0

We are using svn (1.6.3)+sasl integration. After installationon on Suse 10.x (64 bits) I found that svnserve hanged sessions ("ps ax" shows more and more hanged svnserve processes). Xinetd logs contains a lot of lines like this

EXIT: svn signal=11 duration=1(sec)

Svn "make check" shows that all stages passed as "success".

Does anybody try to resolve similar issue? Any help will be appreciated.

Amandasaurus
  • 31,471
  • 65
  • 192
  • 253
FoxyBOA
  • 417
  • 1
  • 6
  • 12
  • Signal 11 on Unix-like systems is SIGSEGV, or "Segmentation fault". – user1686 Jul 31 '09 at 13:49
  • Did you install from source or pre-compiled packages? – Swoogan Jul 31 '09 at 16:42
  • I compile svn from the source code. I found similar issue here and try to recompile svn with native arp, but w/o success - http://linuxadminzone.com/fix-subversion-svn-child-exit-signal-segmentation-fault-error-in-apache/ – FoxyBOA Aug 01 '09 at 07:29

2 Answers2

1

Enable core dumping by calling:

ulimit -c 9999999999

then start svnserve and let it crash (signal 11). You should find a core dump in the directory where you were as svnserve started. Load the core dump using gdb:

gdb --core=core svnserve

where "core" ist the file name of the core dump. gdb should spit out the function where the crash happend. If you need more information execute the "bt" gdb command. This will give you a back trace of the call. Probably you could guess than where the problem is.

If you don't see anything usefull just memory addresses. Try to recompile with debug switches "-g" and optimisation switches turned off "-O0".

Eduard Wirch
  • 352
  • 4
  • 14
0

Founded workaround: daily (cron.daily)

/usr/bin/killall svnserve
FoxyBOA
  • 417
  • 1
  • 6
  • 12