1379437 - nsCSSFrameConstructor::IsValidSibling while interacting with YouTube videos, followed by CPU spike and inactive tabs

Status: RESOLVED DUPLICATE of bug 1380340

(Core :: CSS Parsing and Computation, defect, P3)

Description

[smartfon.reddit]

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:56.0) Gecko/20100101 Firefox/56.0
Build ID: 20170708030206

Steps to reproduce:

Open a dozen YouTube video tabs. Play and close some of them. Right click on a video while it plays and click "stats for nerds". The tab crashes. Not reproducible.
https://crash-stats.mozilla.com/report/index/efd12e0a-23ec-44d9-a27a-6251a0170709

Play another video but don't open State for Nerds. Interact with progress bar. Tab crashes. Not reproducible.
https://crash-stats.mozilla.com/report/index/969c6fb1-3ab2-4b01-8619-e7d1d0170709

Open another YouTube tab and play the video. The video plays but the tab is all blank with spinning icon in center. Click to close that tab. About 5 seconds later it closes but the audio continues to play for 10 seconds.

Laptop fans work very hard. CPU around 25% with no videos playing, while normally should be 6%. One of the Nightly processes uses lots of RAM. In this example it was using 1.1GB. Earlier today it progressively went up to 2.5GB. I think closing the offending tab makes the RAM go down after a while.

Possibly related:
I have enabled Stylo few days ago. 
Earlier today uBlock Origin developer said EasyList tracking blocker (which I use) had a problem that was fixed earlier. I followed his advice and purged/updated EasyList.

Comment 1

[MrOtherGuy]

I'm going to guess this is Stylo related because I've experienced something similar but only with Stylo enabled.

Actually I was trying to find a consistent way to reproduce other issue which is reasonably easy to reproduce while interacting Youtube video controls.

The main issue is that when interacting with the video in a way that would cause controls to show the content process locks up and uses about one full core of cpu time, No meaningful increase in RAM though. It's as if main process loses all contact to that process since about:performance doesn't recognize it anymore. There's nothing in browser console either. And it prevents about:debugging from listing any tabs until the process in killed. Closing about:debugging shows a error message in browser console: "error occurred while processing 'listTabs: TypeError: this.conn is null", followed by the stack.

Additionally, main process may still try to put new tabs in that process, which obviously will fail and will result in endless loading.

This doesn't seem to affect the browser in any other way. All tabs in other content processes work fine and main process doesn't seem to suffer too much.


Unfortunately I can't crash the process in such way that would produce a crash report. Windows Taskkill /PID /F will end the process but no crash report is produced.

While trying to find a way to consistently reproduce this I also twice experienced content crashes @nsIFrame::GetDepthInFrameTree - these would happen when on page load but after first paint.

These would be:

https://crash-stats.mozilla.com/report/index/5244d8c1-e040-4b54-8030-6f3bf0170709
and 
https://crash-stats.mozilla.com/report/index/dec730b5-b627-46a7-9103-b90d30170709

Comment 2

[Jet Villegas (inactive)]

Please reply if you're still able to reproduce this crash. Thx!

Comment 3

[smartfon.reddit]

Problem has been fixed.(In reply to Jet Villegas (:jet) from comment #2)
> Please reply if you're still able to reproduce this crash. Thx!

Works for me. Haven't seen this crash for weeks or months.