[Bug debug/13231] Bad DWARF2 CFI when there is no frame pointer

["kettenis at chello dot nl" <gcc-bugzilla at gcc dot gnu dot org>]

------- Additional Comments From kettenis at chello dot nl  2003-12-12 15:26 -------
Subject: Re:  Bad DWARF2 CFI when there is no frame pointer

   Date: 2 Dec 2003 00:00:53 -0000
   From: "rth at gcc dot gnu dot org" <gcc-bugzilla@gcc.gnu.org>

   Try this.  Certainly it removes the annotation for the instruction 
   you pointed out, but things still don't seem quite right.  The backtrace
   I get out of gdb skips main.  Of course, this may be a separate problem;
   if so we should handle that with a separate PR.

Sorry, but it took a bit of time before I was able to test the patch.
I've tested the patch with mainline gcc-3.4 on
x86_64-unknown-freebsd5.2 and with the gcc-3.3 branch in
i386-unknown-freebsd4.7 (with -fomit-frame-pointer).  It's defenitely
an improvement, although -fomit-frame-pointer on i386 still seems to
have an awful lot of other debug-info related problems.  On x86_64 I'm
now able to get some meaningful backtraces when debugging GDB itself.

About the backtrace skipping main: I recently (2003-11-19) fixed a bug
in GDB where it was skipping main.

   Anyway, I would like to know the effect of this patch on the gdb testsuite.

The effects are minimal.  I have one FAIL->PASS in selftest.exp.  The
rest of the testsuite seems unaffected by this patch.

AFIAC this patch can go in.  Thanks,

Mark


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13231