Description of problem: amarok will not run unless run with LD_ASSUME_KERNEL=2.2.5 (it crashes, backtrace for amarokapp below). with LD_ASSUME_KERNEL=2.2.5, it seems to run normally. Program received signal SIGSEGV, Segmentation fault. ---Type <return> to continue, or q <return> to quit--- [Switching to Thread -1208882720 (LWP 22426)] 0x00000000 in ?? () (gdb) bt #0 0x00000000 in ?? () #1 0x0022b955 in pthread_once () from /lib/libpthread.so.0 #2 0x0048a95e in glXChannelRectSyncSGIX () from /usr/lib/libGL.so.1 #3 0x004bc2b8 in ?? () from /usr/lib/libGL.so.1 #4 0x0048acb0 in _init () from /usr/lib/libGL.so.1 #5 0xbf942908 in ?? () #6 0x00000000 in ?? () (gdb) Version-Release number of selected component (if applicable): 1.3 Additional info: marking this as high severity, as it cannot be run at all without this workaround and that without linuxthreads (i.e.: rawhide and FC5) it won't run at all. is this an OpenGL bug ?
Emily, please fix the Product Version you chose for this ticket, so we know what distribution release you refer to. Also include the RPM package version and release. Amarok has not been included in old fedora.us before FC2. Please include full unedited output of: rpm -qf /usr/lib/libGL.so.1 rpm -q libstdc++ glibc Rest assured, Amarok works just fine for a high number of users.
sorry about the product version and the other missing info. i was in a hurry. thanks. i know i'm not the only one seeing this, though. some people in #fedora were complaining about this very problem, so i installed it for the sole purpose of filing a bug on their behalf. sorry for sounding alarmist. should i reclassify this with a lower priority ? $ rpm -qf /usr/lib/libGL.so.1 xorg-x11-Mesa-libGL-6.8.2-37 $ rpm -q libstdc++ glibc libstdc++-4.0.1-4.fc4 glibc-2.3.5-10 $ rpm -q amarok amarok-1.3-4.fc4
> some people in #fedora were complaining about this very problem That's the wrong place. If it really bugs them, they ought to submit a problem report, preferably also at http:/amarok.kde.org I don't know yet whether Aurelien can reproduce this, but FWIW, I can't. Only crash I've found during the past hours of test-driving Amarok 1.3 is a SIGSEGV in xorg-x11 mga_dri.so when switching through the graphical analysers. Are there any special steps on how to reproduce a crash like yours? Can you reproduce it with amarok-debuginfo package installed, so proper debugging information is included in your backtrace? Your backtrace does not include any symbols from Amarok, so it's not even clear that Amarok is the culprit.
i was unable to locate an amarok-debuginfo of any kind (doing a yum list amarok*). i can tell you that the crash is 100% reproducable on my end, and it appears that the call stack leaves the amarok code well before the crash. it's like a threading error in libGL ? i'm not good enough to figure it out. i wish someone more capable than i could reproduce it. all i had to do was yum install amarok and then try to run it. nothing special. for what it's worth, i'm running nvidia's proprietary drivers. maybe i should give this a try on the x.org's nv driver and see if the same thing happens. okay, here it is ten minutes later (my computer crashed at first, trying to use two drivers at the same time on different displays, then i rebooted and changed drivers the real way), and nope, running with the nv driver makes no difference. i'm out of ideas, but i don't think i'm the only one seeing it.
Something is extremely odd here, because LD_ASSUME_KERNEL=2.2.5 isn't supposed to be functional anymore. LinuxThreads support is no longer in the default library paths. See FC4 release notes. http://fedora.redhat.com/docs/release-notes/fc4/errata/#sn-nptl
isn't linuxthreads still _available_ in FC4 ? i know the situation in rawhide is different. i had rawhide glibc installed, and the LD_ASSUME_KERNEL trick fell flat. downgraded to glibc that came with FC4 and it did indeed let me run the program. does that clarify anything ?
*-debuginfo packages here: http://fedoraproject.org/extras/4/i386/debug/ [...] Re: comment 5 - FC4 still includes the LinuxThreads library paths, also at run-time, see "LD_ASSUME_KERNEL=2.2.5 ldd ...", it's FC5 that will no longer be able to compile for LinuxThreads. And yes, that's what the release notes say. [...] > i'm running nvidia's proprietary drivers I'm not.
I've been running Amarok without problem for the last few days, and I'm using NVidia's binary drivers.
i guess my question is, is anybody actually seeing this crash besides me ? i could swear a lot of people had complained about it in #fedora, but maybe it's some obscure corner case after all. well, i'll get the debuginfo and follow up with a backtrace soon.
the backtrace is identical with debuginfo installed.
Well, then other debuginfo packages are missing for the "in ?? ()" lines. I fail to see how Amarok is involved when the stracktrace does not lead back into Amarok's code.
the only reason amarok is involved is because it somehow causes the crash. it's the only thing i know that can cause the crash. should i do something ? or should this bug be closed ?
Searching Google yields many other problems where glXChannelRectSyncSGIX () is involved, even a crash report at nvidia's forum.
is there a way to use amarok without using /usr/lib/libGL.so.1 ? so that we can maybe narrow it down ? this problem happens whether i use the nv driver or the nvidia driver out of livna, like i said. did any of those google hits yield another workaround ? i guess i'm going to go look myself right after i post this.
Keep this bug open and talk to upstream about the issue. This is probably their responsibility to fix.
to upstream amarok ? i personally don't know if i feel motivated to go so far for a program i don't use... they'll probably just think it's a fedora specific thing and close the bug anyways.
I'm closing this bug, if it still occurs with the latest version of Amarok (currently 1.4.4), feel free to reopen.