1528876 - enable AudioWorklet in Nightly Web Audio wpt

Status: RESOLVED FIXED

(Core :: Web Audio, enhancement, P2)

Description

[Karl Tomlinson (:karlt)]

This is Nightly and Dev Edition only at this stage due to https://hg.mozilla.org/mozilla-central/rev/f99b99cc076d06251bf16725647d52f65bf17d4d#l1.13

Comment 1

[Karl Tomlinson (:karlt)]

Attachment: Bug 1528876 enable AudioWorklet in Nightly Web Audio wpt r?padenot

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/6fdef2d548e7
enable AudioWorklet in Nightly Web Audio wpt r=padenot

Comment 3

[Cosmin Sabou [:CosminS]]

https://hg.mozilla.org/mozilla-central/rev/6fdef2d548e7

Comment 4

[guest271314]

#3 Is AudioWorklet disabled in Nightly except for wpt tests?

Comment 5

[Paul Adenot (:padenot)]

Yes. For now, we have just enough things working to run a few test to make sure we're on the right track. More is coming soon, for example in bug 1558123.

Comment 6

[guest271314]

#5 From what have been able to gather, there appears to be a bug where when requestFrame() of canvas.captureStream(0) https://bugzilla.mozilla.org/show_bug.cgi?id=1231131 is used and a MediaStreamTrack from Web Audio is recorded using MediaRecorder the last 1 second or less of audio is clipped when compared to using canvas.captureStream(60) without using requestFrame(), see https://bugzilla.mozilla.org/show_bug.cgi?id=1539186 and https://bugzilla.mozilla.org/show_bug.cgi?id=1545008.

The bug should be able to be reproduced at Firefox/Nightly without using AudioWorklet.

Will probably shortly file a wpt issue for the purpose of standardizing a test to confirm the bug and bugs reports at Firefox and Chromium.

Comment 7

[guest271314]

#5 Did you delete the comment #6 as "off-topic"? Am trying to confirm a bug that apparently appears in both Chromium and Nightly. Just happened to narrow the cause down to use of requestFrame() in code that uses AudioWorklet and was attempting to confirm the cause by running the same code at Nightly. In any event https://github.com/web-platform-tests/wpt/issues/17558

Comment 8

[Paul Adenot (:padenot)]

If you see a bug, please open a bug with a test case. What you're describing has nothing to do with AudioWorklet.

Comment 9

[guest271314]

#8 Perhaps you did not gather what was asking at #6? Was asking if it would be possible to test the same code that ran at Chromium at Nightly. Prior to using the AudioWorklet (to use process() in place of requestAnimationFrame() by posting message to main thread to execute requestFrame()) code at Chromium had not previously observed the audio being muted for the last 1 second at Chromium, though which had previously observed at Nightly. After running the code over and over with and without captureStream(0) and requestFrame() found the common denominator which lead to last 1 second or less being muted was requestFrame(). The simplest test solution would be to run the exact same code at both Chromium and Firefox (save for the requestFrame() being defined differently at the browsers) to confirm the cause of the muting of audio was in fact <canvas>. Since testing of AudioWorklet is being enabled for wpt it is sensible to ask if users in the wild can also test AudioWorklet at Nightly, due to the fact that actual human tests can and do find edge cases that automated tests generally cannot and do not find. Am currently composing a version using requestAnimationFrame() .

Comment 10

[guest271314]

#8 This is the original code https://plnkr.co/edit/NoVi5Q?p=preview. Still working on composing comparable code without using AudioWorklet