[Bug 1140184] New: Falkon segfaults at exit, could be mesa related?
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 Bug ID: 1140184 Summary: Falkon segfaults at exit, could be mesa related? Classification: openSUSE Product: openSUSE Distribution Version: Leap 15.1 Hardware: Other OS: Other Status: NEW Severity: Normal Priority: P5 - None Component: KDE Applications Assignee: opensuse-kde-bugs@opensuse.org Reporter: stakanov@freenet.de QA Contact: qa-bugs@suse.de Found By: --- Blocker: --- Created attachment 809232 --> http://bugzilla.opensuse.org/attachment.cgi?id=809232&action=edit backtrace of the segfault, required libraries of falcon and mesa falkon did crash at exit with segfault. The installation of debug libraries does reveal a connection to Mesa (this is why I post it, konqi did not give an address to post so maybe this is not upstream. Feel free to point me upstream if applicable). The backtrace is joint with the attachment. Repeatable: well it crashed before, but is irregular, does not happen always. And this is the first time I get a usable backtrace. What the application does: segfaults What the application should do: should close without segfault. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c1 Wolfgang Bauer <wbauer@tmo.at> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |wbauer@tmo.at Component|KDE Applications |X.Org Assignee|opensuse-kde-bugs@opensuse. |xorg-maintainer-bugs@forge. |org |provo.novell.com QA Contact|qa-bugs@suse.de |xorg-maintainer-bugs@forge. | |provo.novell.com --- Comment #1 from Wolfgang Bauer <wbauer@tmo.at> --- (In reply to Stakanov Schufter from comment #0)
Created attachment 809232 [details] backtrace of the segfault, required libraries of falcon and mesa
falkon did crash at exit with segfault. The installation of debug libraries does reveal a connection to Mesa (this is why I post it, konqi did not give an address to post so maybe this is not upstream. Feel free to point me upstream if applicable). Mesa is maintained by the X.Org team, reassigning.
FTR, it doesn't crash here, but I'm currently using Falkon 3.1.0 with Qt 5.13 on Leap 15.1 (with the default Mesa though, but an intel chip) I suppose setting the environment variable QT_WEBENGINE_DISABLE_GPU=1 (to force QtWebEngine to use software rendering instead) would help? E.g. run "QT_WEBENGINE_DISABLE_GPU=1 falkon", or add "QT_WEBENGINE_DISABLE_GPU=1" to /etc/environment and logout/login. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c2 Stefan Dirsch <sndirsch@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Priority|P5 - None |P3 - Medium Status|NEW |IN_PROGRESS CC| |stakanov@freenet.de Flags| |needinfo?(stakanov@freenet. | |de) --- Comment #2 from Stefan Dirsch <sndirsch@suse.com> --- Hmm. Can't reproduce this on my RADEON(0): Chipset: "OLAND" (ChipID = 0x6611) on my Leap 15.1 installation - using radeon driver. falcon starts and can be terminated without a segfault. Which Radeon GPU are you using? Could you attach your /var/log/Xorg.0.log, please? -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c3 --- Comment #3 from Stakanov Schufter <stakanov@freenet.de> --- I have an intel chipset. However, if I uninstall the mesa driver on my system, the dependency break the x-server. Several games and also the xserverpackage itself seem to require the presence of mesa drivers (Gallium) on my system with i945 Intel graphics (this is a Lenovo X201). I will post var/log/Xorg.0.log right now as an attachment. Do you want also dmidecode? -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c4 Stakanov Schufter <stakanov@freenet.de> changed: What |Removed |Added ---------------------------------------------------------------------------- Flags|needinfo?(stakanov@freenet. | |de) | --- Comment #4 from Stakanov Schufter <stakanov@freenet.de> --- Created attachment 809296 --> http://bugzilla.opensuse.org/attachment.cgi?id=809296&action=edit Xorg.0.log as requested With this you should have what you asked for so I take off need info. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c5 --- Comment #5 from Stakanov Schufter <stakanov@freenet.de> --- (In reply to Wolfgang Bauer from comment #1)
(In reply to Stakanov Schufter from comment #0)
Created attachment 809232 [details] backtrace of the segfault, required libraries of falcon and mesa
falkon did crash at exit with segfault. The installation of debug libraries does reveal a connection to Mesa (this is why I post it, konqi did not give an address to post so maybe this is not upstream. Feel free to point me upstream if applicable). Mesa is maintained by the X.Org team, reassigning.
