Re: [libunwind] endless backtrace

[David Mosberger]

>>>>> On Thu, 18 Nov 2004 14:39:50 -0500 (EST), Richard C Bilson 
>>>>> <address@hidden> said:

  Richard> I've been noticing strange behavior with relatively simple
  Richard> local unwinding.  Sometimes my attempts to construct a
  Richard> backtrace will result in endless cycling at the bottom of
  Richard> the stack.  Typically, my code traces down to the bottom of
  Richard> the stack (to a function such as __do_global_ctors_aux)
  Richard> using unw_step, but rather than unw_step returning 0 it
  Richard> returns a positive value, causing my code to keep looping,
  Richard> revisiting the same frame over and over.

  Richard> Obviously, I could put a hack in to detect this and stop
  Richard> looping, but I'd rather find some piece of code to blame
  Richard> and hopefully fix.  I'm doing this on an Itanium-2 Debian
  Richard> unstable system (i.e. gcc 3.3.5, glibc 2.3.2).  I've tried
  Richard> both the system libunwind and versions of 0.98.1 and 0.98.2
  Richard> that I compiled myself, and the problem still arises.

Most probably you're hitting a known bug in Debian/unstable.

See the patches in this mail:

        http://lists.debian.org/debian-ia64/2004/10/msg00026.html

With those patches applied, you should get proper backtraces.

I suspect the particular problem you're hitting is that the current
Debian/unstable GCC doesn't have unwind info for crt{begin,end}.

        --david