The WebRTC echo canceller (AEC2) is sensitive to incorrect the timestamps that are reported on Chrome OS. |
||||||||||||||
Issue descriptionWhen the the audio timestamps reported for the playout and capture audio by Chrome to WebRTC are incorrect, the WebRTC echo canceller (AEC2) has been observed ending up in a state where there is full and persistent echo. Since the audio timestamps reported by Chrome OS are sometimes incorrect, the AEC sensitivity to that is an issue on the Chrome OS platform.
,
Dec 22 2017
This issue seems to be affecting VoIP calls from all Chrome OS devices. We would therefore like to merge the fixing CL to both M63 and M64.
,
Dec 22 2017
,
Dec 22 2017
When did this issue arise? M63, or? I see that this is a high priority, but can you share details on Customer impact (% of customers, what they'd experience, etc.)? Also added release labels.
,
Dec 22 2017
As far as we know, the issue is affecting all VoIP calls using Chrome OS, but most calls don't seem to end up echoing. From the reports we have, it seems that once a machine ends up echoing, it continues to do that throughout the call, severely disturbing the call. The recordings we have made and logs captured strongly indicates that the CL https://chromium-review.googlesource.com/c/chromium/src/+/592403 is what started the regression. That CL changed to using timestamps provided by the OS and from what we see those are not always correct on Chrome OS. A merge of a revert of that CL was deemed too hard to achieve since it was landed so long ago. The above CL landed in M62 so the regression started in M62. It was, however, not discovered until Dec 13. An issue (https://bugs.chromium.org/p/chromium/issues/detail?id=797272&desc=5) has been filed for adressing the above CL, but in the meantime we would prefer to solve the issue by this merge.
,
Dec 22 2017
Please lower the priority for P0. This isn't impacting all users and has been ongoing for some time. Approving the merge request, however, for M62 specifically.
,
Dec 22 2017
,
Dec 22 2017
Apologies; my mistake #6 should have read, "Approving the merge request, however, for M64 specifically." Fat fingered; PEBCAK
,
Dec 22 2017
,
Dec 22 2017
The following revision refers to this bug: https://webrtc.googlesource.com/src.git/+/9c262749f5bbb2dc825c57468328315945a5d293 commit 9c262749f5bbb2dc825c57468328315945a5d293 Author: Per Åhgren <peah@webrtc.org> Date: Fri Dec 22 23:39:44 2017 Merge of Enforcing a stream delay of 0 to be assumed in the AEC on Chrome OS This CL forces the AEC2 to assume a stream delay of 0, thereby avoiding that the incorrect stream delays reported on Chrome OS causes echo issues. TBR=henrik.lundin@webrtc.org (cherry picked from commit 11556464a62ea14db9eee5e6c2a79a40aa1700ab) Bug: chromium:797274 , chromium:797272 Change-Id: I10f295c9f1d735622c55fc56be99a14c6cdd88a2 Reviewed-on: https://webrtc-review.googlesource.com/36081 Reviewed-by: Per Åhgren <peah@webrtc.org> Reviewed-by: Henrik Lundin <henrik.lundin@webrtc.org> Commit-Queue: Per Åhgren <peah@webrtc.org> Cr-Original-Commit-Position: refs/heads/master@{#21432} Reviewed-on: https://webrtc-review.googlesource.com/36221 Cr-Commit-Position: refs/branch-heads/64@{#7} Cr-Branched-From: aede67a199ae0552074bfec4bb03cc9a6a5fba0f-refs/heads/master@{#20918} [modify] https://crrev.com/9c262749f5bbb2dc825c57468328315945a5d293/modules/audio_processing/BUILD.gn [modify] https://crrev.com/9c262749f5bbb2dc825c57468328315945a5d293/modules/audio_processing/echo_cancellation_impl.cc [modify] https://crrev.com/9c262749f5bbb2dc825c57468328315945a5d293/modules/audio_processing/echo_cancellation_impl.h
,
Jan 2 2018
M63 went stable a few weeks ago. Only accepting urgent merges e.g. P0 security fixes at this point.
,
Jan 2 2018
,
Jan 3 2018
The following revision refers to this bug: https://webrtc.googlesource.com/src.git/+/dc5904661c4c8b357e19595d80b2881db3f4a71b commit dc5904661c4c8b357e19595d80b2881db3f4a71b Author: Per Åhgren <peah@webrtc.org> Date: Wed Jan 03 00:15:38 2018 Merge of Enforcing a stream delay of 0 to be assumed in the AEC on Chrome OS This CL forces the AEC2 to assume a stream delay of 0, thereby avoiding that the incorrect stream delays reported on Chrome OS causes echo issues. TBR=henrik.lundin@webrtc.org (cherry picked from commit 11556464a62ea14db9eee5e6c2a79a40aa1700ab) Bug: chromium:797274 , chromium:797272 Change-Id: I10f295c9f1d735622c55fc56be99a14c6cdd88a2 Reviewed-on: https://webrtc-review.googlesource.com/36081 Reviewed-by: Per Åhgren <peah@webrtc.org> Reviewed-by: Henrik Lundin <henrik.lundin@webrtc.org> Commit-Queue: Per Åhgren <peah@webrtc.org> Cr-Original-Commit-Position: refs/heads/master@{#21432} Reviewed-on: https://webrtc-review.googlesource.com/36220 Cr-Commit-Position: refs/branch-heads/63@{#14} Cr-Branched-From: bef8a5d2ca5413c680995584b8c0976852ba5f25-refs/heads/master@{#20237} [modify] https://crrev.com/dc5904661c4c8b357e19595d80b2881db3f4a71b/modules/audio_processing/BUILD.gn [modify] https://crrev.com/dc5904661c4c8b357e19595d80b2881db3f4a71b/modules/audio_processing/echo_cancellation_impl.cc [modify] https://crrev.com/dc5904661c4c8b357e19595d80b2881db3f4a71b/modules/audio_processing/echo_cancellation_impl.h
,
Jan 8 2018
,
Feb 12 2018
This issue has been approved for a merge. Please merge the fix to any appropriate branches as soon as possible! If all merges have been completed, please remove any remaining Merge-Approved labels from this issue. Thanks for your time! To disable nags, add the Disable-Nags label. For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
,
Feb 13 2018
|
||||||||||||||
►
Sign in to add a comment |
||||||||||||||
Comment 1 by bugdroid1@chromium.org
, Dec 22 2017