962332 - [B2G][Media Recorder]Unable to simultaneously record audio in single stream/process if second recording has timeslice but first does not.

Status: RESOLVED INVALID

(Core :: Audio/Video: Recording, defect)

Description

[Ross Kuhlman ([e-mail])]

Attachment: TwoRecordersOneStreamOneTimeslice.txt

Device is simultaneously creating two recordings in a single stream and content process. If the second recording has a timeslice and the first does not, then the second audio track will not record properly. Attempting to launch the Blob URL will display a warning: "Video can't be played because the file is corrupt."

Repro Steps:
1) Updated Buri to BuildID: 20140121004137
2) Go to http://mozilla.github.io/qa-testcase-data/webapi/mediarecorder/index.html.
3) Select microphone checkbox, and hit setup with 2 provided in number of recorders.
4) Grant permissions.
5) Select 'start recording' on the first set of media recorder controls.
6) Select 'start recording' with a 250 timeslice in the 2nd set of media recorder controls.
7) After five seconds of generating sound, stop each active recording.
8) Examine each blob URL generated.

Actual:
The media from the first recorder does not display a video length. Only the current timestamp of the video is displayed. The media from the second recorder is not available. The user is presented with a warning "Video can't be played because the file is corrupt.

Expected:
Both audio tracks are saved properly. User can play them without issues, and the length of the clip is displayed.

Environmental Variables:
Device: Buri v1.3 Moz RIL
BuildID: 20140121004137
Gaia: 47049555282a9a01fb60d1e1421b57e2810c96f5
Gecko: 6f7dfe36ab6c
Version: 28.0a2
Firmware Version: v1.2-device.cfg

Notes:
This issue only occurs if the second recorder has a timeslice. If the first recorder has a timeslice, then both audio tracks function appropriately.

Repro frequency: 100%
Link to failed test case: https://moztrap.mozilla.org/manage/case/10170/
See attached: logcat, screenshot.

Comment 1

[Ross Kuhlman ([e-mail])]

Attachment: TrackScreenshots.jpg

Issue also occurs in 1.2

Environmental Variables:
Device: Buri v1.2 Moz RIL
BuildID: 20140114004002
Gaia: 539a25e1887b902b8b25038c547048e691bd97f6
Gecko: 42a1c35fc831
Version: 26.0
Firmware Version: v1.2-device.cfg

Comment 2

[Jason Smith [:jsmith]]

I can reproduce this on desktop.

Comment 3

[Randy Lin [:rlin] - Be Mozillian on 2015/01/26]

Can't reproduce on 29.0a1 (2014-01-25) nightly...=.=

Comment 4

[Jason Smith [:jsmith]]

(In reply to Randy Lin [:rlin] from comment #3)
> Can't reproduce on 29.0a1 (2014-01-25) nightly...=.=

What STR are following? I just double checked on Nightly 1/25/2014 & I was still able to reproduce.

Comment 5

[Jason Smith [:jsmith]]

Attachment: No Timeslice oga file (Working)

Comment 6

[Jason Smith [:jsmith]]

Attachment: Timslice Oga File (Not Working)

Comment 7

[Jason Smith [:jsmith]]

One idea you might want to try here is to start up a debugger & playback the not working oga file. Then, identify using the debugging why decoding of that oga file failed. That could point out what failed in the Media Encoder.

Comment 8

[Randy Lin [:rlin] - Be Mozillian on 2015/01/26]

I don't fill the Number of Recorders [], just click setup twice.
The broken file seems miss the ogg header.

Comment 9

[Jason Smith [:jsmith]]

(In reply to Randy Lin [:rlin] from comment #8)
> I don't fill the Number of Recorders [], just click setup twice.
> The broken file seems miss the ogg header.

Ah - that's the problem. You need to fill number of recorders to 2.

Comment 10

[Jason Smith [:jsmith]]

(In reply to Jason Smith [:jsmith] from comment #9)
> (In reply to Randy Lin [:rlin] from comment #8)
> > I don't fill the Number of Recorders [], just click setup twice.
> > The broken file seems miss the ogg header.
> 
> Ah - that's the problem. You need to fill number of recorders to 2.

The difference btw in my test app of clicking setup twice vs. using # of recorders to 2 is that clicking setup twice would have 2 media recorder objects using separate media streams from each other, where as doing # of records to 2 would have 2 media recorder objects using the same media stream.

Comment 11

[Randy Lin [:rlin] - Be Mozillian on 2015/01/26]

OK. got it. Thanks.

Comment 12

[Randy Lin [:rlin] - Be Mozillian on 2015/01/26]

Attachment: test case patch

I have some fix with the test case.
blobDataAvailable obj on 2nd onstop callback changed by others instance.

Comment 13

[Jason Smith [:jsmith]]

Comment on attachment 8365832 [details] [diff] [review]
test case patch

Review of attachment 8365832 [details] [diff] [review]:
-----------------------------------------------------------------

The intention of the patch looks right, although there's a bug in it.

I'll close this out as invalid, as this isn't a bug in gecko.

Feel free to open a pull request against https://github.com/mozilla/qa-testcase-data with the fix here. Otherwise, I'll probably get to this tomorrow.

::: qa/common.js
@@ +28,4 @@
>  
>    for(var i = 0; i < numberOfRecorders; i++){
>      var mediaRecorder = new MediaRecorder(stream);
> +    mediaRecorder.blobdata = [];

I don't think this is going to work, as there's a variable inconsistency. This uses mediaRecorder.blobData as a start, but later evt.target.blobdata.push is used. We should be using mediaRecorder.blobData throughout the code here.

Comment 14

[Jason Smith [:jsmith]]

Leaving needinfo on myself to ensure that the test case gets fixed here.

Comment 15

[Jason Smith [:jsmith]]

Fixed this here - https://github.com/mozilla/qa-testcase-data/commit/9d4b2e082f8e1af65462478ce26bb373d84ae4b5

Comment 16

[Jason Smith [:jsmith]]

Looks like this bug still reproduces with that test case fix.

Comment 17

[Randy Lin [:rlin] - Be Mozillian on 2015/01/26]

I think the mediaRecorder would reference to wrong object, so I change to use evt.target.
You can dump the object send before updateBlobURLUI and found the data would be change in this case.
+      updateBlobURLUI(new Blob(mediaRecorder.blobData, { 'type' : mimeType }));
+      mediaRecorder.blobData = [];

Comment 18

[Jason Smith [:jsmith]]

Ah that worked. Thanks for the help.