Also reported in  issue 589925 , but on Android for normal Chrome usage. The issues should be kept separate, since the particular reason for the overflow in the other issue should be investigated.

It would be good to know why this happens in the ClusterFuzz case. The overflow means that the render process doesn't consume the data fast enough, and we'll lose data. The fifo is large (1 second). Can this be expected? Could losing mic data it cause problems in the tests were it happens?

If this is expected, and don't cause any problems, would it be enough with capping the log output to say 50 lines? I'll do that anyway.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/8da4609ddcb6d0394b78fd83b15a25647622383a

commit 8da4609ddcb6d0394b78fd83b15a25647622383a
Author: grunell <grunell@chromium.org>
Date: Wed Jun 29 07:32:36 2016

Cap AudioInputSyncWriter "no room in fifo" log messages.

Also change log errors to warnings.

BUG= 589925 , 623598 

Review-Url: https://codereview.chromium.org/2107723002
Cr-Commit-Position: refs/heads/master@{#402749}

[modify] https://crrev.com/8da4609ddcb6d0394b78fd83b15a25647622383a/content/browser/renderer_host/media/audio_input_sync_writer.cc

Log output is now capped to 50 lines. Assigning to Patrik to answer the questions in comment #1.

Re #1: Anything can happen on ClusterFuzz :) We can flip audio features on or off and make arbitrary changes to the SDP and choose arbitrary codecs and any number of things.

Assuming fixed; inferno or mbarbella, please verify things got better or if we need to do more here.