[Bug 811435] New: Could not launch application: Failed to fork (Cannot allocate memory), in fallback mode
https://bugzilla.novell.com/show_bug.cgi?id=811435 https://bugzilla.novell.com/show_bug.cgi?id=811435#c0 Summary: Could not launch application: Failed to fork (Cannot allocate memory), in fallback mode Classification: openSUSE Product: openSUSE 12.3 Version: Final Platform: x86-64 OS/Version: Other Status: NEW Severity: Normal Priority: P5 - None Component: GNOME AssignedTo: bnc-team-gnome@forge.provo.novell.com ReportedBy: gp@suse.com QAContact: qa-bugs@suse.de Found By: Product Management Blocker: --- Created an attachment (id=531553) --> (http://bugzilla.novell.com/attachment.cgi?id=531553) Screenshot of gnome-panel; nothing unusual, basically what I had for years I am running GNOME fallback mode, and after about two days I can no longer start applications from the panel. When clicking on one of the icons, I get an error message box: Could not launch application Failed to fork (Cannot allocate memory) Doing the same from the command-line works just fine and the entire system is not out of memory (no swapping or anything). HOWEVER, top shows the following: PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 1294 gp 20 0 2238m 667m 18m S 22,3 35,1 352:46.02 firefox 1210 gp 20 0 4204m 190m 8072 S 0,3 10,0 2:26.59 gnome-panel This is 4GB of virtual memory for gnome-panel. Sounds like a huge leak. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=811435 https://bugzilla.novell.com/show_bug.cgi?id=811435#c1 Vincent Untz <vuntz@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |vuntz@suse.com --- Comment #1 from Vincent Untz <vuntz@suse.com> 2013-03-25 16:28:14 UTC --- I assume you didn't have that issue in 12.2... These are the commits that touch codes between 3.4 and 3.6: na: apply style after realize panel: Fix launcher icon animation ending with black artifact Use-after-free in panel_object_loader_stop_loading() Avoid double forking when launching apps; it breaks "pkexec" clock: Set correct month on startup panel: Remove ellipses from user menu to match GNOME Shell 3.6 panel: Remove Online Accounts item from the user menu panel: Rearrange user menu order to match GNOME Shell 3.6 Port to new documentation infrastruture Rewrite evolution-data-server detection Remove markup from translations libpanel-util: adjust to DConf API changes Clock Applet: port to modern GNOME technologies clock: Adapt to Evolution-Data-Server API changes (yes, gnome-panel is kind of dead) To me, it sounds like an issue either in one library used by gnome-panel or in the new clock applet code. Could you possibly provide a valgrind log? I looked around upstream, and there is a similar bug: https://bugzilla.gnome.org/show_bug.cgi?id=688967 For the time being, a workaround is to stop using locations in the clock applet. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=811435 https://bugzilla.novell.com/show_bug.cgi?id=811435#c2 --- Comment #2 from Gerald Pfeifer <gp@suse.com> 2013-03-25 20:49:00 CET ---
I assume you didn't have that issue in 12.2...
Yes, this is new. With 12.2 I did not see this with uptimes of two three weeks even.
To me, it sounds like an issue either in one library used by gnome-panel or in the new clock applet code. Could you possibly provide a valgrind log?
How can I do so? Online I found https://live.gnome.org/MemoryReduction/Tools/GetValgrindInToughPlaces but that does not work in GNOME 3.6 any more (g-s-delete does not exist, g-s-p looks differently).
For the time being, a workaround is to stop using locations in the clock applet.
Another is to simply `pkill gnome-panel` once a day. That will restart the panel and seems to work just fine. :-) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=811435 https://bugzilla.novell.com/show_bug.cgi?id=811435#c3 --- Comment #3 from Vincent Untz <vuntz@suse.com> 2013-03-26 11:05:06 UTC --- (In reply to comment #2)
To me, it sounds like an issue either in one library used by gnome-panel or in the new clock applet code. Could you possibly provide a valgrind log?
How can I do so? Online I found https://live.gnome.org/MemoryReduction/Tools/GetValgrindInToughPlaces but that does not work in GNOME 3.6 any more (g-s-delete does not exist, g-s-p looks differently).
"gnome-panel --replace" should work, so you can start this in valgrind. Alternatively, the usual trick of "mv /usr/bin/gnome-panel /usr/bin/gnome-panel.real; vi /usr/bin/gnome-panel" works. Note that you'll need debuginfo packages for gnome-panel and libgweather at least (glib2 would be nice too).
For the time being, a workaround is to stop using locations in the clock applet.
