Bug 184574 – [GTK] Webkit 2.20.x crashes frequently on Youtube.com when changing resolution during video playback, and sometimes while clicking the link to another video.

• Component: WebKitGTK
• Status: RESOLVED MOVED

[Ryan Farmer]

While I was testing to see if the distorted video rendering related to bug 173796 was still an issue, I noticed that while Youtube videos now play fine, at least initially, the Webkit Web Process crashes frequently if you change the resolution during playback and (occasionally) if you click on another video while one is playing.

Tested on Fedora 28 with the Gstreamer VA-API plugin and the Intel VA driver installed on Iris 540 graphics on a Skylake Core i7-6560U.

[Michael Catanzaro]

Please post a backtrace

[Hyunjun Ko]

(In reply to Ryan Farmer from comment #0)
> While I was testing to see if the distorted video rendering related to bug
> 173796 was still an issue, I noticed that while Youtube videos now play
> fine, at least initially, the Webkit Web Process crashes frequently if you
> change the resolution during playback and (occasionally) if you click on
> another video while one is playing.
> 
> Tested on Fedora 28 with the Gstreamer VA-API plugin and the Intel VA driver
> installed on Iris 540 graphics on a Skylake Core i7-6560U.

Could you let us know which version of gstreamer you're using?

[Hyunjun Ko]

I found an issue in gst-vaapi, i'm still not sure if that is same as this issue.
If you're using gst-vaapi 1.14, it's likely to be same.

[Ryan Farmer]

Yes, it's Gstreamer 1.14. 

Sorry for the delay. The bug mail was getting sent into my spam folder.

[Michael Catanzaro]

Backtrace still needed :)

[Ryan Farmer]

Created attachment 339047 [details]
Stack Trace

[Ryan Farmer]

I can't get ABRT to come up anymore when this happens, but this is what it leaves in my system log. 

https://www.youtube.com/watch?v=07G4xhSefuI

There's a video that does it every time. Just click the gear icon and change the resolution to 720p.

[Michael Catanzaro]

Hi, please see https://wiki.gnome.org/Community/GettingInTouch/Bugzilla/GettingTraces for instructions on getting a higher-quality backtrace. Be sure to install debuginfo for all the libraries that appear in the backtrace, and take the backtrace in gdb using 'thread apply all bt full'.

[Michael Catanzaro]

(In reply to Ryan Farmer from comment #7)
> I can't get ABRT to come up anymore when this happens, but this is what it
> leaves in my system log. 

Ah, sorry. Fedora makes this easy: just run 'coredumpctl gdb'. Then quit gdb and run the dnf debuginfo-install command it gives you. Then 'coredumpctl gdb' again,  'thread apply all bt full'.

[Ryan Farmer]

Created attachment 339049 [details]
GDB Stack Trace

[Ryan Farmer]

I hope that's better. I had to install a _lot_ of debuginfo packages to get it. 

It still complained that a few were needed, but when I went to install the last batch, dnf says I already have them. :/

[Michael Catanzaro]

(In reply to Ryan Farmer from comment #11)
> I hope that's better. I had to install a _lot_ of debuginfo packages to get
> it. 
> 
> It still complained that a few were needed, but when I went to install the
> last batch, dnf says I already have them. :/

Don't worry, what you posted there is great, thanks. That should be enough for Hyunjun to figure out if he's seeing the same bug as you.

I'll also CC Victor since this appears to be a bug in gst-vaapi.

[Hyunjun Ko]

(In reply to Ryan Farmer from comment #10)
> Created attachment 339049 [details]
> GDB Stack Trace

This is same as mine :)

I'm working on this, refer to the issue https://bugzilla.gnome.org/show_bug.cgi?id=795391.

[Michael Catanzaro]

I guess it's sufficient to track this on GNOME Bugzilla, right? Closing this, then. Thanks Hyunjun!

[Hyunjun Ko]

(In reply to Michael Catanzaro from comment #14)
> I guess it's sufficient to track this on GNOME Bugzilla, right? Closing
> this, then. Thanks Hyunjun!

That's fine. 

@Ryan, for more information, you can test with patches on https://bugzilla.gnome.org/show_bug.cgi?id=793643. (on the master now)

[Víctor M. Jáquez L.]

(In reply to Hyunjun Ko from comment #15)
> (In reply to Michael Catanzaro from comment #14)
> > I guess it's sufficient to track this on GNOME Bugzilla, right? Closing
> > this, then. Thanks Hyunjun!
> 
> That's fine. 
> 
> @Ryan, for more information, you can test with patches on
> https://bugzilla.gnome.org/show_bug.cgi?id=793643. (on the master now)

Also they are merged in the branch 1.14

[Ryan Farmer]

Interestingly, I can't make this crash happen no matter what I do. 

Something in WebkitGTK 2.20.2 appears to have fixed it?

[Michael Catanzaro]

*** Bug 185642 has been marked as a duplicate of this bug. ***