I emerged gtk+ 2.4 and it's depends (glib 2.4) then re-emerged all the GTK+ engines and librsvg. After this I logged out and logged back in. (just to get everything to use the new GTK+). Gaim started up just fine but as soon as I double click on a SN to send an IM it closes/crashes. I ran gaim with -d. And I get this output when I double click on a SN. g_log: Module error: /usr/lib/enchant/libenchant_ispell.so: undefined symbol: __gxx_personality_v0 g_log: Module error: /usr/lib/enchant/libenchant_myspell.so: undefined symbol: __gxx_personality_v0 GLib: gmem.c:157: failed to allocate 2691625484 bytes Aborted so then I did emerge sync && emerge aspell aspell-en enchant gtkspell gaim. To rebuild all of that spell stuff. (those files are in the enchant package). Same problem as above when I tried again. Reproducible: Always Steps to Reproduce: 1. emerge glib gtk+ 2. run gaim 3. double-click on buddy emerge info Portage 2.0.50-r1 (default-x86-1.4, gcc-3.3.3, glibc-2.3.3_pre20040207-r0, 2.6.3-gentoo-r2) ================================================================= System uname: 2.6.3-gentoo-r2 i686 Mobile Intel(R) Pentium(R) 4 - M CPU 2.40GHz Gentoo Base System version 1.4.3.13p1 distcc 2.13 i686-pc-linux-gnu (protocols 1 and 2) (default port 3632) [disabled] Autoconf: sys-devel/autoconf-2.59-r3 Automake: sys-devel/automake-1.8.2 ACCEPT_KEYWORDS="x86 ~x86" AUTOCLEAN="yes" CFLAGS="-march=pentium4 -O3 -pipe -fomit-frame-pointer -fexpensive-optimizations -ffast-math -falign-functions=4 -fforce-addr -funroll-loops -maccumulate-outgoing-args -frerun-loop-opt -fprefetch-loop-arrays -frerun-cse-after-loop -mfpmath=sse" CHOST="i686-pc-linux-gnu" COMPILER="gcc3" CONFIG_PROTECT="/etc /usr/X11R6/lib/X11/xkb /usr/kde/2/share/config /usr/kde/3.1/share/config /usr/kde/3.2/share/config /usr/kde/3/share/config /usr/lib/mozilla/defaults/pref /usr/share/config /var/qmail/control" CONFIG_PROTECT_MASK="/etc/gconf /etc/terminfo /etc/env.d" CXXFLAGS="-march=pentium4 -O3 -pipe -fomit-frame-pointer -fexpensive-optimizations -ffast-math -falign-functions=4 -fforce-addr -funroll-loops -maccumulate-outgoing-args -frerun-loop-opt -fprefetch-loop-arrays -frerun-cse-after-loop -mfpmath=sse" DISTDIR="/usr/portage/distfiles" FEATURES="autoaddcvs ccache sandbox" GENTOO_MIRRORS="http://gentoo.oregonstate.edu http://distro.ibiblio.org/pub/Linux/distributions/gentoo" MAKEOPTS="-j2" PKGDIR="/usr/portage/packages" PORTAGE_TMPDIR="/var/tmp" PORTDIR="/usr/portage" PORTDIR_OVERLAY="/usr/local/portage" SYNC="rsync://rsync2.us.gentoo.org/gentoo-portage" USE="X alsa apm avi berkdb cdr crypt cups dvd encode esd foomaticdb footmaticdb gdbm gif gnome gnomedb gpm gstreamer gtk gtk2 imlib ipv6 java jpeg libg++ libwww mad mikmod motif mozilla moznocompose moznoirc moznomail mozsvg mpeg mysql ncurses nogcj nptl oav oggvorbis opengl oss pam pcmcia pdflib perl pic png pnp ppds python qt quicktime readline samba sdl slang slp smooth spell sse ssl svga tcltk tcpd tiff truetype usb wmf x86 xml2 xmms xv zlib"
I tried to emerge gaim with USE="-spell" so it doesn't use any of the spell checker stuff. Similar problem happens when I follow the steps below. GLib: gmem.c:157: failed to allocate 2414792517 bytes Aborted The problem also exists when someone tries to IM me. So it's not just one way. It looks like when a chat window tries to open the problem happens.