Another is to simply `pkill gnome-panel` once a day. That will restart the panel and seems to work just fine. :-)
Heh, if that's good enough for you, that works ;-) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=811435 https://bugzilla.novell.com/show_bug.cgi?id=811435#c Dominique Leuenberger <dimstar@opensuse.org> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |dimstar@opensuse.org See Also| |https://bugzilla.gnome.org/ | |show_bug.cgi?id=688967 -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=811435 https://bugzilla.novell.com/show_bug.cgi?id=811435#c4 --- Comment #4 from Gerald Pfeifer <gp@suse.com> 2013-04-26 00:46:14 CEST --- Soo, I finally managed to get those debug channels and packages added and the machine rebooted into a fresh stage. Alas, when running valgrind gnome-panel --replace I get to that unhappy computer "Something went wrong, log out" quickly. Quickly, as within a second or two. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=811435 https://bugzilla.novell.com/show_bug.cgi?id=811435#c5 --- Comment #5 from Gerald Pfeifer <gp@suse.com> 2013-04-26 00:47:45 CEST --- Created an attachment (id=536950) --> (http://bugzilla.novell.com/attachment.cgi?id=536950) Output of valgrind gnome-panel --replace, though probably not that useful -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=811435 https://bugzilla.novell.com/show_bug.cgi?id=811435#c6 --- Comment #6 from Vincent Untz <vuntz@suse.com> 2013-04-26 07:23:27 UTC --- (In reply to comment #4)
Alas, when running valgrind gnome-panel --replace I get to that unhappy computer "Something went wrong, log out" quickly. Quickly, as within a second or two.
Doh, I see. That's an old bug with gnome-session being a bit too aggressive when detecting issues. I don't remember if/when this was fixed -- maybe it's just in 3.8... I'd recommend the trick I mentioned in comment 3, in that case (moving gnome-panel to gnome-panel.real and making gnome-panel a wrapper script that calls gnome-panel.real with valgrind). -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=811435 https://bugzilla.novell.com/show_bug.cgi?id=811435#c7 --- Comment #7 from Gerald Pfeifer <gp@suse.com> 2013-04-27 09:21:45 CEST --- Created an attachment (id=537173) --> (http://bugzilla.novell.com/attachment.cgi?id=537173) Output of valgrind using the "replace /usr/bin/gnome-panel" trick This is using the alternate approach. It really hurt my system, badly when I used --leak-check=full, and seems to lack the desired output? How do I have to close/kill gnome-panel in this case? A simple logout does not seem to do it? Nor does killing /usr/lib64/valgrind/memcheck-amd64-linux? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=811435 https://bugzilla.novell.com/show_bug.cgi?id=811435#c8 --- Comment #8 from Vincent Untz <vuntz@suse.com> 2013-05-24 09:07:03 UTC --- Sorry for the late reply; this got lost in my "I left for vacation" mails ;-) (In reply to comment #7)
This is using the alternate approach. It really hurt my system, badly when I used --leak-check=full, and seems to lack the desired output?
Indeed, it lacks what we want. Not sure why :/
How do I have to close/kill gnome-panel in this case? A simple logout does not seem to do it? Nor does killing /usr/lib64/valgrind/memcheck-amd64-linux?
You should be able to issue a "gnome-panel --replace" to start a new instance of the panel. But logging out certainly should be enough to have it stop. Maybe there's some race because valgrind makes it so slow, and it doesn't catch the signal? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=811435 https://bugzilla.novell.com/show_bug.cgi?id=811435#c9 --- Comment #9 from Vincent Untz <vuntz@suse.com> 2013-05-24 09:13:20 UTC --- Hrm: http://valgrind.org/docs/manual/faq.html#faq.hiddenbug suggests that this could happen if program is statically linked. Which is not the case. I'm lost here. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=811435 https://bugzilla.novell.com/show_bug.cgi?id=811435#c10 --- Comment #10 from Gerald Pfeifer <gp@suse.com> 2013-05-27 01:00:24 CEST --- Created an attachment (id=541250) --> (http://bugzilla.novell.com/attachment.cgi?id=541250) New run of gnome-panel under valgrind (In reply to comment #8)
Sorry for the late reply; this got lost in my "I left for vacation" mails ;-)
No worries. Vacation is good.
Indeed, it lacks what we want. Not sure why :/
I fiddled with this a bit more. How about the attached? To me this clearly points to a smoking gun: ==18637== 2,538,880 bytes in 19,835 blocks are possibly lost in loss record 13,679 of 13,682 ==18637== at 0x4C2C27B: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==18637== by 0x4C2C52F: realloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==18637== by 0x75CD4AE: g_realloc (gmem.c:224) ==18637== by 0x759CD09: g_ptr_array_maybe_expand (garray.c:1093) ==18637== by 0x759DE62: g_ptr_array_add (garray.c:1350) ==18637== by 0x1B040733: location_new_from_xml (gweather-location.c:239) ==18637== by 0x1B040877: location_new_from_xml (gweather-location.c:231) ==18637== by 0x1B04071F: location_new_from_xml (gweather-location.c:221) ==18637== by 0x1B0406A7: location_new_from_xml (gweather-location.c:208) ==18637== by 0x1B040A86: gweather_location_new_world (gweather-location.c:310) ==18637== by 0x1B038126: gweather_info_set_location_internal (weather.c:1750) ==18637== by 0x7343CBB: g_object_constructor (gobject.c:1358) There are a couple of similar ones, but all have location_new_from_xml and gweather_location_new_world. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=811435 https://bugzilla.novell.com/show_bug.cgi?id=811435#c11 --- Comment #11 from Gerald Pfeifer <gp@suse.com> 2013-07-21 07:49:04 CEST --- Created an attachment (id=548759) --> (http://bugzilla.novell.com/attachment.cgi?id=548759) Another log file showing the leaks, this a lot more extensive I created another log file that covers a longer period and hopefully points to the issue more clearly. Any update, Vincent? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
participants (1)
-
bugzilla_noreply＠novell.com