Re: Is GDB broken on ia64?

[David Mosberger <davidm@napali.hpl.hp.com>]

>>>>> On Thu, 19 Feb 2004 14:14:59 -0800, Richard Harke <rharke@earthlink.net> said:

  Richard> When I look at the back trace, the first three entries look
  Richard> legit but higher up, the back trace has call sequences that
  Richard> are not possible.

Older GDB use code-reading to create a backtrace, which can't be
reliable, especially with newer compilers.  The latest GDB will use
libunwind to build the backtrace.  I don't know if any distro has
picked that up yet, but it's definitely in the CVS tree.

  Richard> But the call parameters printed out at the breakpoint are
  Richard> completely different from the call parameters printed after
  Richard> the seg fault.

Depends on the details, but optimizations can confuse GDB in this fashion.

  Richard> I have trouble stepping thru C code. It seems to jump around
  Richard> in a manner inconsistent with teh code.

That's normal when code gets rescheduled.  The easiest way to avoid
this problem is to build with -g -O (instead of "-g -O2").

There is also a debugger from Intel (idb), which has
gdb-compatibility, but I never tried it myself.

	--david