Re: BUG: g++ 3.x.x -fomit-frame-pointer: exception handling doesn't work

[Denis Zaitsev <zzz at anda dot ru>]

On Mon, Aug 09, 2004 at 09:16:57PM -0700, James E Wilson wrote:
> I wasn't able to reproduce the first problem you reported.  But I was 
> able to eventually reproduce the problem with the second example.  I 
> configured gcc using
> CFLAGS="long string of options" CXXFLAGS="long string of options" 
> ../gcc/configure ...
> where "long string of options" came from your message.  When I built gcc 
> this way, your new[-1] example did fail.  Unfortunately, -g was not in 
> your list of options, so I was unable to debug the result.
> 
> I never did get the first example to fail.

It's very strange.  Did you use -O2?  The both examples utilize the
same error...

> At this point, I've spent about as much time on this as I care to.  I 
> suggest filing a bug report, or trying to debug this yourself.  If you 
> file a bug report, be sure to include enough info for someone else to 
> reproduce the problem.  If the problem is that libsupc++ has 
> dependencies on being compiled with a frame pointer, then maybe 
> something as simple as adding -fno-omit-frame-pointer to CFLAGS in the 
> libsupc++ Makefile would solve the problem.

No.  Adding the option will not cure the error, but just hide it.
And, more over, not only libsupc++, but any C++ library compiled along
with GCC should suffer from the bug.  As for now, I'm nearly sure that
it's DWARF-related problem.  I has debugged something alike some time
ago.  And the current picture is very similar to the previous one.
So, I will try to debug this problem.  But can you please make some
navigation for me: where in GCC sources is the dwarf info (used for
the exception handling) generated and emited into the object code?  I
surely don't remember where I've found it last time.  And sources are
_huge_...

Thanks in advance.