848368 - pages not loading and tabs not changing

Status: RESOLVED WORKSFORME

(Core :: Graphics, defect)

Description

[Paul [pwd]]

User Agent: Mozilla/5.0 (Windows NT 6.1; rv:22.0) Gecko/20130305 Firefox/22.0
Build ID: 20130305030833

Steps to reproduce:

I'm unable to pin down exactly what's causing this, but it's a real issue of which I run into a couple times a day. I attempted to provide relevant logcats in bug 845795 although they don't seem to have helped. That said, I feel that bug 845795 was prematurely marked as invalid and hope that won't happen with this one.

Comment 1

[Paul [pwd]]

Attachment: another LogCat taken after 'stuck' viewport

Comment 3

[Kartikaya Gupta (email:[e-mail])]

Unfortunately the log doesn't have a lot of useful information (mostly because we don't log things that are relevant to this sort of bug). I created a patch which adds some more relevant logging and pushed it to try to build it. If you have some time please use that build and when you see the problem, grab the logcat and attach it here.

Try build at https://tbpl.mozilla.org/?tree=Try&rev=bb74508309f3

Comment 4

[Kevin Brosnan [Ex-Mozilla]]

The unable to switch tabs is likely bug 844526

Comment 5

[Kartikaya Gupta (email:[e-mail])]

I think that's a different bug.

Comment 6

[Paul [pwd]]

(In reply to Kartikaya Gupta (email:[e-mail]) from comment #3)
> Unfortunately the log doesn't have a lot of useful information (mostly
> because we don't log things that are relevant to this sort of bug). I
> created a patch which adds some more relevant logging and pushed it to try
> to build it. If you have some time please use that build and when you see
> the problem, grab the logcat and attach it here.
> 
> Try build at https://tbpl.mozilla.org/?tree=Try&rev=bb74508309f3

Just to confirm, we're talking about this build right? http://ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/[e-mail]-bb74508309f3/try-android/

Comment 7

[Kartikaya Gupta (email:[e-mail])]

Assuming you're using an ARMv7 device, yes. If you're using an ARMv6 device it's the build directory at http://ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/[e-mail]-bb74508309f3/try-android-armv6/

Comment 8

[Paul [pwd]]

Attachment: Stuck viewport with kats try build (1)

Comment 9

[Paul [pwd]]

Attachment: Stuck viewport with kats try build (2)

Comment 10

[Paul [pwd]]

Attachment: Stuck viewport with kats try build (3)

Comment 11

[Kartikaya Gupta (email:[e-mail])]

(In reply to Paul [sabret00the] from comment #8)
> Created attachment 722743 [details]
> Stuck viewport with kats try build (1)

In this log I see bug 846774 show up at 22:13:51 when the awesome screen is entered, followed by some EGL errors at 22:14:12. Then there's a bunch of tab switching at 22:14:26 - 22:14:34 where it seems like each tab switch gets painted and composited just fine, although if the GL errors were a problem then that might cause the composites to fail here. FF is put into the background at 22:14:37 and there's nothing else interesting after that.

(In reply to Paul [sabret00the] from comment #9)
> Created attachment 722744 [details]
> Stuck viewport with kats try build (2)

In this one FF starts up at 07:47:07 and creates the compositor around 07:47:13, no problems there. Awesome screen comes up at 07:47:17, the GL surface is destroyed at 07:47:18.443 and then at 07:47:19.493 the GL surface is up again (presumably the awesome screen was dismissed) but then we get EGL_BAD_ALLOC errors from eglMakeCurrent without any indication why. From 07:47:22 to 07:47:55 there's a bunch of tab switching/painting/compositing that seems fine but again may not composite properly because of the previous error. FF goes in the background at around 07:47:57, and that's the end of Act II.

(In reply to Paul [sabret00the] from comment #10)
> Created attachment 722745 [details]
> Stuck viewport with kats try build (3)

This one is fine up until 08:49:07 when for no apparent reason we get the EGL_BAD_ALLOC error while handling a surface change notification. It's not until 08:50:29 that a couple of tab switches happen. Same as before.

So in a nutshell this seems to happen as a result of EGL_BAD_ALLOC errors but I'm not sure what the root cause is here.

Comment 12

[Paul [pwd]]

So what can I do to further help identify the issue?

Comment 13

[Kartikaya Gupta (email:[e-mail])]

Sorry for the delay. I'm not really sure what we can do here, if anything. According to the docs at http://www.khronos.org/registry/egl/sdk/docs/man/xhtml/eglMakeCurrent.html an EGL_BAD_ALLOC is returned basically if there aren't enough resources left, so this seems like a graphics driver problem.

Moving to core graphics as they have more experience in this area.