FTR, it doesn't crash here, but I'm currently using Falkon 3.1.0 with Qt 5.13 on Leap 15.1 (with the default Mesa though, but an intel chip)
I suppose setting the environment variable QT_WEBENGINE_DISABLE_GPU=1 (to force QtWebEngine to use software rendering instead) would help? E.g. run "QT_WEBENGINE_DISABLE_GPU=1 falkon", or add "QT_WEBENGINE_DISABLE_GPU=1" to /etc/environment and logout/login.
I did try what you suggested. Substantially for now I see no difference (also because it is difficult as it does not crash so often, only "from time to time"). What triggered me to report was the fact that it wanted mesa debug libraries. I can set this as command of the start up of Falkon and see if on the long run it reproduces...or not. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 Stefan Dirsch <sndirsch@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #809296|text/x-log |text/plain mime type| | -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c6 --- Comment #6 from Stefan Dirsch <sndirsch@suse.com> --- Ok. Seems _mesa_delete_ati_fragment_shader crashes on an Intel Ironlake GPU (8086:0046). Weird. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c7 Stefan Dirsch <sndirsch@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Flags| |needinfo?(stakanov@freenet. | |de) --- Comment #7 from Stefan Dirsch <sndirsch@suse.com> --- Does it help to run falkon without GL_ATI_fragment_shader extension? MESA_EXTENSION_OVERRIDE="-GL_ATI_fragment_shader" falkon -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c8 Stakanov Schufter <stakanov@freenet.de> changed: What |Removed |Added ---------------------------------------------------------------------------- Flags|needinfo?(stakanov@freenet. | |de) | --- Comment #8 from Stakanov Schufter <stakanov@freenet.de> --- (In reply to Stefan Dirsch from comment #7)
Does it help to run falkon without GL_ATI_fragment_shader extension?
MESA_EXTENSION_OVERRIDE="-GL_ATI_fragment_shader" falkon
Well, as I said I cannot guarantee for the stability as crashes are irregular. With these settings I did not have a crash and what is weird the pages load as it seems faster. Actually much faster. Without there is always a stop of loading of about 1 second. With the setting the loading is immediate. That does of course not guarantee that this solves the problem but I like that it loads faster to tell the truth. Where can I force it to do this as "default" start settings, so to check if all is ok on the long run? -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c9 --- Comment #9 from Wolfgang Bauer <wbauer@tmo.at> --- (In reply to Stakanov Schufter from comment #8)
Where can I force it to do this as "default" start settings, so to check if all is ok on the long run? As I wrote earlier, you could add it to /etc/environment e.g., i.e. edit the file in a text editor (as root) and in this case add the line: MESA_EXTENSION_OVERRIDE="-GL_ATI_fragment_shader"
Or maybe add this line to ~/.profile: export MESA_EXTENSION_OVERRIDE="-GL_ATI_fragment_shader" (it will only affect your user account then) -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c10 --- Comment #10 from Stakanov Schufter <stakanov@freenet.de> --- (In reply to Wolfgang Bauer from comment #9)
(In reply to Stakanov Schufter from comment #8)
Where can I force it to do this as "default" start settings, so to check if all is ok on the long run? As I wrote earlier, you could add it to /etc/environment e.g., i.e. edit the file in a text editor (as root) and in this case add the line: MESA_EXTENSION_OVERRIDE="-GL_ATI_fragment_shader"
Or maybe add this line to ~/.profile: export MESA_EXTENSION_OVERRIDE="-GL_ATI_fragment_shader" (it will only affect your user account then)
Well, done. If I do not complain it simply works. If not you are going have me reporting here again with backtrace. Thank you for the fast reply. Much appreciated. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c11 --- Comment #11 from Wolfgang Bauer <wbauer@tmo.at> --- Just to be sure: you shouldn't be using QT_WEBENGINE_DISABLE_GPU=1 now then, otherwise you'd probably avoid the problem by other means (as it should not use Mesa at all then)... ;-) -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c12 --- Comment #12 from Stakanov Schufter <stakanov@freenet.de> --- (In reply to Wolfgang Bauer from comment #11)
Just to be sure: you shouldn't be using QT_WEBENGINE_DISABLE_GPU=1 now then, otherwise you'd probably avoid the problem by other means (as it should not use Mesa at all then)... ;-)
No, I did not use it. I only used the second solution as of my comments. The first one did not have a fastening effect (strange enough that it had). -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c13 --- Comment #13 from Wolfgang Bauer <wbauer@tmo.at> --- (In reply to Stakanov Schufter from comment #12)
(In reply to Wolfgang Bauer from comment #11)
Just to be sure: you shouldn't be using QT_WEBENGINE_DISABLE_GPU=1 now then, otherwise you'd probably avoid the problem by other means (as it should not use Mesa at all then)... ;-)
No, I did not use it. I only used the second solution as of my comments. The first one did not have a fastening effect (strange enough that it had). Ok, I just wanted to mention it to avoid misunderstandings.
