| Summary: | www-cilent/firefox-42[system-cairo]: Crash when opening a new tab. | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Jeremy Murphy <jeremy.william.murphy> |
| Component: | Current packages | Assignee: | Mozilla Gentoo Team <mozilla> |
| Status: | RESOLVED OBSOLETE | ||
| Severity: | normal | CC: | aoaaxy+gentoobugzilla, mads, zazdxscf+bugs.gentoo.org |
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | AMD64 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: |
# emerge --info -v
gdb backtrace of firefox crashing on exit steps I tried and emerge --info here's the diff between 4.2.6 and 4.2.6-r1 kernels' source code kernel .config |
||
|
Description
Jeremy Murphy
2015-11-05 02:12:44 UTC
Please rebuild with -system-cairo, I believe we are gonna drop the useflag and force everyone back to bundled for the time being until we can sort out cairo-1.14.x breakage. Created attachment 416842 [details]
# emerge --info -v
Hi. For me it doesn't crash (built with -system-cairo already) but only the first loaded tab (eg. on firefox startup, first loaded tab(the focused one) of those auto-restored) loads okay. All others are hanging on "Waiting for ..." as if the network is slow.
The issue happens only with hardened-sources-4.2.6 but not with (tested) hardened-sources-4.2.5-r2 (and firefox 42.0)
Firefox does crash on exit though. I could include a full all threads stacktrace. I have one from yesterday, but I should probably regen a new one, next. (I am currently writing this from 4.2.5-r2 kernel, or else I couldn't have logged in, even if I was using only the loaded tab, with 4.2.6 kernel and firefox 42.0)
Once I was able to open two new windows(on firefox startup) and both loaded okay - they were of bugs.gentoo.com but log in was impossible after this due to the same issue: hung on "Waiting for..."
What other info should I give?
Created attachment 416850 [details]
gdb backtrace of firefox crashing on exit
(pls ignore the "-g45b4b78" kernel local version which I forgot in .config for every kernel that I've had since I've switched from git kernel to hardened.)
Sorry for the possible hijacking (since I haven't tested with system-cairo to reproduce the initial). Let me know if I should file a new bug instead (which I probably won't, 'cause I dislike the rigmarole, but I will happily provide more information should the bug already exist filed - I will test any patches happily; just don't make me file a NEW bug :D - hate that process)
(In reply to Emanuel Czirai from comment #3) > Created attachment 416850 [details] > gdb backtrace of firefox crashing on exit > > (pls ignore the "-g45b4b78" kernel local version which I forgot in .config > for every kernel that I've had since I've switched from git kernel to > hardened.) > > Sorry for the possible hijacking (since I haven't tested with system-cairo > to reproduce the initial). Let me know if I should file a new bug instead > (which I probably won't, 'cause I dislike the rigmarole, but I will happily > provide more information should the bug already exist filed - I will test > any patches happily; just don't make me file a NEW bug :D - hate that > process) If you want to test with system-cairo please use the mozilla overlay for testing. I am getting ready to push firefox-42.0-r1 to tree in next few days so more testing is always welcomed. (In reply to Jory A. Pratt from comment #4) > If you want to test with system-cairo please use the mozilla overlay for > testing. I am getting ready to push firefox-42.0-r1 to tree in next few days > so more testing is always welcomed. emerge cannot find firefox-42.0-patches-0.2.tar.xz on a bunch of mirrors that I had in the list. Is this what you meant by not being in the tree? Where can I wget it directly? (apparently it's not in the usual location that I check when firefox isn't on any of the mirrors, which is https://ftp.mozilla.org/pub/mozilla.org/firefox/releases/latest/source/ ) I would like to otherwise test this with -egl and system-cairo (also xcb had to be enabled on cairo) Thanks (In reply to Emanuel Czirai from comment #5) > (In reply to Jory A. Pratt from comment #4) > > If you want to test with system-cairo please use the mozilla overlay for > > testing. I am getting ready to push firefox-42.0-r1 to tree in next few days > > so more testing is always welcomed. > > emerge cannot find firefox-42.0-patches-0.2.tar.xz on a bunch of mirrors > that I had in the list. Is this what you meant by not being in the tree? > Where can I wget it directly? (apparently it's not in the usual location > that I check when firefox isn't on any of the