Hmm works fine for me on three different machines I use. fwiw, all use gcc 3.2.3 still. I'll dig into it, hold tight.
I know right off the bat the gaim people will go nuts looking at all of your optimizations. Can you try with CXXFLAGS="-march=pentium4 -O3" and rebuild gaim ( at least).
Also regardless of what your flags are I need you to build gaim with USE="debug" and run it in gdb and attach the output of backtrace after it fails.
For more details: http://gaim.sourceforge.net/gdb.php
Actually I see that I'm getting those two enchant errors in debug window when I open a conversation window. However my spell-checking still works and gaim doesn't crash. I would file a bug with enchant on that one anyway. If you could get a backtrace on that gmem.c crash I'd appreciate it.
USE="debug" emerge gaim . . . make[3]: Entering directory `/var/tmp/portage/gaim-0.75-r10/work/gaim-0.75/plugins/docklet' source='docklet.c' object='docklet.lo' libtool=yes \ depfile='.deps/docklet.Plo' tmpdepfile='.deps/docklet.TPlo' \ depmode=gcc3 /bin/sh ../../depcomp \ /bin/sh ../../libtool --silent --mode=compile gcc -DHAVE_CONFIG_H -I. -I. -I../.. -DDATADIR=\"/usr/share\" -DVERSION=\"0.75\" -I../../src -DG_DISABLE_DEPRECATED -DGDK_DISABLE_DEPRECATED -DGDK_PIXBUF_DISABLE_DEPRECATED -DGTK_DISABLE_DEPRECATED -I/usr/include/gtk-2.0 -I/usr/lib/gtk-2.0/include -I/usr/include/atk-1.0 -I/usr/include/pango-1.0 -I/usr/X11R6/include -I/usr/include/freetype2 -I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include -march=pentium4 -O3 -pipe -fomit-frame-pointer -fexpensive-optimizations -ffast-math -falign-functions=4 -fforce-addr -funroll-loops -maccumulate-outgoing-args -frerun-loop-opt -fprefetch-loop-arrays -frerun-cse-after-loop -mfpmath=sse -Wall -g3 -c -o docklet.lo `test -f 'docklet.c' || echo './'`docklet.c In file included from docklet.c:42: ../../src/gtkblist.h:65: error: syntax error before "GtkItemFactory" ../../src/gtkblist.h:65: warning: no semicolon at end of struct or union ../../src/gtkblist.h:89: error: syntax error before '}' token ../../src/gtkblist.h:89: warning: type defaults to `int' in declaration of `GaimGtkBuddyList' ../../src/gtkblist.h:89: warning: data definition has no type or storage class In file included from ../../src/gtkutils.h:29, from docklet.c:47: ../../src/gtkconv.h:77: error: syntax error before "GtkItemFactory" ../../src/gtkconv.h:77: warning: no semicolon at end of struct or union ../../src/gtkconv.h:77: warning: no semicolon at end of struct or union ../../src/gtkconv.h:79: warning: type defaults to `int' in declaration of `menu' ../../src/gtkconv.h:79: warning: data definition has no type or storage class ../../src/gtkconv.h:89: error: syntax error before '}' token make[3]: *** [docklet.lo] Error 1 make[3]: Leaving directory `/var/tmp/portage/gaim-0.75-r10/work/gaim-0.75/plugins/docklet' make[2]: *** [all-recursive] Error 1 make[2]: Leaving directory `/var/tmp/portage/gaim-0.75-r10/work/gaim-0.75/plugins' make[1]: *** [all-recursive] Error 1 make[1]: Leaving directory `/var/tmp/portage/gaim-0.75-r10/work/gaim-0.75' make: *** [all] Error 2 !!! ERROR: net-im/gaim-0.75-r10 failed. !!! Function src_compile, Line 78, Exitcode 2 !!! Make failed So I can't give you that output. Looks like some of the add-on patches in the ebuild apply ok when debug isn't on.
Alright I'm getting this too. Digging in.
Looks like the problem with debug is upstream, as I'm getting it out of the tarball. Sounds like gtk+ deprecated some stuff, and enabling debug puts the clamps down on deprecated stuff. I'll file a bug upstream on that. For now can you humor me and try to get a backtrace on your normal build?
