The millisecond field is deprecated and will be removed once the
dependencies have been updated.
BUG=webrtc:7760
Review-Url: https://codereview.webrtc.org/3010623002
Cr-Commit-Position: refs/heads/master@{#19622}
This CL introduces common functions for adding and releasing
native references to reduce the number of JNI functions.
BUG=webrtc:8043
Review-Url: https://codereview.webrtc.org/3003213002
Cr-Commit-Position: refs/heads/master@{#19547}
This function is currently only available in EglBase14, which is not in
the api. This CL adds the swapBuffer method to the public EglBase
interface. For EglBase10, the presentation time is just ignored.
BUG=webrtc:8155
Review-Url: https://codereview.webrtc.org/3003843002
Cr-Commit-Position: refs/heads/master@{#19518}
The field trials enables producing new VideoFrames in camera classes.
This field trial should be enabled if VideoSinks are used.
BUG=webrtc:7749, webrtc:7760
Review-Url: https://codereview.webrtc.org/2984633002
Cr-Commit-Position: refs/heads/master@{#19467}
VideoSink is not assumed to take ownership of the frame, so it is not
necessary to release it unless it is retained for future use.
BUG=webrtc:7749
Review-Url: https://codereview.webrtc.org/3003563002
Cr-Commit-Position: refs/heads/master@{#19440}
Also fixes a small bug in EglRenderer where if it was passed a custom
frame, it would leak a reference.
BUG=b/64887218
Review-Url: https://codereview.webrtc.org/3003533002
Cr-Commit-Position: refs/heads/master@{#19432}
We are already asserting in tryDeliverTextureFrame() that we are on the
SurfaceTextureHelper handler thread, but this behavior isn't guaranteed
by Android.
Also updated docs for OnTextureFrameAvailableListener, since other
classes (e.g. capturers) were also asserting that they were called on
the SurfaceTextureHelper thread.
Bug: webrtc:8087, b/64609528
Change-Id: I47148c452f66f24477e438b06ef124a96965853d
Reviewed-on: https://chromium-review.googlesource.com/610495
Reviewed-by: Sami Kalliomäki <sakal@webrtc.org>
Commit-Queue: Jonathan Yu <yujo@chromium.org>
Cr-Commit-Position: refs/heads/master@{#19398}
VideoSinks receive the new kind of VideoFrames and will replace
VideoRenderers. Converting from old texture frames to VideoFrames will
involve conversion to I420 so it is not recommended to use VideoSinks
before all sources produce VideoFrames.
BUG=webrtc:7749, webrtc:7760
Review-Url: https://codereview.webrtc.org/3002553002
Cr-Commit-Position: refs/heads/master@{#19335}
This has been a frequent source of confusion, especially since the
method names don't match anything in the standard exactly.
BUG=None
NOTRY=True
Review-Url: https://codereview.webrtc.org/2994733002
Cr-Commit-Position: refs/heads/master@{#19290}
Previously, the matrix in VideoFrame was used to crop and scale the
frame. This caused complications because webrtc::VideoFrame doesn't
include a matrix. cropAndScale method is added to VideoBuffer class for
cropping and scaling instead.
BUG=webrtc:7749, webrtc:7760
Review-Url: https://codereview.webrtc.org/2990583002
Cr-Commit-Position: refs/heads/master@{#19179}
When you create multiple "PeerConnectionFactory"s, they end up using
the same NetworkMonitor singleton. But the second one's
"AndroidNetworkMonitor" class (in C++) wasn't getting the expected
network list update, and as a result it wasn't binding sockets to
networks successfully, acting as if the networks didn't exist.
The solution is just to move "updateActiveNetworkList" to
"startMonitoring". This CL also does some other minor
cleanup/refactoring, and fixes a more corner-casey issue where, if the
first PeerConnection is destroyed, the second one would stop receiving
network updates.
BUG=webrtc:7946
Review-Url: https://codereview.webrtc.org/2990693002
Cr-Commit-Position: refs/heads/master@{#19156}
HardwareVideoEncoderFactory can now take an EglBase.Context on creation.
When it does, it creates video encoders in texture mode. It uses the
COLOR_FormatSurface colorFormat. It passes the EglBase.Context to the
HardwareVideoEncoder.
The HardwareVideoEncoder sets up an input surface for its codec and handles
incoming frames by drawing them onto the input surface.
BUG=webrtc:7760
R=pthatcher@webrtc.org, sakal@webrtc.org
Review-Url: https://codereview.webrtc.org/2977153003 .
Cr-Commit-Position: refs/heads/master@{#19083}
HardwareVideoDecoder is now a listener for SurfaceTextureHelper. It takes a
SurfaceTextureHelper on construction. If it is non-null, it operates in texture
mode instead of byte-buffer mode.
When in texture mode, the HardwareVideoDecoder renders output frames to a Surface,
listens for the texture frame to become available, wraps it in a VideoFrame, and
pushes it to the decoder callback.
