You are not logged in.
I've been getting segmentation faults on Firefox 29.0.1 on two computers. It happens as soon as I go to the following address:
https://robertsspaceindustries.com
I see in the console:
"No permission to use the keyboard API for https://robertsspaceindustries.com"
Right before the segmentation fault. It also claims to be dumping a core file but doesn't actually do it... nothing shows up in systemd-coredumpctl either. If I run it through gdb I get the following backtrace:
http://bpaste.net/show/270258/
I've also been able to reproduce this by linking it on IRC and confirming other Firefox users on Arch are seeing it as well. Also of note, it doesn't happen on 29.0 or on aur/firefox-beta-bin, or any other browser I've tried.
Offline
Offline
Same for me, the full output when opening roberstspaceindustrues.com on my system is:
Failed to open VDPAU backend libvdpau_i965.so: Kann die Shared-Object-Datei nicht öffnen: Datei oder Verzeichnis nicht gefunden
No permission to use the keyboard API for https://robertsspaceindustries.com
ATTENTION: default value of option force_s3tc_enable overridden by environment.
Can't find symbol 'glXCreateContextAttribsARB'.
zsh: segmentation fault (core dumped) firefox
Offline
Mine crashes when trying things with WebGL.
ie. http://thp.io/2014/petals/play/ and the webgl aquarium thing.
Started happening after the latest update which included a new kernel and latest mesa (radeon)
Last edited by 89c51 (2014-05-12 16:30:31)
Offline
Just to make sure it is the Firefox package and not any other component that got updated, I downgraded to Firefox 29.0. The old version works without crashing.
Offline
Potentially related to this bug?
no, it's a different bug
I can confirm the segfaults with firefox 29.0.1 with these 2 websites :
https://robertsspaceindustries.com/
http://thp.io/2014/petals/play/
it seem a critical bug because a hacker can deliberately create a website who can exploit this bug, in order to crash firefox
Last edited by Potomac (2014-05-12 22:42:42)
Offline
I visited the sites posted above as "sites that crash FF" but I had not problems whatsoever.
I, however, do not use the FF version packaged by Arch. In fact, there are only two apps. I use straight from the developers: Firefox and Thunderbird.
So ... that makes me think that perhaps there is a problem specifically related to Firefox in Arch and not necessarily to Firefox.
May be try FF from their web site, see if you get a better result?
R.
Offline
I visited the sites posted above as "sites that crash FF" but I had not problems whatsoever.
I, however, do not use the FF version packaged by Arch. In fact, there are only two apps. I use straight from the developers: Firefox and Thunderbird.
check if the version you use is really 29.0.1 and not the aur version
Offline
The output I posted above is a red herring, I get the same output with Ffx 29.0 from Arch and the official Ffx 29.0.1 binaries. Both do not segfault. So it's likely an issue specific to the Arch package of Ffx 29.0.1.
Btw, Potamac has opened a bug report at https://bugs.archlinux.org/task/40331?project=1&cat[0]=2&string=firefox
Offline
As workaround you can set webgl.disabled to true in about config.
Offline
Was Firefox 29.0 (not 29.0.1) built with GCC 4.9? Most other distros haven't moved to it yet, which could explain why nobody else is seeing it.
Offline
I think the real question is whether a proper flyspray has been submitted. If not, please do so as the forums are not a bugtracker.
Offline
@WonderWoofy
Btw, Potamac has opened a bug report at https://bugs.archlinux.org/task/40331?project=1&cat[0]=2&string=firefox
Offline
ralvez wrote:I visited the sites posted above as "sites that crash FF" but I had not problems whatsoever.
I, however, do not use the FF version packaged by Arch. In fact, there are only two apps. I use straight from the developers: Firefox and Thunderbird.check if the version you use is really 29.0.1 and not the aur version
aur ???
Not sure what you mean, I use the download from Mozilla.org therefore the version I have is 29.0.1
R.
Offline
@WonderWoofy
phw wrote:Btw, Potamac has opened a bug report at https://bugs.archlinux.org/task/40331?project=1&cat[0]=2&string=firefox
Excellent!
