Version: KMail 1.12.4 (using KDE 4.3.5) Installed from: Ubuntu Packages Sometimes when I open the messages composer (either by selecting "New message" or by hitting "reply" in an existing message), the whole X server is crashing. I am usually writing 5 to 15 mails each day, and this happens every 1 to 2 weeks - but when it happens, it's of course really annoying as *all* applications are terminated. The .xsession-errors file contains some messages like this: kdeinit4: Fatal IO error: client killed kdeinit4: sending SIGHUP to children. but since there is no timestamp, it's hard to tell when they occurred. The Xserver log file contains this backtrace: Backtrace: 0: /usr/bin/X(xorg_backtrace+0x3b) [0x8133d6b] 1: /usr/bin/X(xf86SigHandler+0x55) [0x80c7d35] 2: [0x739400] 3: /usr/bin/X(FindGlyphByHash+0x3c) [0x816e71c] 4: /usr/bin/X [0x817b13c] 5: /usr/bin/X [0x81742e5] 6: /usr/bin/X(Dispatch+0x35f) [0x808d17f] 7: /usr/bin/X(main+0x395) [0x8072515] 8: /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe6) [0x22cb56] 9: /usr/bin/X [0x80719c1] Saw signal 11. Server aborting. Which is always the same except for the hex numbers in line 2 and 8. This backtrace is of course rather useless without debug symbols, but I don't know how to install them and get a good backtrace without having to compile X myself. I am using Kubuntu 9.10 and X.Org X Server 1.6.4. KMail is the only application which ever triggered an Xserver crash so far.
I can confirm this on KMail 1.12.4, KDE 4.3.4, X.org 1.7.5 (Debian Testing). X.org log file contains the following: Backtrace: 0: /usr/bin/X (xorg_backtrace+0x3b) [0x80e68cb] 1: /usr/bin/X (0x8048000+0x60ad5) [0x80a8ad5] 2: (vdso) (__kernel_rt_sigreturn+0x0) [0xb77bb40c] 3: /usr/bin/X (FindGlyphByHash+0x3c) [0x81b2edc] 4: /usr/bin/X (0x8048000+0xcfa22) [0x8117a22] 5: /usr/bin/X (0x8048000+0xca553) [0x8112553] 6: /usr/bin/X (0x8048000+0x2afa7) [0x8072fa7] 7: /usr/bin/X (0x8048000+0x1e95a) [0x806695a] 8: /lib/i686/cmov/libc.so.6 (__libc_start_main+0xe5) [0xb74cfb55] 9: /usr/bin/X (0x8048000+0x1e541) [0x8066541] Segmentation fault at address 0x4 I just installed (what appears to be) the X.org debugging symbols (package xserver-xorg-core-dbg) and will try to reproduce.
