GNOME Bugzilla – Bug 363585
crash in Evolution: Closing a image attachem...
Last modified: 2006-10-21 07:23:38 UTC
What were you doing when the application crashed? Closing a image attachement (PNG), sending a mail (disconnected mode). Distribution: Gentoo Base System version 1.12.5 Gnome Release: 2.16.1 2006-10-16 (Gentoo) BugBuddy Version: 2.16.0 Memory status: size: 104689664 vsize: 0 resident: 104689664 share: 0 rss: 29696000 rss_rlim: 0 CPU usage: start_time: 1161329078 rtime: 0 utime: 489 stime: 0 cutime:455 cstime: 0 timeout: 34 it_real_value: 0 frequency: 0 Backtrace was generated from '/usr/bin/evolution-2.8' (no debugging symbols found) Using host libthread_db library "/lib/libthread_db.so.1". (no debugging symbols found) [Thread debugging using libthread_db enabled] [New Thread -1231931712 (LWP 25192)] [New Thread -1288037488 (LWP 25211)] [New Thread -1276859504 (LWP 25210)] [New Thread -1278215280 (LWP 25209)] [New Thread -1267848304 (LWP 25206)] [New Thread -1248638064 (LWP 25204)] 0xb7fb3410 in ?? ()
+ Trace 77275
(In reply to comment #0) > Closing a image attachement (PNG), sending a mail (disconnected mode). Can you please be more detailed about the attachment part? Did you close the inline view of an attached image in your mail folders? Thanks for the bug report. Unfortunately, that stack trace is not very useful in determining the cause of the crash. Could you please install some debugging packages [1] and reproduce the crash, if possible? Once bug-buddy pops up, you can find the stacktrace in the "details", now containing way more information. Please copy that stacktrace and paste it as a comment here. Thanks! [1] debugging packages for evolution, evolution-data-server and gtkhtml, plus debugging packages for some basic GNOME libs. More details can be found here: http://live.gnome.org/GettingTraces/DistroSpecificInstructions
Unable to reproduce after at least 100 open/close of the same image and 20 email sendings in disconnected mode. I don't even know why I send this report with my awful over-optimized stack trace ^^ I recompiled Evolution with debug symbols so if it happens I'll have mode interesting things to tell ("what? you are unable to translate a memory address in a function name?" ;-)).
Heh, actually we got the function names. Or rather, "some" function names. That stacktrace unfortunately is broken and does not show any info about the real cause of the crash. Pretty common these days, but enabling debugging usually results in perfect traces. Btw, sorry for the "package" reference. I overlooked you're a Gentoo user. :) Since you just rebuilt Evo -- did you rebuild the exact same code, or did you re-emerge? What's the exact version , both now and at the time of the crash? Any custom or Gentoo specific patches? Also, could you please answer the first paragraph of comment 1?
1) The full spec on my Gentoo build : mail-client/evolution-2.8.1-r1 USE="crypt dbus hal ipv6 pda spell ssl -bogofilter -debug -doc -kerberos -krb4 -ldap -mono -nntp -profile" Usefull complementary info : gnome-extra/evolution-data-server-1.8.1 USE="ipv6 ssl -debug -doc -kerberos -keyring -krb4 -ldap -nntp" 2) I don't mind the "package" references :-) I came from LFS to Gentoo just because it was what I was building around LFS... 3) I re-emerged the exact same version, using the exact same ebuild. A small analysis of the ebuild process reveals that 4 patches are applied. They are called evolution-2.3.7-configure_plugins, evolution-2.8.0-uri.patch, evolution-2.8.1-64-bit and evolution-2.8.1-filter-bar-crash. I'll attach them. 4) Yes, it was an inline view, in the main Evo window. I can't be more detailled, it was a screenshot in the PNG format. I can attach the mail if you feel it useful (it was from my local LUG mailing list, nothing private).
Created attachment 75100 [details] [review] patch #1
Created attachment 75101 [details] [review] patch #2
Created attachment 75102 [details] [review] patch #3
Created attachment 75103 [details] [review] patch #4
Created attachment 75105 [details] [review] patch #2 - the real one (oops)
Thanks for all these details, Mikael. :) OK, so now I got a pretty clear picture. According to the description, this is a crash that bites sometimes (though not always) when showing or hiding images attached to a mail. More precisely, this most likely is bug 333864 yet again. The fix for this crash is not yet part of the Gentoo ebuild. Checked the provided reference patches. Marking this bug as a duplicate of bug 333864. Unfortunately you built with USE flag -debug, so we won't get a better stacktrace, unless you rebuild. Mikael, if you ever can reproduce this particular crash, please enable debugging and reproduce it. ;) FWIW, this crash is fixed in Evolution 2.8.1.1. Thanks again for your fast replies and your love to details. Much appreciated. (Marking the patches obsolete, so bugzilla won't report them.) *** This bug has been marked as a duplicate of 333864 ***
> Unfortunately you built with USE flag -debug, so we won't get a better stacktrace, unless you rebuild. Mikael, if you ever can reproduce this particular crash, please enable debugging and reproduce it. ;) In fact I modified my CFLAGS removing the -fomit-frame-pointer and adding -ggdb, so we would have something better. Anyway, I put the debug useflag on Evo so if I it that bug again I'll have something interesting to tell ;-) My Evo is use to crash less than once every 6 month, and I can tell you, I use lots of its possibilities, disconnected mode, slow connection (here in New Caledonia, you have at least 600ms to the rest of the world -- Evo delay in closing is quite anoying in this regard, there should be a way to stop tasks more cleanly and *fastly*, and moreover not having other tasks starting while closing), thousands of mails, including those on my machine, where I sometimes receive database dumps from stupid users ;-) I appreciated your fast replies too, I was even impressed at the first one. Really good contact, too, informative, short, you clearly master customer relation.
Oops, sorry -- guess I got the use flag the wrong way around. Not used to Gentoo myself. ;) Anyway... Thanks for your kind words, Mikael. :-) Btw, from a Gentoo maintainer I do happen to know that the Evolution 2.8.1.1 (which fixes this bug) ebuild should be rolled "today". Considering timezones, it probably is out already.
True, I build evo-2.8.1.1 with debug info ;-) For the use-flags, you were absolutely right, the debug useflag is the one to use. It's just that I'm used to tweak CFLAGS... LFS background ;-)