Alright I just fixed my use of the debug flag. Gaim devs say not to make use of --enable-debug in their configure. So I'm instead making use of the debug.eclass which will add -g to the CFLAGS and also does the nostrip that I was neglecting earlier. Give it about 30 minutes, rsync, and try again.
Well I ran my normal build in gdb and of course there was no problem at all. It ran as expected. :( But it still crashes when I just run gaim. So then I did USE="debug" emerge gaim. And it built gaim with debug symbols properly and ran it in gdb. And it didn't crash. I ran it out of gdb and it crashed. It doesn't produce a core file. Debugging isn't my speciality. Any other ideas? I've stepped it down to -mmarch=pentium4 -O2 -pipe and it still does it.
Run it from xterm with "gaim -d > gaim.log", attach log here.
So I was able to get the compiled with CFLAGS="-march=pentium4 -O2 -pipe" version. (that's glib and gaim compiled like that) to core dump. This is the non-USE="debug" version. (gdb) backtrace #0 0x40687031 in kill () from /lib/libc.so.6 #1 0x4003764d in pthread_kill () from /lib/libpthread.so.0 #2 0x400379a1 in raise () from /lib/libpthread.so.0 #3 0x40686de4 in raise () from /lib/libc.so.6 #4 0x406881e0 in abort () from /lib/libc.so.6 #5 0x4050b2c5 in g_logv () from /usr/lib/libglib-2.0.so.0 #6 0x4050b2f9 in g_log () from /usr/lib/libglib-2.0.so.0 #7 0x40508e4e in g_malloc0 () from /usr/lib/libglib-2.0.so.0 #8 0x411d80f3 in rsvg_start_text () from /usr/lib/librsvg-2.so.2 #9 0x0840e2ec in ?? () #10 0x00000001 in ?? () #11 0x0140f330 in ?? () #12 0x4003c060 in ?? () from /lib/libpthread.so.0
Hmmm there is nothing there that indicates gaim. Was the segfault message the same gmem.c message?
Yep The same. I was able to get gaim do it compiled with debug and running -d. The output is just my buddy list loading and signing on and sitting there idle until it opens a chat window. That's when g_log: Module error: /usr/lib/enchant/libenchant_ispell.so: undefined symbol: __gxx_personality_v0 g_log: Module error: /usr/lib/enchant/libenchant_myspell.so: undefined symbol: __gxx_personality_v0 GLib: gmem.c:157: failed to allocate 2691625484 bytes Aborted Appears in the log file. The # of bytes can vary but I've seen the same #s before. But like you were saying. This looks like it's an isolated glib problem. Trying to remember if I upgraded from gentoo-dev-sources-2.6.3-r2 to 2.6.4-r1 the same time I upgraded glib and gtk+ and got the problem.
I've noticed that compiled with CFLAGS="-march=pentium4 -O2 -pipe" it'll crash in gdb. But with my regular flags it won't. If I run it through gdb it'll work as ordered. Also, any reason it's trying to allocate 2.5GB of memory for a simple chat window? Sounds like an unsigned long gone crazy. That's probably where the failure comes from. And if I had to make a shot in the dark guess, it looks like librsvg is loading one of the SVG graphics that's in the chat window. A bug in there or the interface between glib and librsvg is causing the problem... Sooooo.... I checked it out. And sure enough.... glib-2.4 and gtk+-2.4 want you to rebuild librsvg after you compile them (as noted in the gtk+-2.4 ebuild). However, they rebuild librsvg-2.4.0-r3 and not librsvg-2.6.3 which is the latest stable version for Gnome 2.6. However librsvg-2.6.3 is setup for gtk+-2.4 and glib-2.4 in mind. Soooo... I emerged librsvg-2.6.3 and tried it. Now I didn't get a crash but I got this... free(): invalid pointer 0x840b928! Gaim has segfaulted and attempted to dump a core file. This is a bug in the software and has happened through no fault of your own. It is possible that this bug is already fixed in CVS. If you can reproduce the crash, please notify the gaim maintainers by reporting a bug at http://gaim.sourceforge.net/bug.php Please make sure to specify what you were doing at the time, and post the backtrace from