Closed Bug 907213 Opened 11 years ago Closed 9 months ago

Small hangs on webRTC calls initiated from Linux

Categories

(Core :: WebRTC: Audio/Video, defect, P4)

24 Branch
defect

Tracking

()

RESOLVED WORKSFORME
Tracking Status
firefox24 - ---
backlog parking-lot

People

(Reporter: phorea, Unassigned)

Details

Attachments

(1 file)

https://wiki.mozilla.org/Releases/Firefox_24/Test_Plan#Beta_4 While testing webRTC on the following environments, we found small hang issues (the video and the sound stopped for about one second), sometimes encountered if other tabs were opened or closed: Chrome Stable on Ubuntu x64 (caller) to Firefox 24.0b4 on Win8 (callee) Caller: small hangs Callee: small hangs if browsing other websites during chat Firefox 23.0.1 on Ubuntu x64 (caller) to Firefox 24.0b4 on Mac OS X 10.8.4(callee) Caller: small hangs Callee: small hangs No hangs were noticed on previous versions of Firefox, so this seems to be a Firefox 24 regression. Will investigate on which beta this regressed and come back with details.
Is this worse or the same as in 23.0.1?
Needinfo'ing jesup here, so this is his on his radar. QA, will be helpful if we can check if this is a 24.0b4 regression specifically or did this happen in any other Fx24 beta builds ?
Flags: needinfo?(rjesup)
24beta1 and b2 were affected by bug 904598. The fix there was also checked in for 24b3. I think what we're looking for is if this is worse than in 23.0.1 or not.
We tried with Firefox 23.0.1 on Ubuntu x64 (caller) to Firefox 23.0.1 on Mac OS X 10.8.4(callee) and encounter only one hardly noticed hang. Will investigate which versions of Firefox 24 caused the regression and come back with the results.
Firefox 24b4 on Ubuntu 13.04 x64 to Firefox 24b4 on Mac 10.7 Firefox 24b3 on Ubuntu 13.04 x64 to Firefox 24b3 on Mac 10.7 On these beta versions, the hangs are worse than Fx 23.0.1. The hangs were noticed more easily while opening flash games in other tabs (not reproduced on Chrome to Chrome). Please let me know how could I help you more.
(In reply to petruta.rasa from comment #5) > Firefox 24b4 on Ubuntu 13.04 x64 to Firefox 24b4 on Mac 10.7 > Firefox 24b3 on Ubuntu 13.04 x64 to Firefox 24b3 on Mac 10.7 > On these beta versions, the hangs are worse than Fx 23.0.1. > > The hangs were noticed more easily while opening flash games in other tabs > (not reproduced on Chrome to Chrome). > > Please let me know how could I help you more. Are you able to recover from the hangs easily (within how many minutes) ? If its minor and recoverable I am not too worried. Also is there a way to get regressing bug here ? :jesup, do you suggest some logging here for you to be able to see what's going on?
Most of the hangs were 1 to 3 seconds. We encounter only very few hangs that lasted 10-20 seconds.
Chrome isolated content in different tabs into different processes, so it isn't that surprising that opening flash content in other tabs would affect us and not Chrome. Obviously we'd like to avoid any stalls while something like that is happening, but that will require some performance analysis to see where the problem actually lies. Do these hangs show up when *not* using other tabs? Since the comment made originally was "when opening and closing other tabs", does it happen opening/closing about:blank tabs or when opening an google.com tab? (Open in New Tab, not doing a search from a titlebar widget) Does it still happen if you set "browser.pageThumbs.enabled" to false (on 24) or "browser.pagethumbnails.capturing_disabled" to true (on 26)? Are these calls using Betas using the same profiles and machines as the calls using 23.0.1? Are there any extensions installed? If this is clearly repeatable and a regression in 24b3 and b4, can you continue to go back to find the start of the regression?
Flags: needinfo?(rjesup)
We had issues with the https://apprtc.webrtc.org page and couldn't figure it out what happened. We couldn't make any connection between Windows, Linux and Mac OS neither on different networks or the same network. "Connecting..." is displayed on the bottom bar for a long period of time and no connection is made. We tried with latest Nightly, Fx 23, Fx 24 beta 2 and beta 3. Unfortunately we can't help until this problem is solved. Is there any other page on which we can test? Thank you.
Can you provide the JS logs from your testing?
(In reply to petruta.rasa from comment #9) > Is there any other page on which we can test? Thank you. http://codeshare.io, tap the camera icon
I attached the logs obtained using latest Nightly. https://apprtc.webrtc.org is still not working on Nightly, but it works on Fx 24 beta 6. We encounter the same situation with http://codeshare.io page (works only on 24b6).
This would match - we disabled TURN on 23.0.1 and 24 (recent betas) because of problems including that we're preferring TURN over better connections (like p2p). This has been compounded in the last few days by a bug at apprtc that's given invalid TURN addresses. From the log: typ relay raddr 0.0.0.0 rport 0 (etc). It's giving 0.0.0.0 addresses, which are invalid, and then we're trying to use them. See bug 904598 and bug 905150
Unclear if there is anything to do on our side within the Fx 24 timeframe and this does not sounds like a Fx24 regression and is not a shipping blocker. Feel free to renom if needed.
Severity: normal → minor
Priority: -- → P3
backlog: --- → parking-lot
Is this working better now?
Flags: needinfo?(petruta.rasa)
We performed calls from Ubuntu 12.04 32-bit, Fx 41.0RC to Mac OS X 10.9.5 and Win 10 32-bit, Fx 42 beta 5: - There are no hangs while normal browsing (including tabs with youtube videos and flash games) - Small hangs (3-5 seconds) for both callers while loading websites from http://www.ebizmba.com/articles/best-flash-sites page
Flags: needinfo?(petruta.rasa)
I'm going to go out on a limb that trying to find a regression range from the Fx23/Fx24 era isn't likely to help much at this point, so I'm dropping that flag. Will leave the bug open for more investigation of the overall perf issue, however.
Mass change P3->P4 to align with new Mozilla triage process.
Priority: P3 → P4
Severity: minor → S4

I could no longer reproduce this issue on newer Firefox versions. Considering this and the fact that webrtc tests are run each cycle, I'm closing it as RWFM.

Status: NEW → RESOLVED
Closed: 9 months ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: