Also do some naming cleanup + do ms->us seek conversion
in ExoPlayerImpl, since that's where we now do all the
conversions coming out of the player.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=125662973
ExoPlayer.EventListener.onPositionDiscontinuity is notified during seeking and
transitioning from one source to the next.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=125578976
This CL only provides checks for HEVC codecs and adds codec information to Format
instances in DASH. Right now, we check that the supported profiles are advertised
individually and that the supported level is equal or higher than the requested codec
level.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=125470489
- AllowedVideoJoiningTimeMs must be set to 0 for tests so
that tests which disable/enable video renderers don't
register a large number of dropped frames.
- Fixed a threading issue that could cause occassional test
failure.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=124978843
The ones that are commented out are legitimately broken
and will need a library fix.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=124978438
When buffering playlist items after the currently-playing one,
continueBuffering should take a position that is negative because the playback
position is before the start of the source being buffered. This change makes
sure that ExtractorSampleSources always have a known duration when they are
fully buffered, which means that the correct (negative) source-relative
position can be calculated.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=124949409
ExtractorSampleSource covers almost all use cases (except for AMR, AVI and
legacy Widevine). The FLAC extension must be compiled for support for FLAC
extraction.
FrameworkSampleSource has device-specific issues, makes blocking calls to the
underlying MediaExtractor, and does not provide much control over buffering.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=124850102
- Event listener is now at the SampleSource level, since the
individual ChunkTrackStream instances are created internally.
- Event listener receives events corresponding to loads made
at the SampleSource level (e.g. manifest fetches) in addition
to those made by the ChunkTrackStream instances.
- Added ability for FormatEvaluators to propagate arbitrary
information (as an Object) through to the listeners. This was
a request from YouTube.
- Added significantly more information to each event, for example
the DataSpec that defines the load being made is now passed, as
is timing information.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=124732984
DashTest will be migrated separately, since it's a little
more work.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=124722462
If the same instance is reset then rapidly re-enabling the
renderer after disabling it can prevent an application from
reading the final values at the time when the renderer was
disabled. Hence we now instantiate a new CodecCounters each
time. This is needed to migrate DashTest to V2.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=124721106
TextTrackRenderer would unconditionally flip input buffers read from its
source TrackStream, but the data could be null in the case where the buffer
signaled end of stream. Only flip if the corresponding flag is not set.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=124559077
Things like manifests are loaded not in the context of a Chunk,
but we want to eventually be reporting all loading events. So
it makes senes to define data types at a global level.
Also added {@code} in a few places for consistency.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=124549737
DASH and SS manifests define potentially useful information
that isn't present in the actual media streams. Primarily
the representation id, but bitrate, fps and language may
also be defined only at the manifest leve. This change merges
the information into the sample format that's propagated to
the renderers.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=124225175
- Parse APIC and TextInformation frames.
- In MPEG-TS, don't mind if packets contain end padding.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=124079930
On older devices getCapabilitiesForType calls fail for some
secondary codecs. We didn't used to call getCapabilitiesForType
on secondary codecs, and so avoided this issue.
This change suppresses any exceptions encountered when
querying secondary decoders on API levels prior to N, to
restore previous behavior when a failure occurs.
Issue: #1534
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=124010242
This allows the TrackRenderer superclass to do things when the renderer is
reset, and makes resetting consistent with other 'events' on renderers.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123974633
The allocator and buffering policy (e.g. how large the buffer
is) is moving to the top level as part of playlist support,
so it no longer makes sense to inject these parameters as
args into ExtractorSampleSource's constructor.
Instantiating the allocator and buffer size inside of the
source is temporary and only until they're moved up.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123968976
It is possible to add use of this feature in the demo app, restricting the selected
tracks to the viewport size. But it should be added in a future CL.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123965232
The bug this cl fixes is basically an uncaught case of []
Issue: #1567
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123963779
This removes "message sent on dead thread" warnings in nearly
all cases, and guarantees delivery of load cancelation to event
listeners.
Issue: #426
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123957691
- The main goal of this change is to remove the need for
discardSamplesForDisabledTracks() in continueBuffering
in HlsTrackStreamWrapper and ExtractorSampleSource. As
a result we'll also save one Allocation per disabled
track.
- Another benefit of this change is that Allocator.trim
calls can no longer be dropped. This could previously
happen on player release if the playback thread's
Looper quit before delivery of onLoadCanceled which
performed the trim in the case that a load needed to be
canceled at the time of the release.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123957434
Plus remove some unused util methods + make a few
warnings go away.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123888046
This is similar to suppression of failures disabling the
renderers (in the block above).
