New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 762556 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Last visit > 30 days ago
Closed: Dec 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Linux , Android
Pri: 2
Type: Bug-Regression



Sign in to add a comment

42.8% regression in webrtc_perf_tests at 19663:19666

Project Member Reported by ehmaldonado@chromium.org, Sep 6 2017

Issue description

See the link to graphs below.
 
All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=762556

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=14892b03a3b78ff516c3e7d1a4a121dc4eeb4d23e046139631f1932fac0a494e


Bot(s) for this bug's original alert(s):

webrtc-android-tests-nexus5
Components: Blink>WebRTC>Video
Labels: OS-Android OS-Linux
Cc: philipel@chromium.org
Owner: sprang@chromium.org
Status: Assigned (was: Untriaged)
Could this be related to your change?

https://chromium.googlesource.com/external/webrtc/+log/a0a5b98937de0f0bf90e17011d2cbc8c632cb409..f33cee7534f0469330b6b7aa57f84eec4b082598
Added some more Ulpfec alerts to the report.
Cc: -philipel@chromium.org sprang@chromium.org
Owner: philipel@chromium.org
I've done some testing and concluded that these regressions are not caused by the nack change linked. 
Instead, I can repro them locally by rolling past

commit d4fac6957e14068f8dbf2ab4a20006b1df044274
Author: philipel <philipel@webrtc.org>
Date:   Mon Sep 4 07:03:46 2017 -0700

    Unwrap picture ids in the RtpFrameReferencerFinder.


Is the even a negative btw? Isn't being able to achieve a more stable high media and fec bitrate a good thing?

In any case, philipel@ ptal
Possible that it's not negative, but I'd like to understand why. :)
Maybe it makes sense then that we only see Ulpfec alerts, but no Flexfec alerts: Ulpfec packets share the sequence number space with the media packets, but Flexfec does not.
Yes, that sounds like a reasonable explanation. Sounds like we might have a bug here.
Any progress on this one?
Status: Fixed (was: Assigned)

Sign in to add a comment