the core file. If you do not know how to get the backtrace, please get instructions at http://gaim.sourceforge.net/gdb.php. If you need further assistance, please IM either RobFlynn or SeanEgn and they can help you. Aborted ah ha! Problem seems to be rooted @ glib and librsvg interaction. The Gnome guys are CC'd on this but I don't know who else should be made aware. So of course it didn't dump a core file even though it should have. So I just ran it in gdb gaim.. And this is what I get. (gdb) run Starting program: /usr/bin/gaim Detaching after fork from child process 23642. Detaching after fork from child process 23643. Detaching after fork from child process 23644. free(): invalid pointer 0x840e2f8! (I hit Ctrl-C after 5min here) Program received signal SIGINT, Interrupt. 0x406d117a in mallopt () from /lib/libc.so.6 (gdb) backtrace #0 0x406d117a in mallopt () from /lib/libc.so.6 #1 0x406d0366 in malloc () from /lib/libc.so.6 #2 0x41149929 in xmlCtxtReadIO () from /usr/lib/libxml2.so.2 #3 0x411f3755 in xmlDictSize () from /usr/lib/libxml2.so.2 #4 0x4121c8c8 in ?? () from /usr/lib/libxml2.so.2 #5 0x00000005 in ?? () #6 0x411f3022 in xmlDictLookup () from /usr/lib/libxml2.so.2 #7 0x084036d8 in ?? () #8 0x4078c430 in __after_morecore_hook () from /lib/libc.so.6 #9 0x08402ff3 in ?? () #10 0x00000080 in ?? () #11 0x083fef50 in ?? () So I did a little "use debug && inherit debug" magic at the top of the librsvg-2.6.3 ebuild. And I got this instead.... (gdb) run Starting program: /usr/bin/gaim Detaching after fork from child process 1653. Detaching after fork from child process 1654. Detaching after fork from child process 1655. free(): invalid pointer 0x8404620! Program received signal SIGSEGV, Segmentation fault. 0x406d16b9 in mallopt () from /lib/libc.so.6 (gdb) backtrace #0 0x406d16b9 in mallopt () from /lib/libc.so.6 #1 0x406d0504 in free () from /lib/libc.so.6 #2 0x40508ef9 in g_free () from /usr/lib/libglib-2.0.so.0 #3 0x405199e9 in g_string_free () from /usr/lib/libglib-2.0.so.0 #4 0x4123c5ac in rsvg_style_handler_free (self=0x408) at rsvg.c:667 #5 0x4123a704 in rsvg_end_element (data=0x8409a10, name=0x8400463 "style") at rsvg.c:1051 #6 0x4114b896 in xmlCtxtReadIO () from /usr/lib/libxml2.so.2 (gdb) So just to test my theory out a little. I added some "use debug && inherit debug" magic to the librsvg-2.4.0-r1 ebuild. And then tried to run gaim. No problems. Gaim has been running solid with 3 convos opened for a couple of minutes. And if I can just have another shot in the dark guess why the librsvg-2.6.3 is failing. It uses an external CSS handler libcrono or something like that. A <style> tag caused the problem in the crash above and that is CSS so... maybe it's something with that external library. I don't know. Don't have time to play with it.
Doug, fantastic sleuthing. Gnome team you want to take this back into your court then?
@comment #16 : the line given at the end of the gtk+ ebuild just makes you rebuild all stable engines, it shouldn't really matter what version you are using vs. what version of gtk/glib. I haven't used gaim in a while, but what svg icons does it use ? Or is this basic gtk theming, so a pixbuf team doesn't show this behaviour ? I'll ask someone of the librsvg team to come and have a look.
I personally don't see it using any SVG icons "qpkg -l gaim | grep svg" but I rebuilt my engines and didn't rebuild librsvg after emerging gtk+-2.4 and there were several icons in Gaim that didn't display. All I got was the white box with the red X on it. Until after I rebuilt librsvg and restarted gaim did they appear. So that's what makes me comment on the use of the librsvg. Also if you notice in the debug backtrace there's references to that library. So your guess is as good as mine why librsvg is used by Gaim. It probably shouldn't be. But you're probably right that it's just basic gtk+ theming.
