RE: [libunwind] libunwind segv with gcc 2.96 programs run on RedhatEL 3 with GLIBC 2.3.2

[David Mosberger]

>>>>> On 11 Feb 2004 11:04:52 -0800, Jim Wilson <address@hidden> said:

  Jim> If libunwind segfaults or aborts because of errors in the unwind info,
  Jim> then that would be a problem for gdb, as that would kill gdb.  It would
  Jim> be better if libunwind returned an error indication, but it isn't clear
  Jim> if there is a problem, or whether it is worthwhile to fix.

If libunwind segfaulted in the gdb-case (which means remote, not local
debugging), that would almost certainly be a libunwind bug which I'd
want to fix.  I don't think it would segfault though, it would most
likely just cause a bad backtrace to be shown.

  Jim> If there is debug info, then you can example the dwarf2 producer string
  Jim> to find which compiler generated it.  This is probably not very
  Jim> convenient.

  Jim> address@hidden gcc]$ readelf -dw a.out | grep producer
  Jim> DW_AT_producer    : (indirect string, offset: 0x683): GNU C 3.2

  Jim> 20020903 (Red Hat Linux 8.0 3.2-7)

Cool trick! ;-)

        --david