precedence of spans seems risky - I can't find it
defined anywhere. It might have changed in Android 6.0?
https://stackoverflow.com/q/34631851
PiperOrigin-RevId: 287989365
-- Optimize Mp4Extractor for AC-4
-- Optimize FragmentedMp4Extractor for AC-4
-- Add test case for AC-4 in MP4
-- Add test case for AC-4 in Fragmented MP4
It's unreliable for at least one IMA ADPCM file I've found.
Calculating the blockIndex to seek to using exact properties
also seems more robust.
Note this doesn't change anything for the existing PCM test,
since averageBytesPerSecond is set correctly. It does make
a difference for an upcoming IMA ADPCM test though.
PiperOrigin-RevId: 287814947
An experiment with this algorithm didn't show positive results. We can
therefore keep the simpler default algorithm.
Startblock:
<unknown commit> is submitted
PiperOrigin-RevId: 287807538
Where media segments are specified using byte ranges, it makes
sense that a server might return 416 (which we don't consider
for blacklisting) if the segment is unavailable, rather than
the 404 (which we do consider for blacklisting) that we expect
when media segments are only specified using a URL.
Issue: #6775
PiperOrigin-RevId: 286620698
The Download constructor considers it invalid to have a failure
reason if the download isn't in the failed state. Unfortunately,
calling DefaultDownloadIndex.removeAllDownloads when there's a
failed download will change the state without clearing the reason.
If the downloads are then read back from the DefaultDownloadIndex
we end up violating the Download constructor assertion.
This change clears the failed reason for any existing rows in the
invalid state, and also fixes the root cause that allows invalid
rows to enter the table in the first place.
Issue: #6785
PiperOrigin-RevId: 286576242
Without this @Nullable, potential subclasses can't override the
method to return null if they don't want to use the renderer as a
media clock.
Issue:#6792
PiperOrigin-RevId: 286545736
This class is only used to hold temporary data while we parse the
settings and text, so we don't need it outside the Parser class.
Also remove all state from WebvttCueParser - this increases
the number of allocations, but there are already many
and subtitles generally aren't very frequent (compared to
e.g. video frames).
PiperOrigin-RevId: 286200002
Add support for MP3 as an encoding format for passthrough.
This change does not change the observable
behavior of Exoplayer.
Also name the magics.
#exo-offload
PiperOrigin-RevId: 286146539
*** Original commit ***
Move getting-stuck-prevention into DefaultLoadControl.
We recently added code that prevents getting stuck if the buffer is low and
the LoadControl refuses to continue loading (b84bde0252).
Move this logic into DefaultLoadControl to keep the workaround, and also apply the
maximum buffer size check in bytes if enabled. ExoPlayerImplInternal will now
throw if a LoadControl lets playback get stuck. This includes the case where
DefaultLoadControl reaches its maximum buffer size and not even a mim...
***
PiperOrigin-RevId: 286071115
We recently added code that prevents getting stuck if the buffer is low and
the LoadControl refuses to continue loading (b84bde0252).
Move this logic into DefaultLoadControl to keep the workaround, and also apply the
maximum buffer size check in bytes if enabled. ExoPlayerImplInternal will now
throw if a LoadControl lets playback get stuck. This includes the case where
DefaultLoadControl reaches its maximum buffer size and not even a mimimal buffer
duration.
PiperOrigin-RevId: 285979989
This offset allows to improve the calculated live offset because it
can take known client-server time offsets into account.
PiperOrigin-RevId: 285970738
I want to add fields related to vertical text support, and neither
adding another constructor nor updating all call-sites of existing
constructors seemed like attractive propositions.
PiperOrigin-RevId: 285956024
- Remove duplicated null assignments
- Move mediaCryptoRequiresSecureDecoder reset to be with all the
other mediaCrypto stuff.
PiperOrigin-RevId: 285955134
MultiLockAsyncMediaCodecAdapter is an implementation of the
MediaCodecAdapter that uses multiple locks to synchronize access to its
data compared to the single-lock approach used in
DedicatedThreadAsyncMediaCodecAdapter.
PiperOrigin-RevId: 285944702
- Make extractor output samples that are uniformly distributed
with respect to time, with a target of ~10 samples per second.
The old approach could in theory put every frame into its own
sample, which would be very inefficient downstream because we'd
need to pass them individually to MediaCodec. It could also put
data corresponding to a long duration of time into a single
sample (e.g. if the sample rate of the content is low), which
is bad downstream because we decide whether to set the decodeOnly
flag on a per sample basis. More generally, the new approach
is more predictable :).
- Stop using the WavSeekMap to get sample timestamps, and instead
calculate them directly from the number of frames output. It's
more obviously correct, particularly for data formats like IMA
ADPCM where we'll need to adjust the data prior to output.
PiperOrigin-RevId: 285750010
In MatroskaExtractor, if the last cue time exceeds the duration
specified in the header, then we end up generating a negative
duration chunk as the last item in the SeekMap. We should probably
not do this, so drop it instead.
