RE: 64bit gdb for KAI/KCC compiler

["Chu, Josh" <josh dot chu at retek dot com>]

Yeah this is for Solaris 64bit.
The KCC support for gdb v5.0 was actually fine for 32bit...it just didn't
compile under 64bit (5.0 doesn't compile for 64bit by itself either).
I'm also checking into the KCC support group to see if they will support
this.  I will post the patches once(if) we get this working
Thanks for taking a look at the code!
--JC

-----Original Message-----
From: Daniel Jacobowitz [mailto:drow@mvista.com]
Sent: Friday, June 28, 2002 8:31 AM
To: Chu, Josh
Cc: 'gdb@sources.redhat.com'
Subject: Re: 64bit gdb for KAI/KCC compiler


On Thu, Jun 27, 2002 at 05:36:03PM -0500, Chu, Josh wrote:
> Ok, so here's the deal : KAI released a patch for gdb 5.0 that will allow
> gdb to read the EDG-style mangled names.  I've attached the patch files
that
> were to be applied to 5.0 (gdb.patch-5.0 and libiberty.patch-5.0).  I
tried
> manually applying these modifications to the 5.2 source, and it builds.
> However, there are some definite problems because gdb will now core dump
> occasionally, not to mention the names are still mangled.
> I've included the patch files to 5.0 and my modifications to the files in
> 5.2 :
>  
> gdb/libiberty:
>     cplus-dem.c
>  
> gdb/gdb:
>     dwarf2read.c
>     hpread.c
>     stabsread.c
>     symfile.c

> After this is done, you can try to build and run.  (export CC='gcc -m64'
and
> ./configure first)

What platform is this, Solaris?  Looks like it from the m64.

> I suppose this is a fairly vague question : but does anyone have a clue
why
> the patches aren't working with 5.2?
> This would be a huge help for anyone who has a KAI compiler on a 64-bit
> system, as it would allow people to debug 64bit programs compiled with
KCC.
> Or if someone who's familiar with the above-mentioned files could comment
on
> the patched files and what could potentially be wrong with them, I would
> very much appreciate it.
> Thanks so much

I'm more curious why they worked with 5.0.  I imagine they had very
limited functionality, since there is a great deal more to do in C++
support than demangle names properly.

I skimmed the patches and nothing jumps out at me as immediately wrong
except for:
!       if ( !EDG_DEMANGLING && !is_edg_temporary_var(sym) )
!         add_symbol_to_list (sym, &local_symbols);

That should probably be || if you ever want to see a local static
variable in GDB.... also, you do not need to comment out
finish_cv_type.



-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer