GitLab

media/gpu/video_encode_accelerator_tests: Fix the timing of FlushDoneTask() when VEA doesn't support Flush()

When VideoEncodeAccelerator (VEA) doesn't support Flush(),
VideoEncoderClient simulates the Flush() flow. VideoEncoderClient
starts the Flush event when the last frame is fed to VEA.
VideoEncoderClient, however, invokes FlushDoneEvent() when the
VEA read the last frame. This may be before the encoded bitstream
of the last frame is returned by BitstreamBufferReady().
This not only breaks VEA API rule; FlushDoneTask() is called
after the encoded bitstream of the last frame is returned, but
also BitstreamBufferProcessors may process less number of frames
than expected.
This CL fixes the timing of FlushDoneTask().

Bug: b:159384004, 1045825
Test: video_encode_accelerator_tests on hana
Change-Id: I47e3bc39699a727f5bdb9ccc0043dc2bd096fa6d
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2340853
Commit-Queue: Hirokazu Honda <hiroh@chromium.org>
Reviewed-by: David Staessens <dstaessens@chromium.org>
Cr-Commit-Position: refs/heads/master@{#798001}