As in MediaCodecVideoDecoder, it may queue up to three buffers while waiting for
the surface to become available for rendering. If more buffers are queued, it will
drop the oldest.
This change also implements the VideoFrame.TextureBuffer and reorganizes code
for wrapping an existing ByteBuffer into an I420Buffer. This makes it easier
to implement the texture buffer's ToI420() method.
BUG=webrtc:7760
R=pthatcher@webrtc.org, sakal@webrtc.org
Review-Url: https://codereview.webrtc.org/2977643002 .
Cr-Commit-Position: refs/heads/master@{#19081}
The surface view renderer size was set to match parent so it couldn't
adjust based on the frame size. The size is now set to wrap_content
which allows the renderer to adjust. The root element of the call
activity is changed to FrameLayout to allow the renderer to center.
requestLayout is added to SurfaceView setScalingType so onMeasure gets
called again.
BUG=webrtc:7901
Review-Url: https://codereview.webrtc.org/2978173002
Cr-Commit-Position: refs/heads/master@{#19073}
If a SurfaceViewRenderer is reinitialized, the onFirstFrameRendered
callback is not fired.
Ensure that we reset the flag when the SurfaceViewRenderer is
initialized.
BUG=webrtc:7985
Review-Url: https://codereview.webrtc.org/2981793002
Cr-Commit-Position: refs/heads/master@{#19016}
The conversion code was wrong because it assumed the 3x3 matrix is a
XYZ-matrix when it really is XYW-matrix. We have to override the matrix
for I420 frames to flip the vertically before rendering.
R=magjed@webrtc.org
Bug: webrtc:7760
Change-Id: I1f08c1a929bf5721706e2a902701100cf7a9c31d
Reviewed-on: https://chromium-review.googlesource.com/541346
Reviewed-by: Magnus Jedvert <magjed@webrtc.org>
Commit-Queue: Sami Kalliomäki <sakal@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#18801}
The implementation creates an Android hardware video decoder. It is built
around the same patterns as the HardwareVideoEncoderFactory.
This change pulls some shared code and constants into a common "utils" class.
Finally, adds an instrumentation test for the HardwareVideoDecoder.
BUG=webrtc:7760
Change-Id: Iea6eaae7727925743cb54f7c3153a6c07d62f55d
Reviewed-on: https://chromium-review.googlesource.com/536254
Commit-Queue: Bjorn Mellem <mellem@webrtc.org>
Reviewed-by: Sami Kalliomäki <sakal@webrtc.org>
Reviewed-by: Peter Thatcher <pthatcher@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#18686}
Adds the VideoEncoderFactory interface and implements it for use with HardwareVideoEncoder. This uses MediaCodecVideoEncoder's initialization code as an example.
BUG=webrtc:7760
Change-Id: I9fbc93ce9ac4ad866750a4386c4f15e800a3073e
Reviewed-on: https://chromium-review.googlesource.com/530063
Commit-Queue: Bjorn Mellem <mellem@webrtc.org>
Reviewed-by: Sami Kalliomäki <sakal@webrtc.org>
Reviewed-by: Peter Thatcher <pthatcher@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#18636}
Guarded by field trial - similar to high profile encoder.
If high profile is requested, but device do not support it
then fallback to baseline profile.
BUG=b/34816463
Review-Url: https://codereview.webrtc.org/2936313002
Cr-Commit-Position: refs/heads/master@{#18619}
This field shouldn't have been in the class in the first place.
Bug: webrtc:7760
Change-Id: If3c1d24f18a643249da1ed072bdfe06a37a7da12
Reviewed-on: https://chromium-review.googlesource.com/535539
Reviewed-by: Magnus Jedvert <magjed@webrtc.org>
Commit-Queue: Sami Kalliomäki <sakal@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#18593}
These interfaces will be used by the future refactoring that will
allow clients to provide custom codec implementations.
Change-Id: If199bc2807e1c27094c05983c62fa43d2eec5700
Bug: webrtc:7760
Reviewed-on: https://chromium-review.googlesource.com/522065
Commit-Queue: Sami Kalliomäki <sakal@webrtc.org>
Reviewed-by: Magnus Jedvert <magjed@webrtc.org>
Reviewed-by: Bjorn Mellem <mellem@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#18441}
Color distortion also happens on Android L. Tested on the Mi 4.
BUG=webrtc:7681
Review-Url: https://codereview.webrtc.org/2894643003
Cr-Commit-Position: refs/heads/master@{#18423}
This new VideoFrame class closesly matches the C++ webrtc::VideoFrame
and webrtc::VideoFrameBuffer classes. It's supposed to replace the
existing VideoRenderer.I420Frame. The purpose is to clean up the code
and support more frame formats.
BUG=webrtc:7749
Review-Url: https://codereview.webrtc.org/2915083002
Cr-Commit-Position: refs/heads/master@{#18404}