Btw, the first one (i.e. QT_WEBENGINE_DISABLE_GPU=1) was not supposed to speed up anything, it might even slow down things as the GPU is not used anymore for rendering at all... -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c14 --- Comment #14 from Stakanov Schufter <stakanov@freenet.de> --- FYI, the MESA_EXTENSION_OVERRIDE="-GL_ATI_fragment_shader" entry has a secondary, positive, side effect on the plasma desktop. Normally since long time I was haunted by the "black screen after changing user with alt+ctl+fn key combination. The system would stay either hanging with a black screen (until reboot) or would enter the normal function only if entering space bar or sometimes enter. This would take about 30 seconds, sometimes more. This is entirely gone with the entry as of above. So on a X201 Lenovo, this change makes not only falkon faster, it makes also the desktop more stable. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c15 --- Comment #15 from Stakanov Schufter <stakanov@freenet.de> --- Created attachment 810494 --> http://bugzilla.opensuse.org/attachment.cgi?id=810494&action=edit falcon segfaults at exit new backtrace with the new settings. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c16 --- Comment #16 from Wolfgang Bauer <wbauer@tmo.at> --- (In reply to Stakanov Schufter from comment #15)
Created attachment 810494 [details] falcon segfaults at exit new backtrace
with the new settings.
Well, that's a completely different crash now. Could you please report it upstream at bugs.kde.org against Falkon please? Thanks. The relevant backtrace now seems to be: #6 QScopedPointer<QObjectData, QScopedPointerDeleter<QObjectData> >::data (this=0x8) at ../../include/QtCore/../../src/corelib/tools/qscopedpointer.h:140 #7 qGetPtrHelper<QScopedPointer<QObjectData> > (p=...) at ../../include/QtCore/../../src/corelib/global/qglobal.h:999 #8 QObject::d_func (this=0x0) at kernel/qobject.h:121 #9 QMetaObject::activate (sender=sender@entry=0x0, signalOffset=3, local_signal_index=local_signal_index@entry=3, argv=argv@entry=0x0) at kernel/qobject.cpp:3639 #10 0x00007f30bd08ac07 in QMetaObject::activate (sender=sender@entry=0x0, m=m@entry=0x7f30be313dc0 <History::staticMetaObject>, local_signal_index=local_signal_index@entry=3, argv=argv@entry=0x0) at kernel/qobject.cpp:3629 #11 0x00007f30bdfee8c3 in History::resetHistory (this=this@entry=0x0) at /usr/src/debug/falkon-3.0.1-lp151.2.6.x86_64/build/src/lib/FalkonPrivate_autogen/ZGCQNSLDUB/moc_history.cpp:185 #12 0x00007f30bdee353f in History::clearHistory (this=0x0) at /usr/src/debug/falkon-3.0.1-lp151.2.6.x86_64/src/lib/history/history.cpp:250 #13 0x00007f30bde88359 in MainApplication::saveSettings (this=0x7ffd858688a0) at /usr/src/debug/falkon-3.0.1-lp151.2.6.x86_64/src/lib/app/mainapplication.cpp:781 #14 0x00007f30bdff6c2d in MainApplication::qt_static_metacall (_o=0x7ffd858688a0, _c=<optimized out>, _id=<optimized out>, _a=<optimized out>) at /usr/src/debug/falkon-3.0.1-lp151.2.6.x86_64/build/src/lib/FalkonPrivate_autogen/VJIZ3MDCXP/moc_mainapplication.cpp:148 #15 0x00007f30bd08a535 in QMetaObject::activate (sender=0x7ffd858688a0, signalOffset=<optimized out>, local_signal_index=local_signal_index@entry=0, argv=argv@entry=0x7ffd85868770) at kernel/qobject.cpp:3767 #16 0x00007f30bd08ac07 in QMetaObject::activate (sender=<optimized out>, m=m@entry=0x7f30bd516640 <QCoreApplication::staticMetaObject>, local_signal_index=local_signal_index@entry=0, argv=argv@entry=0x7ffd85868770) at kernel/qobject.cpp:3629 #17 0x00007f30bd05a0e7 in QCoreApplication::aboutToQuit (this=<optimized out>, _t1=...) at .moc/moc_qcoreapplication.cpp:249 #18 0x00007f30bd05e60b in QCoreApplicationPrivate::execCleanup (this=0x556161beee70) at kernel/qcoreapplication.cpp:1317 #19 0x00007f30bd0629cd in QCoreApplication::exec () at kernel/qcoreapplication.cpp:1301 #20 0x000055616003c856 in main (argc=<optimized out>, argv=0x7ffd85868a98) at /usr/src/debug/falkon-3.0.1-lp151.2.6.x86_64/src/main/main.cpp:155 #6 QScopedPointer<QObjectData, QScopedPointerDeleter<QObjectData> >::data (this=0x8) at ../../include/QtCore/../../src/corelib/tools/qscopedpointer.h:140 #7 qGetPtrHelper<QScopedPointer<QObjectData> > (p=...) at ../../include/QtCore/../../src/corelib/global/qglobal.h:999 #8 QObject::d_func (this=0x0) at kernel/qobject.h:121 #9 QMetaObject::activate (sender=sender@entry=0x0, signalOffset=3, local_signal_index=local_signal_index@entry=3, argv=argv@entry=0x0) at kernel/qobject.cpp:3639 #10 0x00007f30bd08ac07 in QMetaObject::activate (sender=sender@entry=0x0, m=m@entry=0x7f30be313dc0 <History::staticMetaObject>, local_signal_index=local_signal_index@entry=3, argv=argv@entry=0x0) at kernel/qobject.cpp:3629 #11 0x00007f30bdfee8c3 in History::resetHistory (this=this@entry=0x0) at /usr/src/debug/falkon-3.0.1-lp151.2.6.x86_64/build/src/lib/FalkonPrivate_autogen/ZGCQNSLDUB/moc_history.cpp:185 #12 0x00007f30bdee353f in History::clearHistory (this=0x0) at /usr/src/debug/falkon-3.0.1-lp151.2.6.x86_64/src/lib/history/history.cpp:250 #13 0x00007f30bde88359 in MainApplication::saveSettings (this=0x7ffd858688a0) at /usr/src/debug/falkon-3.0.1-lp151.2.6.x86_64/src/lib/app/mainapplication.cpp:781 #14 0x00007f30bdff6c2d in MainApplication::qt_static_metacall (_o=0x7ffd858688a0, _c=<optimized out>, _id=<optimized out>, _a=<optimized out>) at /usr/src/debug/falkon-3.0.1-lp151.2.6.x86_64/build/src/lib/FalkonPrivate_autogen/VJIZ3MDCXP/moc_mainapplication.cpp:148 #15 0x00007f30bd08a535 in QMetaObject::activate (sender=0x7ffd858688a0, signalOffset=<optimized out>, local_signal_index=local_signal_index@entry=0, argv=argv@entry=0x7ffd85868770) at kernel/qobject.cpp:3767 #16 0x00007f30bd08ac07 in QMetaObject::activate (sender=<optimized out>, m=m@entry=0x7f30bd516640 <QCoreApplication::staticMetaObject>, local_signal_index=local_signal_index@entry=0, argv=argv@entry=0x7ffd85868770) at kernel/qobject.cpp:3629 #17 0x00007f30bd05a0e7 in QCoreApplication::aboutToQuit (this=<optimized out>, _t1=...) at .moc/moc_qcoreapplication.cpp:249 #18 0x00007f30bd05e60b in QCoreApplicationPrivate::execCleanup (this=0x556161beee70) at kernel/qcoreapplication.cpp:1317 #19 0x00007f30bd0629cd in QCoreApplication::exec () at kernel/qcoreapplication.cpp:1301 #20 0x000055616003c856 in main (argc=<optimized out>, argv=0x7ffd85868a98) at /usr/src/debug/falkon-3.0.1-lp151.2.6.x86_64/src/main/main.cpp:155 (which you should add to that bug report) I.e. it seems to crash now when saving the settings. You could also try the latest Falkon 3.1.0 from the KDE:Extra repo to see if it might have been fixed in this version. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.opensuse.org/show_bug.cgi?id=1140184 http://bugzilla.opensuse.org/show_bug.cgi?id=1140184#c21 Stakanov Schufter <stakanov@freenet.de> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|IN_PROGRESS |RESOLVED Resolution|--- |FIXED --- Comment #21 from Stakanov Schufter <stakanov@freenet.de> --- Resolved in the final version of 15.2. -- You are receiving this mail because: You are on the CC list for the bug.
participants (2)
-
bugzilla_noreply@novell.com
-
bugzilla_noreply@suse.com