Note: Matroska does have a CueDuration element, but it's not used
in the one problematic file I've found.
PiperOrigin-RevId: 285738418
Currently, resetting the internal state related to the codec in done in 4 places
in MediaCodecRenderer:
1. In the constructor to set some sensible initial default.
2. In flushOrReleaseCodec to reset state after flushing
3. In releaseCodec to reset state to default values for a released codec.
4. In initCodec to reset state to defaul values for a new codec.
There are actuall only two types of state reset operations:
1. Resetting state for a released codec, so that a new codec isn't influenced by
previous codecs.
2. Resetting state after flushing an existing codec. This is a subset of (1).
So to simplify the class, this change moves all state resets to two methods
corresponding to the two cases above.
PiperOrigin-RevId: 285731913
These are transitively provided by testutil's build.gradle.
This makes the core module consistent with the others like ui, hls, dash
PiperOrigin-RevId: 285729566
This may happen for example when trying to load unfinished live media
chunks that can only be loaded at real-time playback speed.
The flag can be used by the sources to indicate that network transfer
will knowingly be throttled, such that transfer listeners like the
bandwidth meter can take this information into account.
PiperOrigin-RevId: 285397100
Also change IcyInfo.rawMetatadata from String to byte[]
ICY doesn't specify the character encoding, and there are streams
not using UTF-8 (issue:#6753). It seems the default of at least one
server is ISO-8859-1 so let's support that as a fallback:
https://github.com/savonet/liquidsoap/issues/411#issuecomment-288759200
Also update IcyDecoder to skip strings it doesn't recognise at all
instead of decoding invalid characters.
The feed from issue:#6753 now decodes accents correctly:
EventLogger: ICY: title="D Pai - Le temps de la rentrée", url="null"
PiperOrigin-RevId: 285388522
We previously used Locale.fromLanguageTag and then toLanguageTag to automatically
filter out invalid tags and to replace deprecated tags. However, this only works
for API 21+.
This change does no longer rely on the platform methods and instead:
1. Keeps invalid tags as they are. This may help if developers rely on
non-spec-complaint language tags (for example, #6681).
2. Adds a list of deprcated tags and their modern replacements to our code
3. Normalizes some short codes that have been superseded by macrolanguage codes by
their macrolanguage equivalent.
Issue:#6681
PiperOrigin-RevId: 285384841
This has been broken since
c3d6be3afd
and broken for ICY (where I noticed the problem) since
5695bae9d8.
ICY symptom is that we see no repeated metadata, because the
Icy-MetaData:1 header doesn't make it to the server so we never get back
icy-metaint.
PiperOrigin-RevId: 285379234
The presentationTimeOffsetMs may be C.TIME_UNSET for VOD content
and shouldn't be used in calculations for the windowStartTime.
PiperOrigin-RevId: 285363095
Retrieve the sample number in the extractor instead of passing a holder
to FlacBinarySearchSeeker. This change makes the code easier to
understand and is required to implement the seeking from the seek table.
PiperOrigin-RevId: 285241862
- WavHeader is now immutable and contains only values parsed out
of the WAVE FMT chunk. It no longer contains a C.PcmEncoding
encoding, or mutable data bounds.
- WavHeaderReader now parses the WAVE header chunks without any
additional logic (e.g. validating the block alignment value,
which is format type dependent).
- The SeekMap part of WavHeader is split out into WavSeekMap.
PiperOrigin-RevId: 285232498
This returns the current offset to the live edge. The calculation is
non-intuitive enough to provide this convenience method.
PiperOrigin-RevId: 285171090
This avoids needing to jump through type paremeter hoops to create a
sub-classable Cue.Builder.
Cue should have a standard set of fields, because it will be
consumed by renderers that don't know what type it is. The
existing subclass fields are only used inside their respective
packages, so can be part of a wrapper object instead.
This lays the groundwork for converting Cue's multiple constructors
into a Builder pattern.
PiperOrigin-RevId: 284992135
Keyframe was rendered rather than skipped when performing
an exact seek to a non-zero position close to the start of
the stream.
PiperOrigin-RevId: 284798460
The nullness checker complains about Integers with @IntDef annotations
so replace pairs with custom pair types for the timed event records in
PlaybackStats.
PiperOrigin-RevId: 284731834
Caveats:
- Block additional data is ignored if the block is laced
and contains multiple samples. Note that this is not
a loss of functionality (SimpleBlock cannot have block
additional data, and lacing was previously completely
unsupported for Block)
- Subrip and ASS samples are dropped if they're in laced
blocks with multiple samples (I don't think this is
valid anyway)
Issue: #3026
PiperOrigin-RevId: 284545197
This change constrains the use of sample state member variables to
writeSampleData, finishWriteSampleData and resetWriteSampleData.
Using them elsewhere gets increasingly confusing when considering
features like lacing in full blocks. For example sampleBytesWritten
cannot be used when calling commitSampleToOutput in this case
because we need to write the sample data for multiple samples
before we commit any of them.
