1165825 - MediaCodecReader: testcase test_playback.html failed.

Status: RESOLVED FIXED

(Core :: Audio/Video, defect)

Description

[Benjamin Chen [:bechen]]

STR:
Enable MediaCodec preference.
Run dom/media/test/test_playback.html on kk-emulator.

Testcase will timeout, seems something wrong while playing mp3 file.

[07:58:49.426] owl-funnier-id3.mp3-54 got ended
[07:58:49.563] owl-funnier-id3.mp3-54 got emptied
[07:58:49.572] vbr-head.mp3-56 got play
[07:58:49.573] vbr-head.mp3-56 got waiting
[07:58:49.574] owl-funnier-id3.mp3-54 got loadstart
[07:58:49.580] owl-funnier-id3.mp3-54 got error
[07:58:49.583] vbr-head.mp3-56 got loadstart
TEST-PASS | unknown test url | vbr-head.mp3: Name should match #3
[07:58:49.635] vbr-head.mp3-56 got suspend
TEST-PASS | unknown test url | vbr-head.mp3: Name should match #1
TEST-PASS | unknown test url | vbr-head.mp3 duration (10.03102) should be around 10
TEST-PASS | unknown test url | vbr-head.mp3 isEncrypted should be true if we have decryption keys
[07:58:50.44]  vbr-head.mp3-56 got loadedmetadata
[07:58:50.74]  vbr-head.mp3-56 got loadeddata
[07:58:50.75]  vbr-head.mp3-56 got canplay
[07:58:50.76]  vbr-head.mp3-56 got playing
[07:58:50.77]  vbr-head.mp3-56 got canplaythrough

Comment 1

[Benjamin Chen [:bechen]]

After I print some log in MediaCodecReader, I found that there are 2 kinds of timeout.

1. In EnsureCodecFormatParsed() function, the return value of dequeueOutputBuffer is INFO_FORMAT_CHANGED.
Then no ERROR_END_OF_STREAM signal in DecodeAudioDataSync()

2. In EnsureCodecFormatParsed() function the return value of dequeueOutputBuffer is ERROR_END_OF_STREAM then INFO_FORMAT_CHANGED.
Still no ERROR_END_OF_STREAM signal in DecodeAudioDataSync().

So the patch in bug 1154194 may need modify again.

Comment 2

[Benjamin Chen [:bechen]]

Correct the comment 1.
2. In EnsureCodecFormatParsed() function the return value of dequeueOutputBuffer is -11 then INFO_FORMAT_CHANGED.
Still no ERROR_END_OF_STREAM signal in DecodeAudioDataSync().

Comment 3

[Benjamin Chen [:bechen]]

The root cause is that in EnsureCodecFormatParsed(), we do not release the bufferIndex which contains INFO_FORMAT_CHANGED, so the INFO_FORMAT_CHANGED buffer still keeps there. Makes the DecodeAudioDataTask() decode nothing and no response for the mAudioPromise.

Comment 4

[Benjamin Chen [:bechen]]

Attachment: bug-1165825.v01.patch

1. Release the buffer which contain INFO_FORMAT_CHANGED (comment 3)
2. If the DecodeVideoFrameSync and DecodeAudioDataSync doesn't decode a valid sample, re-trigger the task again because we still hold a MediaPromise.
(In testcase, the dequeueOutputBuffer might return OK but there is no valid DataBuffer.)

Comment 5

[Benjamin Chen [:bechen]]

Attachment: bug-1165825.v01.patch

r=sotaro
try server: https://treeherder.mozilla.org/#/jobs?repo=try&revision=12ac876bd965

Comment 6

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/52661b02eaad

Comment 7

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/52661b02eaad