Apparently installing the dbg packet you mentioned is not enough to get a good backtrace - it just crashed again here, with this result: Backtrace: 0: /usr/bin/X(xorg_backtrace+0x3b) [0x8133d6b] 1: /usr/bin/X(xf86SigHandler+0x55) [0x80c7d35] 2: [0x9b1400] 3: /usr/bin/X(FindGlyphByHash+0x3c) [0x816e71c] 4: /usr/bin/X [0x817b13c] 5: /usr/bin/X [0x81742e5] 6: /usr/bin/X(Dispatch+0x35f) [0x808d17f] 7: /usr/bin/X(main+0x395) [0x8072515] 8: /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe6) [0x213b56] 9: /usr/bin/X [0x80719c1] Saw signal 11. Server aborting. In order to get the actual source lines, I tried to use addr2line on the Xorg binary contained in the dbg packet, like this: addr2line -fe /usr/lib/debug/usr/bin/Xorg 0x816e71c I am not sure if that is correct, it gives this result: 3: FindGlyphByHash /build/buildd/xorg-server-1.6.4/obj-i486-linux-gnu/render/../../render/glyph.c:263 4: ProcRenderAddGlyphs /build/buildd/xorg-server-1.6.4/obj-i486-linux-gnu/render/../../render/render.c:1131 5: ProcRenderDispatch /build/buildd/xorg-server-1.6.4/obj-i486-linux-gnu/render/../../render/render.c:2092 6: Dispatch /build/buildd/xorg-server-1.6.4/obj-i486-linux-gnu/dix/../../dix/dispatch.c:457 7: main /build/buildd/xorg-server-1.6.4/obj-i486-linux-gnu/dix/../../dix/main.c:399
Got a crash again today. Actually, it happened when I tried to compose an answer to Ralf's last comment - KMail has a strange sense of humor it seems. The backtrace was still cryptic, but I tried addr2line as well and I am attaching the results: Backtrace: 0: /usr/bin/X (xorg_backtrace+0x3b) [0x80e696b] 1: /usr/bin/X (0x8048000+0x60b75) [0x80a8b75] 2: (vdso) (__kernel_rt_sigreturn+0x0) [0xb770e40c] 3: /usr/bin/X (FindGlyphByHash+0x3c) [0x81b346c] 4: /usr/bin/X (0x8048000+0xcfa22) [0x8117a22] 5: /usr/bin/X (0x8048000+0xca553) [0x8112553] 6: /usr/bin/X (0x8048000+0x2b017) [0x8073017] 7: /usr/bin/X (0x8048000+0x1e95a) [0x806695a] 8: /lib/i686/cmov/libc.so.6 (__libc_start_main+0xe5) [0xb7423b55] 9: /usr/bin/X (0x8048000+0x1e541) [0x8066541] Segmentation fault at address 0x4 addr2line output (same command as Ralf): 0: xorg_backtrace /home/bgoglin/debian/git/xserver/xorg-server-1.7.5/build/os/../../os/backtrace.c:46 1: OsSigHandler /home/bgoglin/debian/git/xserver/xorg-server-1.7.5/build/os/../../os/osinit.c:142 3: FindGlyphByHash /home/bgoglin/debian/git/xserver/xorg-server-1.7.5/build/render/../../render/glyph.c:263 4: ProcRenderAddGlyphs /home/bgoglin/debian/git/xserver/xorg-server-1.7.5/build/render/../../render/render.c:1107 5: ProcRenderDispatch /home/bgoglin/debian/git/xserver/xorg-server-1.7.5/build/render/../../render/render.c:2059 6: Dispatch /home/bgoglin/debian/git/xserver/xorg-server-1.7.5/build/dix/../../dix/dispatch.c:440 7: main /home/bgoglin/debian/git/xserver/xorg-server-1.7.5/build/dix/../../dix/main.c:287 Although I'm not sure it's the same binary (backtrace mentions "X", while the debugging symbols refer to "Xorg"). I forgot to mention that I've been experiencing the same crash using both fglrx and radeonhd drivers, therefore it's probably not driver-related. I'll try attaching gdb to X before trying to reproduce it next time, as explained here: http://wiki.debian.org/XStrikeForce/XserverDebugging.
The crash itself happens in 3: FindGlyphByHash, frame 1 and 2 are just the Xorg code that actually prints out the backtrace. There's supposed to be nothing an X client can do to crash the X server, so this is an Xorg bug. Given that it comes from a function trying to find glyphs I would assume you have a corrupted font glyph cache somehow. So you may want to try running fc-cache --really-force to rebuild your font caches and then restart KDE. Either way it's not a KDE bug, although if you find some way we're abusing X that we shouldn't be then please reopen this bug with details. Thanks.
The Xorg bugtracker seems to be https://bugs.freedesktop.org/ Manolis, do you have an account there? I don't, and I think we should not both report the bug there :D
Manolis created an upstream report at https://bugs.freedesktop.org/show_bug.cgi?id=26801 , which was marked as duplicate of https://bugs.freedesktop.org/show_bug.cgi?id=20718