Hi, I suspect you folks can't do anything, since we are dealing with a closed source plugin, but I will bugzilla anyway just in case :) For the past few weeks (maybe more than a month), firefox will crash numerous times per day at seemingly random times. Augh! I installed the debug info and captured 2 bt today. Hope they are useful. firefox-1.5.0.12-11.el5_1 flash-plugin-9.0.115.0-1.el5 nspr-4.6.5-3.el5 [rest is fully updated rhel5.1 (non fastrack) as well, I hope] daryl
Created attachment 295532 [details] Firefox -g bt
Created attachment 295533 [details] Another firefox -g bt
What would be *really* useful is if you can find a reproduce procedure or specific URL's that cause this crash.
Hi Warren, Well, http://www.accuweather.com will crash my firefox each time, but I go to another RHEL5 box and it seems to work just fine (with flash-plugin) as well. Maybe something is hosed on my laptop. The recent firefox errata gave me hope that perhaps this crashing was fixed, but I still see it. I'll see if I can give you something better to work with, thanks for the quick response. :) daryl
Created attachment 295731 [details] Firefox bt I uninstalled flash-plugin and haven't seen many crashes. I did get a crash this evening and have attached the bt, so perhaps something else is wrong and this bz could be moved to firefox component? thanks for the help.
Created attachment 295763 [details] Firefox bt Another crash without flash-plugin
Created attachment 295864 [details] firefox bt Crashes a couple of times per day.
Created attachment 296002 [details] firefox bt Another crash... I tried to refile this under firefox, but may need some help getting it assigned correctly? Thanks!
Created attachment 296008 [details] very long firefox bt
*** Bug 430140 has been marked as a duplicate of this bug. ***
I get periodic crashes when I go to a site and I click on a link to open something like a PDF file. Someone in another bz suggested gtk and this thread: http://www.centos.org/modules/newbb/viewtopic.php?topic_id=11589&viewmode=flat&order=ASC&start=20 I have not tried the suggested workaround.
Created attachment 297447 [details] backtrace from firefox-1.5.0.12-12.el5_1.i386
Can you attach more backtraces? (and omit all crashes in /usr/lib/flash-plugin/libflashplayer.so, it's a crash in flash, not in firefox)
re comment #15, is that directed to me? I continue to see 1-2 crashes per day, they usually come during one of these actions: - Right click to 'Save As' some link on a webpage - During HTTP Auth
yes, please attach some more backtraces.
Created attachment 297621 [details] second firefox backtrace
Created attachment 297624 [details] another similar crash
Created attachment 297626 [details] bt Just got this bt just before attempting to do HTTP Auth
All crashes seems to be inside GTK main loop (g_main_context_iterate) so it doesn't look like a firefox bug. All those instances were run with DE locales - can you try to run it with the English one?
Created attachment 297631 [details] backtrace without DE locales
Created attachment 297632 [details] another non-DE backtrace
Created attachment 297671 [details] firefox bt Another backtrace
Created attachment 297685 [details] firefox bt And another :)
Created attachment 297697 [details] firefox bt I seem to get this each time I visit cnbc.com and click on "Mad Money ReCap", hehe :)
Created attachment 297699 [details] another bt I will stop with posting these unless you want more :)
Please try to find out some steps how to reproduce the crash.
Please test packages at http://people.redhat.com/stransky/mozilla-updated/firefox/ The should fix some regression (but not the gtk dialog crash).
Created attachment 297783 [details] bt from cnbc.com Using firefox-debuginfo-1.5.0.12-12.el5_1 reproducer: Visit cnbc.com scroll down and click on the red banner saying "MadCap Recap", crash.
(In reply to comment #30) > Created an attachment (id=297783) [edit] > bt from cnbc.com > > Using firefox-debuginfo-1.5.0.12-12.el5_1 > > reproducer: > Visit cnbc.com > scroll down and click on the red banner saying "MadCap Recap", crash. This crash is going to be fixed in the upcoming test package (see Bug 437124)
GTK dialog crashes are covered by Bug 437132.
NSS/NSPR crashes are covered by Bug 437135.
Created attachment 297803 [details] bt , looks to be java plugin related Hi, I may just be jinxed. :) Here is another I just got while having a PDF open and attempting to close the window via the 'x' in the corner. It looks new, so I thought I would add it.