Issue: #3026
PiperOrigin-RevId: 284541942
Refator to lift a method to create (and destroy) the OnFrameRenderedListenerV23() and split destroying the Handler from removing the MediaCodec onFrameRenderedListener to avoid double call to codec.setOnFrameRenderedListener()
This makes the format metadata null (instead of an empty Metadata
object) when it is not provided, and is therefore consistent with the
other extractors behavior.
PiperOrigin-RevId: 284171148
The capture frame rate is currently available both via Format.metadata
and decoded in Format.frameRate. As the container Format.frameRate may
be useful to apps, only store the capture frame rate in metadata (apps
will need to decode it but can now access the container frame rate too).
PiperOrigin-RevId: 284165711
- Remove "lacing" from member variables. They're used even if
there is no lacing (and the fact that lacing is the way of
getting multiple samples into a block isn't important).
Issue: #3026
PiperOrigin-RevId: 284152447
- Change sampleHasReferenceBlock to a block reading variable, which is
what it is (the distinction didn't matter previously, but will do so
when we add lacing support in full blocks because there wont be a 1:1
relationship any more.
- Move sampleRead to be a reading state variable.
- Stop abbreviating "additional"
Issue: #3026
PiperOrigin-RevId: 284000937
An I/O error could occur while handling the start of an mdta box, in
which case retrying would cause another ContainerAtom to be added. Fix
this by skipping the mdta header before updating container atoms.
PiperOrigin-RevId: 284000715
Saves around 200 lines of code. High level overview:
- Rename SampleQueue to SampleDataQueue.
- Rename SampleMetadataQueue to SampleQueue.
This CL should not introduce behavior changes. The only significant
changes in synchronization should be:
+ Add synchronized keyword to isReady.
- Seems to be necessary.
+ Add synchronized keyword to SampleQueue.sampleMetadata.
- Before this change, SampleQueue.sampleMetadata could acquire the
SampleMetadataQueue lock three times in a single method call.
Other miscellaneous improvements:
+ Put all private methods at the bottom.
+ Move release() to the right category.
PiperOrigin-RevId: 283795844
This parameter is a little confusing, especially as the behaviour
can be surprising if the intended use-case isn't clear. This change
moves the description of the parameter into the class javadoc,
adds context/justification and slims down each method's
javadoc to refer to the class-level.
Related to investigating/fixing issue:#6700
PiperOrigin-RevId: 283724826
Finailzed logic to update the `SampleQueue` write positions (first index) to push these into `HlsMediaChunk` when the track is initially created (from the Extractor output) and as each new chunk is queued to load (`init()` callback).
Add lots of debuging prints that can come out for the final merge.
Code is very close to a clone of `ChunkSampleStream`.
DASH implements this feature, extend the feature for HLS as well. First change just drops video samples.
For demuxed audio the audio samples will continue to play out to match the
dropped video, so need to keep indexes in all the sample queues related to a chunk and discard them all.
This is a minor change ahead of merging a full variant of
https://github.com/google/ExoPlayer/pull/6706, to make
re-buffers less likely.
Also remove variable substitution when parsing
AVERAGE-BANDWIDTH (it's not required for integer attributes)
PiperOrigin-RevId: 283554106
Calls to new Handler() without arguments are deprecated as of the latest Android
version. Replace them by a Util.createHandler call similar to the ones we
already have.
PiperOrigin-RevId: 283532891
Fixes issue:#6700
sample_cbs_truncated.adts test file produced using
`$ split -b 31795 sample_truncated.adts` to remove the last 10 bytes
PiperOrigin-RevId: 283530136
If LoadControl.shouldContinueLoading returns false and the renderers are not
ready for playback using the already buffered data, playback is stuck.
To prevent this situation, we always continue loading if the buffer is almost
empty. We already have a similar workaround for when
LoadControl.shouldStartPlayback returns false even if loading stopped.
Having both workarounds allows playback to continue even if the LoadControl
tries to prevent loading and playing all the time.
PiperOrigin-RevId: 283516750
*** Original commit ***
Improve `Format` propagation within the `MediaCodecRenderer`.
For example, fix handling of pixel aspect ratio changes in
playlists where video resolution does not change.
Issue:#6646
***
PiperOrigin-RevId: 282903626
When transitioning to a new stream in a different format, the audio
processors are reconfigured. After this, they are drained and then
flushed so that they are ready to handle data in updated formats for the
new stream.
Before this change, some audio processors made the assumption that after
reconfiguration no more input would be queued in their old input format,
but this assumption is not correct: during draining more input may be
queued. Fix this behavior so that the new configuration is not referred
to while draining and only becomes active once flushed.
Issue: #6601
PiperOrigin-RevId: 282515359
flush() is guaranteed to be called in all these cases anyway.
Also clarify documentation for AudioProcessor-specific methods that can
change the 'active' flag.
Issue: #6601
PiperOrigin-RevId: 282515255