mirrors, which is > https://ftp.mozilla.org/pub/mozilla.org/firefox/releases/latest/source/ ) > > I would like to otherwise test this with -egl and system-cairo (also xcb had > to be enabled on cairo) > > Thanks Patches belong to gentoo not upstream. You can find them via my devspace. http://dev.gentoo.org/~anarchy/mozilla/patchsets/firefox-42.0-patches-0.2.tar.xz (In reply to Jory A. Pratt from comment #1) > Please rebuild with -system-cairo, I believe we are gonna drop the useflag > and force everyone back to bundled for the time being until we can sort out > cairo-1.14.x breakage. Thanks, I meant to say earlier that I rebuilt with -system-cairo and it stopped crashing. (In reply to Jeremy Murphy from comment #7) > (In reply to Jory A. Pratt from comment #1) > > Please rebuild with -system-cairo, I believe we are gonna drop the useflag > > and force everyone back to bundled for the time being until we can sort out > > cairo-1.14.x breakage. > > Thanks, I meant to say earlier that I rebuilt with -system-cairo and it > stopped crashing. We have made changes to the eclass in the mozilla overlay which should address the crashing issues. It also will introduce hardware acceleration. You might want to check it out as soon as possible as we will push the new update on sunday. Created attachment 416900 [details]
steps I tried and emerge --info
tl;dr: Issue went away for me, with hardened kernel 4.2.6-r1 (but still couldn't repro. OP's issue with or without system-cairo)
Also, maybe until sunday, for me, -system-cairo without hw accel. is faster (16fps) than with hw. accel (9fps) and than system-cairo (5-9 fps, I forget) - url for testing fps is below.
But here's what I had tried before:
(too long for comment, attached)
Created attachment 416952 [details, diff]
here's the diff between 4.2.6 and 4.2.6-r1 kernels' source code
just for completeness/curiosity
Created attachment 416956 [details] kernel .config uploading kernel .config (for completeness) (I didn't have KVM selected, but do have r8169) Whatever fixed it for me, must be in the above kernels' source code difference :-) Anyway, thanks to whoever fixed it! (fix was within genpatches?) $ equery c hardened-sources-4.2.6-r1 *hardened-sources-4.2.6-r1 (13 Nov 2015) 13 Nov 2015; Anthony G. Basile <blueness@gentoo.org> +hardened-sources-4.2.6-r1.ebuild: version bump to 4.2.6-r1 vanilla-4.2.6 + genpatches-4.2-8 + grsecurity-3.1-4.2.6-201511122040 Package-Manager: portage-2.2.20.1 13 Nov 2015; Anthony G. Basile <blueness@gentoo.org> -hardened-sources-4.2.5-r1.ebuild, -hardened-sources-4.2.5-r2.ebuild, -hardened-sources-4.2.6.ebuild: remove older unstable versions. Package-Manager: portage-2.2.20.1 This is it: https://bugzilla.mozilla.org/show_bug.cgi?id=1136417 There is no need to use bundled cairo if cairo is compiled without xlib-xcb. May I propose a change in mozconfig-*.eclass: -system-cairo? ( >=x11-libs/cairo-1.12[X,xcb] +system-cairo? ( >=x11-libs/cairo-1.12[X,xcb,-xlib-xcb] Works for me. (In reply to Jan Psota from comment #12) > This is it: > https://bugzilla.mozilla.org/show_bug.cgi?id=1136417 > > There is no need to use bundled cairo if cairo is compiled without xlib-xcb. > May I propose a change in mozconfig-*.eclass: > -system-cairo? ( >=x11-libs/cairo-1.12[X,xcb] > +system-cairo? ( >=x11-libs/cairo-1.12[X,xcb,-xlib-xcb] > > Works for me. Well this would be one small fix, it still does not resolve the crashes completely. (In reply to Jory A. Pratt from comment #13) > Well this would be one small fix, it still does not resolve the crashes > completely. FF on my system crashed many times before and I've noticed no crashes after cairo recompilation with -xcb-xlib - that's all I can say. This whole problem more seems like something to be handled with by upstream, just my 2 cents. (In reply to mike from comment #15) > This whole problem more seems like something to be handled with by upstream, > just my 2 cents. Upstream doesn't support external cairo; if the community doesn't fix it then it doesn't get fixed. As most of you are likely aware, I've package.use.masked system-cairo to try and limit exposure to these crashes. I'm certainly open to patches to get it working better but it's beyond mozilla team's abilities and time to fix it ourselves. If firefox bumps their bundled custom cairo then maybe we can revisit. If you feel I have closed your bug and it is still a current issue, please reopen and update it completely. We will not work bugs that have no ebuild in tree any longer or can not be reproduced with a current system. Thank You for your support and understanding The Mozilla Team |