(In reply to comment #34) > Created an attachment (id=297803) [edit] > bt , looks to be java plugin related > > Hi, > > I may just be jinxed. :) Here is another I just got while having a PDF open > and attempting to close the window via the 'x' in the corner. It looks new, so > I thought I would add it. No, It's the NSPR/NSS crash (Bug 437135)
Updated packages (with fixed Bug 437124) will be at http://people.redhat.com/stransky/mozilla-updated/ (firefox-1.5.0.12-13.el5_1)
Thanks for the new builds - I continue to run them in my daily setup as well. The last build, -12, was pretty stable for me but I still did occasionally get a crash. I will run the latest (-13) under gdb and post a backtrace if it is interesting.
I have been having very good success with -13 build. The only bug I am left seeing is the GTK dialog bug noted here: bz 437132 thanks!
thanks for letting us know.
Huh? Why was this closed?
(In reply to comment #40) > Huh? Why was this closed? I am sorry, I misunderstood your comment 37 -- I thought that all your problems (except for bug 437132) are fixed.
That is correct, but the package that is working is non supported and certainly not the "CURRENTRELEASE". Sorry about the confusion.
OK, sorry, that should be NEXTRELEASE (I am almost never use it outside of RHEL bugs, and got confused).
Please leave this bug open.
I just commented in #437124 that the released firefox-1.5.0.12-14.el5_1 is missing the firefox-1.5.0.12-empty-document.patch I see that the rpm does however already include the same fix as was patched by mozilla-279505.patch (at least attempting to apply that reports that all the hunks are already applied) - I'd have mentioned that in #437124 but that is a bugzilla just for the empty-document crash... I'm still not sure if the simplish patch from #437132 ought to be included or not. Certainly it seems to do no harm here - at least when applied to the previous fiefox packages. I'm currently building a test version based on firefox-1.5.0.12-14... with the extra two patches (empty-doc and mozcontainer) and hope to be able to test it soonish. -- Jon
FYI, firefox-1.5.0.12-14.el5_1 (the recent errata release) will routinely become unresponsive and peg the CPU. It won't actually crash tho. This usually happens when closing a tab or after visiting a javascript intensive site like Google Reader or Gmail . I don't have an easy reproducer tho, sorry :( firefox-1.5.0.12-13.el5_1 from stransky's site is working much better for me. daryl
I'm still seeing the same behavior with the new Firefox package, too. I'm running firefox-1.5.0.12-14.el5.centos. I was attempting to download Realplayer when I got the following: Program received signal SIGSEGV, Segmentation fault. [Switching to Thread -1208928032 (LWP 5168)] 0x047f6238 in gtk_widget_get_toplevel () from /usr/lib/libgtk-x11-2.0.so.0 (gdb) continue Continuing. Program received signal SIGSEGV, Segmentation fault. 0x00d91402 in __kernel_vsyscall () (gdb) continue Continuing. Program terminated with signal SIGSEGV, Segmentation fault. The program no longer exists. (gdb) thread apply all bt (gdb) quit If I am able to get a bt, I'll post it.
Created attachment 299632 [details] Backtrace from Firefox The "Download RealPlayer" link from http://www.real.com/linux?pcode=rn&opage=freeplayer_partner&src=freeplayer_partner is pretty consistent with its ability to cause Firefox to seg fault. This backtrace is from the above sequence of links.
Hi Ryan, I think you should check out this bugzilla entry #437132 , it is the GTK bug that I keep hitting many times per day as well. :( daryl
An extra patch (firefox-1.5.0.12-colormap.patch) is in package firefox-1.5.0.12-13.debug.el5 at http://people.redhat.com/stransky/mozilla-updated/. Could you please check if it fixes the colourmap crashes for you?
That seems to have fixed the issue I was seeing. I also *was* experiencing trouble visiting our OpenManage installation, but that appears to have been resolved, as well. I'll keep testing with this patch and update as necessary. Thank-you!
Not that I really want to complain much but would it not make sense to base the debug/test releases on the most recent *released* el5 version of the firefox src.rpm? ie firefox-1.5.0.12-14... if nothing else this test release is considered 'old' by rpm etc.