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123847812
- Remove need for SampleSources to ref ChunkSources.
- Correctly propagate errors loading manifests through
ChunkSource components.
- Fix some misc warnings.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123732918
Aside from deleting code, this change (a) adds retry
functionality for initial manifest fetches, (b) uses
the same loader for the utc timing fetch as for the
manifest.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123634812
The issue is visible for DRM playbacks because
BUFFER_FLAG_ENCRYPTED is being cleared, which results
in trying to play encrypted samples without decryption.
The issue would have also incorrectly cleared key frame
flags too, for both clear and encrypted playbacks, but
I don't think we're using that flag downstream anywhere
and so the issue wasn't visible.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123426582
We allow 2 configurations: Forcing seamless adaptiveness and allowing
mixed mime types adaptation.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123420364
Pull more logic up to HlsSampleSource. Somewhat regretfully,
this also backs out the optimization work done toward the
ref'd issue. I think that's one for another time perhaps...
Issue: #551
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123417413
It's not a comprehensive solution for setting the
playback speed because (a) it doesn't apply when
audio is disabled, and (b) we do some timestamp
interpolation in a few places where we assume real
time. However in practice it works pretty well and
most apps probably don't allow disabling of audio,
so I think it makes sense to expose it.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123406605
This is in preparation for changing the TrackStream while keeping the renderer
enabled, to give seamless transitions for playlists.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123394733
Added a new method TestUtil.consumeTestData() to emulate
the exact behaviour and modified OggExtractorFileTests to
use it. Rest of the test will be fixed in follow up CLs.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123320538
Initially only the first source index is used. In a later change,
ExoPlayerImplInternal will create SampleSources for different playlist item
indices as necessary.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123312595
This change moves generally useful functionality (load timing
and fatal error propagation) inside of Loader, so that callers
don't have to duplicate effort.
The change also makes use of generics so that the callback
receives a Loadable with a more specific type.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123304369
Where multiple messages are required to be sent in order
to perform a player reconfiguration, it will usually be
desirable to process all messages in a single "transaction"
(i.e. without any rendering happening when only some of
the messages have been applied).
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123228334
This should be a no-op change, but it seems sensible to
have them not overlap.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123208140
- Simplify setSurface by always blocking when the surface
is being cleared.
- Add analogous setVolume API.
- Support setting of playback params for extension audio
decoders. We can probably use this to add a setPlaybackRate
API to SimpleExoPlayer for API level 23 and above only,
but we'd probably want to make sure it works properly when
there's no audio track too, so some thought will be
required.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123204581
DrmSessionManager is now an actual manager. For each session
request it may return a shared session (as things work now,
and as is suitable for Widevine VOD) or a separate instance
(e.g. for PlayReady where audio and video are protected with
different keys).
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123203664
DemoPlayer moves into core library as SimpleExoPlayer, which
implements ExoPlayer.
Issue: #383
Issue: #592
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=123090184
With this CL:
* The first supported video track found is selected.
* The first supported audio track with preferred language is selected.
If none, we fallback to the first supported one.
* For text, we only present a selection if one of the tracks has the
preferred language and the track is flagged as forced.
TODO list:
* Add a selection policy for video (probably related with adaptiveness).
* We should decide what to do with the default flag.
* Perhaps, if no audio with the preferred language(assuming there is one)
is found, we should fall back to a text track that has the preferred
language.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122985006
1. Properly split out listening responsibilities so that
DemoPlayer only listens to its own components.
2. Revert StreamingDrmSessionManager UnsupportedDrmScheme
exceptions back to how they worked in V1, and inject
a DrmSessionManager rather than a MediaDrmCallback into
DemoPlayer.
This much better prepares DemoPlayer for promotion into
the core ExoPlayer library, since it removes assumptions
such as what SampleSource and DrmSessionManager impls
might be used with it.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122980952
- Parse APIC and TextInformation frames.
- In MPEG-TS, don't mind if packets contain end padding.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122743786
Bigger changes will be coming, but this gets DRM back
up and working again.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122719062
Previously, rounding down to the nearest frame wasn't
being done correctly; seeking could seek into a sample.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122618668
This field will allow us to flexibly add information provided by the container
or streaming manifests related to the tracks that must be selected.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122559858
- Don't report errors to listeners if playback will immediately
fail. Doing so is redundant, since such errors are immediately
reported through ExoPlayer's listener as a result of playback
failure. We were also reporting these errors inconsistently
across renderers.
- Reduce code duplication through EventDispatcher classes.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122519976
- Parse duration from mehd box for FMP4.
