Coda File System

Re: debug logs & proper bug reports?

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Fri, 18 Jan 2002 16:12:34 -0500
On Fri, Jan 18, 2002 at 02:59:36PM -0600, ctest_at_neural.dlsemc.com wrote:
> We are reliably able to kill venus.  in the logs it will say "*****
> FATAL SIGNAL (11) *****".  I am shure that this is not intended
> operation, and would like to submit a bug report.  what should i send
> in?  should i start the server with debugging on?  how much debugging?
> etc.

Signal 11 is a null-pointer dereference. If venus dies like this, it
doesn't matter how much debugging you enable on the server.

The best way I've found to catch these segfaults is to run a venus
with debugging symbols. I'm not sure whether the binaries in the rpm
package are stripped, but I noticed the other day that the binaries in
the debian packages are and thus won't have these debug symbols.

Then attach gdb and try to reproduce the segfault,

  # gdb venus `pidof venus`
  (gdb) continue			# venus was frozen when gdb attached
					# try to reproduce the segfault
  Interrupted, signal 11 		# gdb actually catches the segfault 
  (gdb) bt				# grab the stacktrace. 

The stacktrace will tell me exactly where venus died, and hopefully will
have enough information to help me figure out how to avoid it.

Also a description of how you crash venus might help me debug the
problem here.

Jan
Received on 2002-01-18 16:12:35