Yeah, that's a good point. I had to remove firefox-1.5.0.12-14 in order to install these test packages. Wouldn't it be easy to roll this patch into the -14 release? I've been using this version most of the day without incident...so far so good.
Well -14 has the recent security fixes but was missing one of the two fixes present in the previous test packages (e.g. -13 sadly no longer visible in the directory mentioned above). Comparing the specfiles from the previous -13 and this -13.debug it just has the extra firefox-1.5.0.12-colormap.patch so doesn't *seem* to have any of the security fixes. I'm about to rebuild based on -14 again with the various test patches added back in again (I already have version with the others added and will just add firefox-1.5.0.12-colormap.patch as well). btw do the gaps I see in the comment numbers on this ticket indicate that there are 'hidden' comments I'm not allowed to read?
Well, an updated package firefox-1.5.0.12-14.1.debug.el5_1 based on the latest security update is at http://people.redhat.com/stransky/mozilla-updated/
Thanks Martin, I am using your updated package now. Yesterday, I had 7 crashes (6 appeared to be flash related and other was GTK). I suppose I could start another bz ticket for my flash crashes?
I just went to Help -> About and the Firefox window disappeared. That has happened from time to time, too. Meaning, a Firefox window causes Firefox to die. I'll grab the latest packages and run them under GDB to (hopefully) capture a bt.
Flash crashes are at Bug 222646
Something else must have been done between -12 and -14. I'm able to get to our internal OpenManage site w/ -12, but not with -14. I have captured a bt and will attach.
Created attachment 300251 [details] Backtrace from visiting internal OpenManage site.
Let me ask a question. I believe RHEL 5.2 is going to re-base to Minefield. Does it make sense to pursue this in great detail if this version of Firefox will no longer be used?
I heard back from my support call w/ Red Hat (1805714) that the FF developers are focussing on FF3 and would rather let FF1.5.x go. From here on out, I will get FF3 and start testing with that version. Thank-you for the help on this.
Hi Martin, I would be happy to keep testing FF 1.5 releases. Currently, I continue to see a 100% CPU usage of the -14 firefox line (both errata and your new package this morning) after visiting some flash sites. I know, you can't do much to help because of flash, but was curious if there is a way to get debugging info when this CPU pegging occurs? Maybe once the GTK bug is fixed, I'll move on the FF3 as well. I am very thankful for the help! daryl
The flash problem (crashes/hangs) should be solved in RHEL-5.2 by a new package, nspluginwrapper. It shields the flash plugin from the browser and the flash crashes don't affect the browser.
(In reply to comment #65) > Hi Martin, > > I would be happy to keep testing FF 1.5 releases. Currently, I continue to see > a 100% CPU usage of the -14 firefox line (both errata and your new package this > morning) after visiting some flash sites. I know, you can't do much to help > because of flash, but was curious if there is a way to get debugging info when > this CPU pegging occurs? gdb --pid=XXX where XXX is a PID of running main firefox process.
gdb --pid=${pidof firefox-bin} is probably slighly more digestable
Created attachment 301408 [details] bt during 100% CPU usage for 2 minutes I generated this bt while my firefox was pegging the CPU for a few minutes. Thanks for the tips in the previous messages and for the continued help.
Created attachment 302028 [details] Core file produced by firefox-1.5.0.12-14.1.debug.el5_1
I found a web page that crashes firefox 100% of the time: http://www.worldtimeserver.com/convert_time_in_US-MN.aspx Sometimes just visiting that page will crash it, but if you choose a country on the right (I've been choosing UTC), firefox will crash. Looking at the bt, it's similar to crashes I get in less reproduceable scenarios. I tested multiple systems with freshly created users and they all crash. I removed all plugins from the systems as well. I've tested the current firefox (firefox-1.5.0.12-14.el5_1) and the debug edition posted here (referenced in comment 57). They both crash on that page. My core is attached to comment 70.
Cool, I can reproduce the crash at http://www.worldtimeserver.com/convert_time_in_US-MN.aspx
Crash at http://www.worldtimeserver.com/convert_time_in_US-MN.aspx is tracked by Bug 442023.
Should be fixed by the RHEL-5.2 update and Firefox 3.