gdb registers multiple definition of symbols : patch to fix this

gdb registers multiple definition of symbols : patch to fix this

[Kalpana Ramamurthy]

Hello,

My code is written for 64 bit support. I have upgraded the compiler to 3.4.6 
which was previously 3.2.2. I think the default debug information generated 
is dwarf-2 (the code was compiled with -ggdb option). I have a class 
definition containing about 10 fields. But when I say "ptype cGroupCache", 
it shows me only 1 field !. This happened to a some of the other classes too 
that I tested. But in those cases, when I start gdb with --readnow option, I 
see all the fields. But for this particular class, --readnow option doesn't 
help either.

Why this patch is required with the code is compiled with gcc 3.4.6 is 
something that I still don't
know.

Currently to workaround this problem, I patched the symtab.c code with the 
following algorithm in the function lookup_symbol_aux_symtabs().

1. Browse across all the definitions of the symbol
2. For each definition found, get the number of fields
3. If the number is greater thatn the number of fields of the last saved 
defintion, save the definition.
4. At the end of the browsing, return the saved definition (if it exists).

Please note that though this patch fixes the problem, its only a workaround. 
The root cause is still
unknown. Why is gdb registering multiple definitions of the symbol ? And 
even if that's necessary,
why is it not picking the correct definition ?

Code snippet that was patched :
***************************************************
/* Add declarations for sym1, block1, s1 and maxfields */

struct symbol * sym, *sym1;
const struct block *block, *block1 ;
struct symtab *s, *s1 ;
int maxfields = -1;

/* Initialize sym1 to NULL */
sym1 = NULL ;

ALL_SYMTABS (objfile, s)
{
    bv = BLOCKVECTOR (s)
    block = BLOCKVECTOR_BLOCK (bv, block_index) ;
    sym = lookup_block_symbol (block, name, linkage_name, domain);

    /* PATCH begins */
    if (sym)
    {
         int fields = 0 ;
         if (sym->type && sym->type->main_type)
          {
               fields = sym->type->main_type->nfields;
          }
          if (fields >= maxfields)
            {
                   maxfields = fields ;
                   sym1 = sym ;
                   block1 = block ;
                   s1 = s ;
             }
        }
        if (sym1)
           {
                 block_found = block1 ;
                 if (symtab != NULL)
                       *symtab = s1 ;
                 sym = fixup_symbol_section (sym1, objfile) ;
                 return (sym1) ;
            }
}
*******************************************************

_________________________________________________________________
Catch all the cricketing action right here. Live score, match reports, 
photos et al. http://content.msn.co.in/Sports/Cricket/Default.aspx

Re: gdb registers multiple definition of symbols : patch to fix this

[Daniel Jacobowitz]

On Tue, Jul 31, 2007 at 11:53:23AM +0000, Kalpana Ramamurthy wrote:
> Why this patch is required with the code is compiled with gcc 3.4.6 is 
> something that I still don't
> know.

It sounds to me like your compiler is generating bogus information.

-- 
Daniel Jacobowitz
CodeSourcery

Re: gdb registers multiple definition of symbols : patch to fixthis

[Kalpana Ramamurthy]

How do we verify the information generated by the compiler ? I wanted to see 
all the definitions
generated by the compiler for that symbol. How can I do that ? I couldn't 
decipher anything
with readelf and objdump.

Regards,
Kalpana
====


>From: Daniel Jacobowitz <drow@false.org>
>To: Kalpana Ramamurthy <kal_pana@hotmail.com>
>CC: gdb@sourceware.org, kalpana.balaji@symphonysv.com
>Subject: Re: gdb registers multiple definition of symbols : patch to 
>fixthis
>Date: Tue, 31 Jul 2007 08:02:44 -0400
>
>On Tue, Jul 31, 2007 at 11:53:23AM +0000, Kalpana Ramamurthy wrote:
> > Why this patch is required with the code is compiled with gcc 3.4.6 is
> > something that I still don't
> > know.
>
>It sounds to me like your compiler is generating bogus information.
>
>--
>Daniel Jacobowitz
>CodeSourcery

_________________________________________________________________
Real Estate classifieds on MSN - for free.www.yello.in 
http://www.yello.in/home.php?utm_source=hotmailtag&utm_medium=email&utm_content=in&utm_campaign=jun

Re: gdb registers multiple definition of symbols : patch to fixthis

[Daniel Jacobowitz]

On Wed, Aug 01, 2007 at 04:44:58AM +0000, Kalpana Ramamurthy wrote:
> How do we verify the information generated by the compiler ? I wanted to see 
> all the definitions
> generated by the compiler for that symbol. How can I do that ? I couldn't 
> decipher anything
> with readelf and objdump.

Sorry, those are the tools you need to use.

-- 
Daniel Jacobowitz
CodeSourcery