Created attachment 3998 [details] Console O/P of the 5 crashing programs I have worked in QA on various Mag3 betas, release candidate and pre-releases. I talk of KDE-only CD, KDE-only DVD, generic DVD - all 32-bit. I have found a common problem on various KDE installs (either unique, or as part of several desktops) that some applications crash on launching *on my system*. No other QA team member could re-produce this. Note that the 'weakness' of my box is its SiS 660 on-board video. Because of a different problem I had to use 'xdriver=vesa' kernel parameter to get these systems installed & running; so this problem should not be SiS related. The attached evidence is from an 18th May (just before Mag3 release) KDE English only Live CD installation. The programs are: Amarok: after its splash screen. Konqueror: works/worked on some installs, crashes/crashed on others. KMail Kontact Skrooge I attach the output from launching them from the console; and screenshots of the KDE Crash Handler.
Created attachment 3999 [details] Amarok KDE crash dialogue
CC: (none) => lewyssmith
Cannot add the other 4 KDE crash dialogues; am repeatedly bounced by the system re cookies.
Created attachment 4001 [details] KMail KDE crash dialogue
Created attachment 4002 [details] Konqueror KDE crash dialogue
Created attachment 4003 [details] Kontact KDE crash dialogue
Created attachment 4004 [details] Skrooge KDE crash dialogue
I have the same problem with a new install of 32 bit Mageia 3 (release version) on an AMD machine which has worked fine with previous versions. The following programs crash every time I try to start them. This is with a new user account, so there are no old settings in config files etc. Amarok (after splash screen) KMail Kontact Skrooge. Konqueror works OK.
CC: (none) => mag.4.shrdlu
Created attachment 4012 [details] Backtrace of Kontact crash
Created attachment 4013 [details] Backtrace of KMail crash
CC: (none) => nicolas.lecureuil, tmb
Created attachment 4014 [details] Backtrace of Amarok crash
please install the debuginfo packages and paste new backtraces.
Created attachment 4015 [details] Backtrace of Skrooge crash
The backtraces for each application crash are the same down to the line marked #14 May indicate that the problem is in libQtWebKit.
yes seems the same pb everywhere. Please install the debuginfo packages and paste new backtraces.
Created attachment 4016 [details] Backtrace of Kmail crash after installing kdelibs4-debuginfo and qtwebkit-debuginfo Information about libQtWebKit is still missing from the trace. Are there other debuginfo packages I could install to show this ?
Attachment 4013 is obsolete: 0 => 1
Attachment 3999 is obsolete: 0 => 1CC: (none) => thierry.vignaud
Attachment 4001 is obsolete: 0 => 1
Attachment 4002 is obsolete: 0 => 1
Attachment 4003 is obsolete: 0 => 1
Attachment 4004 is obsolete: 0 => 1
What's the output of "rpm -q libqtwebkit2.2_4 qtwebkit-debuginfo" ?
Keywords: (none) => NEEDINFO
$ rpm -q libqtwebkit2.2_4 qtwebkit-debuginfo libqtwebkit2.2_4-2.3.0-1.mga3 qtwebkit-debuginfo-2.3.0-1.mga3 $
Created attachment 4017 [details] AMD CPU Info Is the current version of QtWebKit compiled for SSE2 ? This CPU doesn't support it (see attachment). See https://bugs.archlinux.org/task/33418 If so, is there an earlier version or alternative build of QtWebKit I could use on this machine ?
Keywords: NEEDINFO => (none)
maybe by using %ifarch %{ix86} --no-force-sse2 %endif i will try to push an update package asap
Comment on Comment 7: > on an AMD machine which has worked fine with previous versions *My* box is AMD (Sempron 2500, really 1.7MHz). Could the problem lie there?
FWIW I have Suse 12.2 (fully up-to-date) with KDE on the same machine, and the comparative crashes are: - Konqueor crashes! - Kontact does NOT - KMail does NOT - Amarok does NOT Skrooge I cannot find to install (it demands too recent dependancies). BTW I am willing to help if you ask...
Lewis, what is the output of "cat /proc/cpuinfo" on your system ? Could you attach it here ? If the "flags" part doesn't list "sse2", that would be further evidence that QtWebKit in Mageia 3 has been compiled allowing SSE2 instructions, which earlier cpus don't support. Nicolas's update should fix that. It's possible that in SUSE, QtWebKit has not been compiled for SSE2, so you don't see crashes in Kontact etc. However, since Konqueror does crash, perhaps one of the other libraries it uses *has* been compiled for SSE2. This is all a guess though ;-)
cat /proc/cpuinfo processor : 0 vendor_id : AuthenticAMD cpu family : 6 model : 8 model name : AMD Sempron(tm) 2500+ stepping : 1 cpu MHz : 1761.154 cache size : 256 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 1 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 mmx fxsr sse syscall mp mmxext 3dnowext 3dnow bogomips : 3522.30 clflush size : 32 cache_alignment : 32 address sizes : 34 bits physical, 32 bits virtual power management: ts As you suspected, no sse2. Well done!
