Hi,

We are facing the problem that a post-mortem investigation of a
multi-threaded program is not possible under RedHat Linux 7.1 (kernel
version 2.4.2-2), because loading the core into gdb 5.0 does not show
the correct crash location.

Attached is a test program, linux.c.
This is a producer-consumer program, with pairs of threads sending each
other messages. Its structure is described in comments at the very
beginning of source code. If 0 is typed as stdin, a divide-by-zero
exception happens.

If you use GDB to look into the core, you'll see it gets some errors
while loading, and doesn't show the correct line which causes the
arithmetic exception:

rock[19] gdb yahuitest/p15 core
GNU gdb 5.0rh-5 Red Hat Linux 7.1
Copyright 2001 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you
are
welcome to change it and/or distribute copies of it under certain
conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for
details.
This GDB was configured as "i386-redhat-linux"...
Core was generated by `yahuitest/p15 100000'.
Program terminated with signal 6, Aborted.
Reading symbols from /lib/i686/libpthread.so.0...done.

warning: Unable to set global thread event mask: generic error
[New Thread 1024 (LWP 5458)]
Error while reading shared library symbols:
Can't attach LWP 5458: No such process
...

Obviously, if one runs the program in the debugger, it shows the correct
location.

Another problem is that core is not always dumpable. I repeat running
this program with 0.csh script as attached. You'll find it stops with no
core dumped after few iterations.

Does anyone know if this is a problem that we can work-around with the
current versions of the kernel (eg. setting a signal handler, etc)? Do
we need a newer version of GDB?

Thanks.

Yahui