Edward: I saw you had made some changes to the file the test is in. Maybe that was just refactoring stuff, but I wanted to check if you're aware of any changes that could have affected these tests?

I'm not going to disable it since only on the FYI bot. Could it be a bot issue?

Assuming you mean this change (https://codereview.chromium.org/2204773004) I can't see how it might be affecting this tests.
I also doubt it is a bot issue, but I'm not so confident about it.

I just looked at the revision history.
https://chromium.googlesource.com/chromium/src/+log/master/content/browser/webrtc/webrtc_audio_browsertest.cc

OK, then I'll see who should own this.

The test times out waiting for audio to go silent. I'm not sure how to interpret the dumped samples. There's lots of zeroes at the end, but the average is obviously too large. What can we make of this? Is the audio muted but it takes too long? Or is it not muted at all? (I don't know the input signal.)

Patrik, could you have a look at the test output (also attached)? We've some weird issues on some bots before but the muting shouldn't be device related.

Deleted: output.txt 54.1 KB

output.txt 54.1 KB View Download

Looks like Chrome has gotten slower at muting as of late. I think I can make the silence check a bit faster though since we're clearly getting silence at the end, and the test should succeed in those cases. It's not a performance test of how fast mute is.

All right, in the failure there were 130k samples. I'll make the code just look at the last 10k samples and I think we should be fine.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/4589e9a6c8fa01dab372c1b35a149ceaed8215f1

commit 4589e9a6c8fa01dab372c1b35a149ceaed8215f1
Author: phoglund <phoglund@chromium.org>
Date: Tue Sep 27 07:46:17 2016

Only look at the last 10K samples for WebRTC muting tests.

Chrome has maybe gotten a bit slower at muting, but it's at most 1 sec.
This test should not be a performance test of muting speed though,
merely that muting happens, so update the test to just look at the last
10K samples in case we're unlucky and get a beep into the samples
window before the mute kicks in.

BUG= 648879 
R=ehmaldonado@chromium.org

Review-Url: https://codereview.chromium.org/2366303002
Cr-Commit-Position: refs/heads/master@{#421139}

[modify] https://crrev.com/4589e9a6c8fa01dab372c1b35a149ceaed8215f1/content/test/data/media/webrtc_test_audio.js

I wonder why this isn't seen on the ordinary (not FYI) bot?

cc hbos@, current sheriff, fyi and to check if the flake goes away.

I think that's just a coincidence. The ordinary bot is probably just a bit faster or slower. Either way, this should be fixed now.