quassel crashed with SIGSEGV in _IO_vfprintf_internal()

[Rafael Belmonte]

Quassel crashed and also made Xorg to crash when I switch to full screen mode.

ProblemType: Crash
DistroRelease: Ubuntu 11.10
Package: quassel 0.7.3-0ubuntu1
ProcVersionSignature: Ubuntu 3.0.0-10.16-generic 3.0.4
Uname: Linux 3.0.0-10-generic x86_64
ApportVersion: 1.22.1-0ubuntu2
Architecture: amd64
Date: Mon Sep 12 02:31:35 2011
EcryptfsInUse: Yes
ExecutablePath: /usr/bin/quassel
InstallationMedia: Kubuntu 11.10 "Oneiric Ocelot" - Beta amd64 (20110901)
ProcCmdline: /usr/bin/quassel
SegvAnalysis:
 Segfault happened at: 0x7fdd526ec522 <_IO_vfprintf_internal+50>: mov %eax,-0x518(%rbp)
 PC (0x7fdd526ec522) ok
 source "%eax" ok
 destination "-0x518(%rbp)" (0x7fffda001b28) not located in a known VMA region (needed writable region)!
 Stack memory exhausted (SP below stack segment)
SegvReason: writing unknown VMA
Signal: 11
SourcePackage: quassel
StacktraceTop:
 _IO_vfprintf_internal (s=0x7fffda002050, format=0x7fdd54215e0d "LANGUAGE=%s", ap=0x7fffda0021c8) at vfprintf.c:245
 ___vsnprintf_chk (s=0xe88750 "", maxlen=<optimized out>, flags=1, slen=<optimized out>, format=0x7fdd54215e0d "LANGUAGE=%s", args=0x7fffda0021c8) at vsnprintf_chk.c:65
 ___snprintf_chk (s=<optimized out>, maxlen=<optimized out>, flags=<optimized out>, slen=<optimized out>, format=<optimized out>) at snprintf_chk.c:36
 snprintf (__fmt=0x7fdd54215e0d "LANGUAGE=%s", __n=42, __s=<optimized out>) at /usr/include/x86_64-linux-gnu/bits/stdio2.h:66
 KCatalogPrivate::setupGettextEnv (this=0xe71e80) at ../../kdecore/localization/kcatalog.cpp:183
Title: quassel crashed with SIGSEGV in _IO_vfprintf_internal()
UpgradeStatus: No upgrade log present (probably fresh install)
UserGroups: adm admin cdrom dialout lpadmin plugdev sambashare
XsessionErrors:
 (process:3445): libindicate-WARNING **: Unable to get property data: GDBus.Error:com.canonical.indicate.NO_SUCH_PROPERTY: Indicator 1 has no property named 'icon'
 (process:3445): libindicate-WARNING **: Unable to get property data: GDBus.Error:com.canonical.indicate.NO_SUCH_PROPERTY: Indicator 1 has no property named 'count'

[Apport retracing service (apport)]

StacktraceTop:
 _IO_vfprintf_internal (s=0x7fffda002050, format=0x7fdd54215e0d "LANGUAGE=%s", ap=0x7fffda0021c8) at vfprintf.c:245
 ___vsnprintf_chk (s=0xe88750 "", maxlen=<optimized out>, flags=1, slen=<optimized out>, format=0x7fdd54215e0d "LANGUAGE=%s", args=0x7fffda0021c8) at vsnprintf_chk.c:65
 ___snprintf_chk (s=<optimized out>, maxlen=<optimized out>, flags=<optimized out>, slen=<optimized out>, format=<optimized out>) at snprintf_chk.c:36
 snprintf (__fmt=0x7fdd54215e0d "LANGUAGE=%s", __n=42, __s=<optimized out>) at /usr/include/x86_64-linux-gnu/bits/stdio2.h:66
 KCatalogPrivate::setupGettextEnv (this=0xe71e80) at ../../kdecore/localization/kcatalog.cpp:183

[Rafael Belmonte (eaglescreen)]

I can reproduce this crash always.

[Rafael Belmonte (eaglescreen)]

Hello, this bug is still present in Kubuntu 12.10, when I set Quassel to fullscreen, the X server becomes completely crazzy .
Using Intel graphics card.

[Harald Sitter (apachelogger)]

New backtrace and Xorg.log please.

Also, "the X server becomes completely crazzy ." define crazy please.
Also, what does it do actually 'go crazy' or 'crash', that is a rather profound difference.
Also, does memory/cpu usage change drastically when this happens?

[Rafael Belmonte (eaglescreen)]

Okay, I have done some new tests, in two totally different computers, one with Intel graphics card and the other with ATI/AMD graphics card.
With X server to becomes crazy I mean, all windows starting to flash and being totally unresponsive for instance.
An important discovery is that the bug only occurs when Quassel window is maximized and then you put it in full-screen. Then X server freezes screen image and becomes totally unresponsive to mouse or keyboard events (or in rare occasions, it becomes a little crazy). At this point, with Intel graphics computer, I can jump to a tty and restart kdm service to can use X server again. But in the ATI graphics computer, going to tty causes a total system hang, resulting in a dark but flashing screen, Alt + Impr + k does nothing, but Alt + Impr + b hardly reboots the system.
While X server was frozen, I have take a look at CPU and memory usage with top command, and I did not see any strange.

Can't you reproduce the bug? Can I give any other valuable debug information or backtrace?

[Rafael Belmonte (eaglescreen)]

Last day I was halt asleep and I didn't read the line where you request me a new backtrace and Xorg logs.
I have just attached them, I don't see anything strange in the Xorg logs, and I hope the quassel backtrace to be useful.
If you think I need to install more debugging symbols to have a better backtrace, please, tell me.
Thanks.

[Harald Sitter (apachelogger)]

I know the cause and have a fix pending, thanks for the information.

It has to do with how QWidget's setfullscreen behaves with regards to KDE's toggle action system. Quassel currently does not follow the KDE fullscreen toggle actions documentation in that one should use its functions to turn on/off fullscreen rather than the regular QWidget ones as the latter include additional changes that are unrelated and more importantly will cause syncness issues.

[Harald Sitter (apachelogger)]

FWIW quassel basically drowns xorg in silly requests which then renders xorg (or drivers) in a state where it in fact goes crazy. Since it is only semi-related to the issue reported I removed an xorg affectness. Most likely that however happens due to stack exhaustion where xlib then starts reading corrupted memory or something equally random. Should it be a reproducible with some simple xlib test program that shoudl be reported separately as the presented bug is really about quassel not following KDE's toggle action documentation.