(And thanks for the proper wrangling Scimmia)
Offline
I just tried to build firefox with clang, and it doesn't have these crashes anymore. Looks like yet another gcc 4.9 problem.
In case anyone wants to try it: http://pkgbuild.com/~fyan/staging/firef … pkg.tar.xz
Offline
I notice that the repo package for firefox is still set to build using clang. Is this strictly still required?
In fact, I meant to ask earlier, but was there a reason that, unlike other packages with GCC 4.9 issues (e.g. flac), firefox was rebuilt with clang instead simply with GCC 4.8?
ArchLinux | x86_64 | linux-ck-ivybridge
ThinkPad X230 | 12.5" | i5-3320M (2.5GHz) | HD 4000 | 16GB (1600MHz) | 256GB mSATA SSD | 2TB HDD
ThinkPad T430 | 14.1" | i7-3520M (2.9GHz) | GF108M (NVS 5400M) | 16GB (1600MHz) | 256GB mSATA SSD | 1TB HDD | 500GB HDD
Offline
In fact, I meant to ask earlier, but was there a reason that, unlike other packages with GCC 4.9 issues (e.g. flac), firefox was rebuilt with clang instead simply with GCC 4.8?
Simple: GCC 4.8 is not in the repo
Offline
Does WebGL work for you in FF 30???
Offline
Does WebGL work for you in FF 30???
Most of WebGL 1.0.3 works, WebGL 2.0 is not working at all for me.
http://www.khronos.org/registry/webgl/s … tests.html
| alias CUTF='LANG=en_XX.UTF-8@POSIX ' |
Offline
aphirst wrote:In fact, I meant to ask earlier, but was there a reason that, unlike other packages with GCC 4.9 issues (e.g. flac), firefox was rebuilt with clang instead simply with GCC 4.8?
Simple: GCC 4.8 is not in the repo
I can't say I follow the reasoning. As this was true for the other package(s) which were rebuilt with 4.8, I can only assume the package maintainers used some prebuilt 4.8 binaries run from some other directory to build the packages in the repo. Though if you're referring to the inability for someone else to build firefox 'simply' from the repo's PKGBUILD, then I think I understand.
Oh well, in any case, I hope this gets fixed soon, so we can get back to GCC.
Last edited by aphirst (2014-06-12 23:36:14)
ArchLinux | x86_64 | linux-ck-ivybridge
ThinkPad X230 | 12.5" | i5-3320M (2.5GHz) | HD 4000 | 16GB (1600MHz) | 256GB mSATA SSD | 2TB HDD
ThinkPad T430 | 14.1" | i7-3520M (2.9GHz) | GF108M (NVS 5400M) | 16GB (1600MHz) | 256GB mSATA SSD | 1TB HDD | 500GB HDD
Offline
Actually I don't see your problem as a package consumer. This is primarily a build issue, and I can understand that the maintainer does not want to build with an older compiler than what is currently the standard in Arch.
Offline
89c51 wrote:Does WebGL work for you in FF 30???
Most of WebGL 1.0.3 works, WebGL 2.0 is not working at all for me.
http://www.khronos.org/registry/webgl/s … tests.html
This browser does not appear to support WebGL
WTF??? This is what i get from your link.
I tested with https://webglsamples.googlecode.com/hg/ … arium.html and doesn't work.
radeon FOSS drivers if it matters.
Offline
I tested with https://webglsamples.googlecode.com/hg/ … arium.html and doesn't work.
radeon FOSS drivers if it matters.
Then try http://get.webgl.org/ It should provide you with some info on how to enable webgl.
| alias CUTF='LANG=en_XX.UTF-8@POSIX ' |
Offline
89c51 wrote:I tested with https://webglsamples.googlecode.com/hg/ … arium.html and doesn't work.
radeon FOSS drivers if it matters.
Then try http://get.webgl.org/ It should provide you with some info on how to enable webgl.
I didn't do anything (except upgrade to the latest stuff from the official repo) and it used to work.
webgl.disabled is false in the about:config. glxinfo tells me that direct rendering works.
Offline