Here's the answer i got from one of the librsvg developers : <cut> I had a look at that problem for you, personally I don't think it is a librsvg problem. However for some reason when one is running a SVG based theme gaim will crash when only opening a chat window once, however running a pixmap based theme one will need to open the window up, close it and open it up a second time for it to crash in the same manner. I uninstalled librsvg and tried that, it crashed after two attempts. Then I installed librsvg-2.4 and used a SVG theme, it crashed just as fast as it did with 2.7. Gaim does not use librsvg directly or even indirectly through gdk-pixbuf, the only way I can think of this happening is a memory violation in gaim shortly followed by a call to draw the window with gtk or some sort of bug with gdk-pixbuf. Gaim does not use any strange svgs for its graphics or utilise any strange interface. It uses the standard gdk-pixbuf interface that every other gtk app uses with the same svg icons as every other app, I don't see how librsvg could possibly have a bug specific to gaim. As for libcroco, that is not responsible as this bug occurs when librsvg is built without libcroco and even when librsvg is not installed at all. Sorry I'm not much help to you but to be honest I am not sure what I can do. I hope you find the bug, it's really an annoying one. </cut>
Are you all using gnome? I'm on flux and I'm unable to reproduce the crash. Also were the librsvg people using gentoo? If not it sounds like I can consider taking this upstream.
Created attachment 27928 [details] backtrace from gmem.c crash Alright I was able to duplicate this in gnome, where it dawned on me that when you talk about "themes" you meant gnome themes and not gaim themes. I'm attaching my full backtrace, more than likely I'll be sending this upstream to gaim.
http://sourceforge.net/tracker/index.php?func=detail&aid=922518&group_id=235&atid=100235 Filed upstream. Going to hold off on resolving as such until they respond.
Just to add a little tidbit. I'm using Gorilla as my theme. I can try changing to a non-SVG theme and see if the problem happens.
Sure enough. I'm not able to replicate the problem when using a non-SVG theme in Gnome. Basic Formula for Failure: Gtk+-2.4 + Glib-2.4 + Gaim-0.75 + SVG Theme in Gnome = failure
Gaim devs have looked over the backtrace as well as another I have generated and preliminary findings are that this is yet another gtk+ bug, in that gtk+ 2.4 was supposed to be backward compatible and this bug only applies to 2.4. They are going to keep researching to make sure there is nothing that they can do on their end.
Filed with gtk+ bugs. http://bugzilla-test.gnome.org/show_bug.cgi?id=138007
Closing for upstream.
*** Bug 45652 has been marked as a duplicate of this bug. ***
*** Bug 45864 has been marked as a duplicate of this bug. ***
*** Bug 46106 has been marked as a duplicate of this bug. ***
As a sidenote, 0.76 seems to work for me just fine with Gnome 2.6.
I switched back to Gorilla/Industrial after upgrading to Gnome 2.6 and the problem is gone. Also if you notice, --enable-debug is fixed in 0.76 for gtk+-2.4
The problem started for me after installing gnome 2.6 (along with gtk+2.4, glib 2.4, et al). i don't use the "Gorilla" theme, I use Crux in sawfish. It had appeared that everything was working great recently. When I saw today's comment saying that 0.76 seems to have fixed it, I emerged 0.76, and now I can't keep gaim running for longer than 5 minutes at a time. I will emerge with USE=debug and see if I can get a backtrace
Why's it crashing? When you open IM windows? Or it just dies after a couple of min?
Resolved by Gaim-0.76.
To me it appears random; However it could be when someone IMs me, and I just never see those IMs. I tried: USE=debug ACCEPT_KEYWORDS=~x86 emerge gaim and it emerged ok, but running it in gdb still produces no backtrace for some reason. I get a bunch of "no debugging symbols found" messages, and when it crashes, this is what I see: (no debugging symbols found)...(no debugging symbols found)... Program received signal SIGSEGV, Segmentation fault. 0x00000000 in ?? () (gdb) bt #0 0x00000000 in ?? () (gdb) The program is running. Exit anyway? (y or n) y
I re-emerged gaim 0.76, and realized that I hadn't updated the other gaim addons (gaim-encryption, and guifications). After re-emerging those (guifications didn't have an ebuild when I originally installed it), it seems the problem has stabilized now. gaim has been up for 22 hours, so it looks good so far.
