STRs: 1. Use a headset 1. Open https://jsfiddle.net/jib1/cknfm2tL/ in two different tabs. 2. In first tab, click the [Call] button and share cam + mic. Expected result: - When speaking, you hear yourself fine in headset. Actual result: - Raspy distorted voice with static noise and a one second delay. Regression range: 13:07.61 INFO: Narrowed inbound regression window from [69e98254, 73260f7c] (4 revisions) to [00f29274, 73260f7c] (2 revisions) (~1 steps left) 13:07.61 INFO: No more inbound revisions, bisection finished. 13:07.61 INFO: Last good revision: 00f292748b3e702f338589863b986f2ee50d55cf 13:07.61 INFO: First bad revision: 73260f7c6dae11039cc4c738ac8714671d8cd1df 13:07.61 INFO: Pushlog: https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=00f292748b3e702f338589863b986f2ee50d55cf&tochange=73260f7c6dae11039cc4c738ac8714671d8cd1df Regressed in bug 971528, Firefox 55. I found this trying to reproduce bug 1405893. There's a chance they're the same, though the symptoms here are instant.

Note that a headset is not required to reproduce, it just makes the problem stand out more.

According to Alex the patch for bug 1406027 seems to help. If that's the case I would have expected this to be regression from a combination of bug 1319445 and bug 1243857 (for mac), or more recently, bug 1330360. In any case, not bug 971528. Similar to bug 1381638 but I think bug 1406027 fixes the root cause of those distortions.

Regression bug is wrong I'll find a new regression range. Patch from bug 1406027 solves the problem.

mozregression result: 15:19.09 INFO: Narrowed inbound regression window from [15cc8dd6, 967aac2f] (3 builds) to [15cc8dd6, 29ac12c8] (2 builds) (~1 steps left) 15:19.09 INFO: No more inbound revisions, bisection finished. 15:19.09 INFO: Last good revision: 15cc8dd663627ec12ecd4be11a69f0caad8e4e11 15:19.09 INFO: First bad revision: 29ac12c81bcda10a981dfe3371f70a9b17a19edd 15:19.09 INFO: Pushlog: https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=15cc8dd663627ec12ecd4be11a69f0caad8e4e11&tochange=29ac12c81bcda10a981dfe3371f70a9b17a19edd paul@paul.cx Thu Jun 29 20:39:37 2017 +0000 29ac12c81bcd Paul Adenot — Bug 1341555 - Make static analysis happier. 5c2fea87df16 Paul Adenot — Bug 1341555 - Consolidate use of the MSG's AbstractMainThread, and stop having AbstractMainThread on MediaStreams. r=pehrsons e0c4f2ca44b4 Paul Adenot — Bug 1341555 - Label MSG runnables. r=pehrsons 5de53323f3ad Paul Adenot — Bug 1330360 - Create new MSGs for each nsPIDOMWindow. r=jesup 1c4d113ff7fb Paul Adenot — Bug 1371719 - Don't create a MediaStreamGraph without doing anything with it. r=jesup 3da6bb01f447 Paul Adenot — Bug 1372247 - Make the AudioOutputObserver a member of the MediaEngineWebRTCMicrophoneSource so it's not a global singleton anymore. r=jesup

I just redid the regression range and I still get the range in comment 0!

Note I have a new 2016 MBP with stereo mic. Might that explain why I'm seeing a different regression range from Alex and Andreas?

too late for 56, and probably not worth tracking for 57.

[Tracking Requested - why for this release]:

jib, since you had a different regression range, could you verify that this works in latest nightly?

wfm.