Please test the new package in updates_testing
Installed new qtwebkit package from Core Updates Testing. $ rpm -q libqtwebkit2.2_4 libqtwebkit2.2_4-2.3.0-1.1.mga3 $ Amarok, Kontact, Kmail, Skrooge, and Konqueror now work fine. So the bug is fixed, at least for my system. Thank you Nicolas :-)
CC: thierry.vignaud => (none)Summary: Some KDE applications crash on startup on Mageia 3 32-bit => Some KDE applications crash on startup on Mageia 3 32-bit die to qtwebkit using sse2
My thanks to you both. I will test this myself as soon as I can; thanks to Keith for the vital commands! This issue gave me a bad name in QA (happened with me only, my h/w not suitable for Linux etc) so I am pleased with its confirmation & speedy resolution. But this 'sse2' issue could be widespread in compiled packages & affect a lot of older processors. Should they be banned? It is actually something the installers could look for & warn about.
I have installed the updated libqtwebkit2.2_4-2.3.0-1.1.mga3 32-bit (i586) on publicly released Mageia 3 - where initially all except Konqueror had crashed as described. For me, it works! All of the cited programs start correctly. I think this can be closed, resolved. And 'pushed'. Repeat my thanks to all concerned.
For the record Source RPM : qtwebkit-2.3.0-1.1.mga3.src.rpm
Source RPM: (none) => qtwebkit-2.3.0-1.1.mga3.src.rpmWhiteboard: (none) => MGA3-32-OK
*** Bug 10252 has been marked as a duplicate of this bug. ***
CC: (none) => derekjenn
Confirmed fixed. Update validated Could sysadmin please push qtwebkit-2.3.0-1.1.mga3.src.rpm from Core Updates Testing to Core Updates Advisory -------- This updates fixes crashes in various applications that depend on libqtwebkit when they run on older processors.
Keywords: (none) => validated_updateCC: (none) => sysadmin-bugs
Derek, on which arch did you test it? I can see only MGA3-32-OK on whiteboard. Removing validation until confirmed that it's tested on both arch.
Keywords: validated_update => (none)CC: (none) => sander.lepikAssignee: bugsquad => qa-bugs
Sander. Just as well you delayed pushing. I have since tested it again with qupzilla on i586 and found it still crashes. Not much information from dbg other than it is failing with 'Illegal Instruction' in libQtWebKit.so.4 konqueror is failing too also with Illegal Instruction while rendering a web page. Testing was on a PIII computer.
Whiteboard: MGA3-32-OK => MGA3-32-OK feedback
Anything useful in ~/.xsession-errors Derek?
Re comment 31 Surely this problem can only apply to 32-bit. Any 64-bit processor would (I guess) certainly have the necessary sse2 etc instructions and not require this update. Re comment 32 Derek: did these browser crashes happen before the update also? Are these Qupzilla & Konqueror failures the same problem? The original bug [title] was unambiguous, & clearly resolved by the update. I will try & use Konqueror to see whether it crashe in use. FWIW with the update I have: /usr/lib/libQtWebKit.so.4.10.0
(In reply to Lewis Smith from comment #34) > Re comment 31 > Surely this problem can only apply to 32-bit. Any 64-bit processor would (I > guess) certainly have the necessary sse2 etc instructions and not require > this update. Well, if you rebuild the SRPM then there will be new version for 64-bit systems too. And you have to test if it works as before or not..
Testing complete for qtwebkit-2.3.0-1.1.mga3.src.rpm on Mageia release 3 (Official) for x86_64, it's Ok for me.
CC: (none) => geiger.david68210Whiteboard: MGA3-32-OK feedback => MGA3-32-OK feedback MGA3-64-OK
Update validated. Thanks. Advisory: This updates fixes crashes in various applications that depend on libqtwebkit when they run on older processors. SRPM: qtwebkit-2.3.0-1.1.mga3.src.rpm Could sysadmin please push from core/updates_testing to core/updates. Thank you!
Keywords: (none) => validated_update
(In reply to David GEIGER from comment #36) > Testing complete for qtwebkit-2.3.0-1.1.mga3.src.rpm on Mageia release 3 > (Official) for x86_64, it's Ok for me. Did you also have the updates testing of kde installed, while testing this update? If so, it might be better to hold this update, till the kde4 update is validated.
CC: (none) => davidwhodgins
http://svnweb.mageia.org/advisories/10157.adv?view=markup&sortby=date added. Removing validated_update keyword, till comment 38 is answered.
Keywords: validated_update => (none)
Dave No, I did *not* have the KDE update installed (it was not available then). Am happy to do that and re-confirm this update.
Best not to Lewis. It's not yet assigned to us and could still have issues. We need to make sure this updates/works with the current version. That's what Dave was meaning. David G that was your one. Could you just confirm please whether you had current KDE or the one from Testing. Thanks
why hold this until kde update is pushed ? this is not related at all
(In reply to Nicolas Lécureuil from comment #42) > why hold this until kde update is pushed ? this is not related at all I just wanted to ensure it had been tested with the current version of kde. I've seen problems before, when an update passed testing, because another package was in updates testing, but when it was pushed to updates, there were problems, until the other package was also pushed to updates. Anyway, I've gone ahead and tested, to ensure it works with the current kde version, on x86_64, so re-validating the update. Could someone from the sysadmin team push 10157.adv.
http://advisories.mageia.org/MGAA-2013-0043.html
Status: NEW => RESOLVEDCC: (none) => boklmResolution: (none) => FIXED
CC: boklm => (none)