- Handle absent tfdt boxes by accumulating decode time
from one fragment to the next.
Issue: #1529
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122512416
This is required because in V2 we'll be instantiating DRM
session managers before the ExoPlayer instance (and hence
we wont have the Looper). This logic will be further cleaned
up in later CLs that overhaul DrmSessionManager in more
depth.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122453765
Search up to 4 KB for both fragmented and unfragmented files.
Detect files with an mvex box in their moov box as fragmented.
Fix reading of brands.
Issue: #1523
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122429548
Now [DASH/SS]SampleSource instances are creating ChunkTrackStream
instances dynamically, it makes sense to always require that the
min retry count be injected from that level. The SampleSource
implementations should also use the retry count when refreshing
the manifest.
The option to configure the retry count on the SampleSource classes
will arrive in a later CL.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122424774
Allows the user to provide their own criteria for selecting one of the available tracks
without having to reimplement the track assignment logic, i.e. linking type x tracks with
type x renderer.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122422087
1. Merge HlsOutput and HlsSampleSource -> HlsTrackStreamWrapper.
2. Rename HlsSource -> HlsSampleSource2. This will be renamed to
HlsSampleSource in a subsequent CL.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122415970
- Increment skippedBufferCount for VPX and ADTR.
- Set maxConsecutiveDroppedOutputBufferCount count for VPX.
Tweak its meaning to ignore skipped frames.
- Remove outputFormat/outputBuffer changed counts. These add
limited value. Also, MediaCodec is moving toward a model where
you don't see the output buffers changing because you dequeue
them one at a time (like how our extension decoders work).
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122258530
1. AudioDecoderTrackRenderer now reports decoder initialization
and AudioTrack underruns.
2. AudioDecoderTrackRenderer can now render more than one output
buffer per call to doSomeWork, to be consistent with
MediaCodecAudioTrackRenderer. This may also prevent audio
underruns in the case that audio is fragmented into many small
buffers.
3. AudioDecoderTrackRenderer now has an overridable method that
receives the audio session id, to be consistent with
MediaCodecAudioTrackRenderer.
4. Fix unsafe event notification in LibvpxVideoTrackRenderer.
5. Vpx and AudioDecoder extensions now increment the CodecCounter
inputBufferCount field correctly.
6. Decoders now have names :).
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122250009
The FFmpeg extension can support various different MIME types, so pass the
whole format to configure the decoder.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122246589
When skipToData is called the peek position is not equal
to the read position of the input. Hence peeking a chunk
header and then skipping by that amount wont do the right
thing :)!
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122157603
Mechanical step to create a new HlsSource in the library.
Note that naming is now confusing. This will be fixed in
the next CL, when:
HlsSource -> HlsSampleSource
HlsSampleSource+HlsOutput -> HlsOutput
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=122054800
Support for opus content in ogg container.
TODO: Sample duration and seeking.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=121940392
- This means DrmInitData is propagated through sample queues (i.e.
is effectively attached to every sample, so we can see when it
changes when reading from the queue).
- It also allows different DrmInitData per track, which is possible
in muxed MKV/WebM, and per Representation for DASH, although we
wont be able to seamlessly adapt in the latter case.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=121821928
I've seen a sample that has a LIST chunk prior to fmt. It
seems best just to skip all chunks until the format is
located. As per the RIFF spec:
"Programs must expect (and ignore) any unknown chunks
encountered, as with all RIFF forms."
https://www.aelius.com/njh/wavemetatools/doc/riffmci.pdf
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=121682470
This allows limiting the horizontal extension of the cues.
NOTE: So far, we only support percentages for size magnitudes.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=121682404
SubtitleParser includes a method updating the current
playback position for parsers that need to be aware
of the position in order to determine how to batch
captions (i.e. Eia608Parser).
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=121519712
- This moves to a single DrmInitData implementation that supports
both specific and universal UUID matching. You can also do a
combination of the two, which was not previously possible.
- The object model is simplified as a result. This is a precursor
to a change where DrmInitData will be included directly in the
Format to enable key rotation.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=121472592
In order to apply track type specific policies in
the track selection, we need to know the type of
the track renderer, this method returns one of the
constants defined in C.java of the form TRACK_TYPE_*
to allow renderer classification.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=121369021
If the media format changes from one Format object to another
one that's is equal (but not the same object), readData would
end up performing an expensive equality evaluation on every
invocation.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=121362300
This will be required for key rotation to work, since
we'll need a way to determine whether or not the init
data has changed.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=121018211
Only edit lists that truncate the first/last sample are supported, which is
sufficient to handle AAC encoder delay/padding.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=121011278
