GCC 3.4 vs GDB 6.1

GCC 3.4 vs GDB 6.1

[Andrew Cagney]

(kind of a bit late but ...)
Are there any "bugs" GDB would like to see fixed in GCC 3.4?

I ask this 'cos GDB starting to really use GCC's DWARF 2 info means (per 
PPC 64) GDB is also identifying occasional problems in GCC's DWARF 2 code.

I'm wondering if there are problems that people know about.

Andrew

Re: GCC 3.4 vs GDB 6.1

[Daniel Jacobowitz]

On Thu, Jan 08, 2004 at 03:30:58PM -0500, Andrew Cagney wrote:
> (kind of a bit late but ...)
> Are there any "bugs" GDB would like to see fixed in GCC 3.4?
> 
> I ask this 'cos GDB starting to really use GCC's DWARF 2 info means (per 
> PPC 64) GDB is also identifying occasional problems in GCC's DWARF 2 code.
> 
> I'm wondering if there are problems that people know about.

For the biggest one I know of, see the thread I started on gcc@ on ...
Monday? or last week? ... about when to emit debug info for C++
classes.  The discussion trailed off without agreement.  After that the
only huge one I knew about was DW_TAG_namespace, and we've got that
covered now.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: GCC 3.4 vs GDB 6.1

[Eli Zaretskii]

> Date: Thu, 08 Jan 2004 15:30:58 -0500
> From: Andrew Cagney <cagney@gnu.org>
> 
> (kind of a bit late but ...)
> Are there any "bugs" GDB would like to see fixed in GCC 3.4?

How about this thingy:

  http://sources.redhat.com/ml/gdb/2003-12/msg00271.html

Does it sound like a GCC bug?

Re: GCC 3.4 vs GDB 6.1

[Daniel Jacobowitz]

On Fri, Jan 09, 2004 at 10:54:57AM +0200, Eli Zaretskii wrote:
> > Date: Thu, 08 Jan 2004 15:30:58 -0500
> > From: Andrew Cagney <cagney@gnu.org>
> > 
> > (kind of a bit late but ...)
> > Are there any "bugs" GDB would like to see fixed in GCC 3.4?
> 
> How about this thingy:
> 
>   http://sources.redhat.com/ml/gdb/2003-12/msg00271.html
> 
> Does it sound like a GCC bug?

No idea without more information.  Did you ever get dwarfdump to work
on COFF?

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: GCC 3.4 vs GDB 6.1

[Eli Zaretskii]

> Date: Fri, 9 Jan 2004 09:36:18 -0500
> From: Daniel Jacobowitz <drow@mvista.com>
> > 
> >   http://sources.redhat.com/ml/gdb/2003-12/msg00271.html
> > 
> > Does it sound like a GCC bug?
> 
> No idea without more information.  Did you ever get dwarfdump to work
> on COFF?

dwarfdump has no hope of ever running on a platform that doesn't
support ELF, at least as an option, since dwarfdump needs libelf to be
built.

I still have plans to build it on GNU/Linux and try to use it against
the binary that gave me trouble on that box; I don't know if it will
be of any use, though.

Are you saying that this problem never happened on an ELF platform,
like on GNU/Linux?

Re: GCC 3.4 vs GDB 6.1

[Daniel Jacobowitz]

On Sat, Jan 10, 2004 at 01:51:00PM +0200, Eli Zaretskii wrote:
> > Date: Fri, 9 Jan 2004 09:36:18 -0500
> > From: Daniel Jacobowitz <drow@mvista.com>
> > > 
> > >   http://sources.redhat.com/ml/gdb/2003-12/msg00271.html
> > > 
> > > Does it sound like a GCC bug?
> > 
> > No idea without more information.  Did you ever get dwarfdump to work
> > on COFF?
> 
> dwarfdump has no hope of ever running on a platform that doesn't
> support ELF, at least as an option, since dwarfdump needs libelf to be
> built.
> 
> I still have plans to build it on GNU/Linux and try to use it against
> the binary that gave me trouble on that box; I don't know if it will
> be of any use, though.
> 
> Are you saying that this problem never happened on an ELF platform,
> like on GNU/Linux?

I have never seen it before with DWARF-2.  So there is no reasonable
way to guess what is wrong without someone debugging it.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: GCC 3.4 vs GDB 6.1

[Michael Elizabeth Chastain]

> (kind of a bit late but ...)
> Are there any "bugs" GDB would like to see fixed in GCC 3.4?

  http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12267
  [3.4 regression] function-at-a-time generates stabs in bad order

This imparirs stabs+ support significantly.

Anonymous unions are another regression area.

I have "compare by gcc" tables but I haven't analyzed them yet.
This one bug causes so many problems that it's hard to look at
other stabs+ problems.

Grossly, there are about 27 new FAILS with gcc HEAD -gdwarf-2
versus gcc 3.3.2 -gdwarf-2, and about 28 new FAILS with gcc HEAD
-gstabs+ versus gcc 3.3.2 -gstabs+.

Michael C