Re: [PATCH RFA] dbxread.c: Complain when local symbols are discarded

[Jim Blandy <jimb@redhat.com>]

Looks good to me.

Kevin Buettner <kevinb@redhat.com> writes:

> I've found that some older versions of gcc incorrectly output N_LSYM
> entries after the corresponding (enclosing) N_LBRAC entry.  The real
> fix for this problem is to patch gcc with the following patch from
> Richard Henderson:
> 
>     http://gcc.gnu.org/ml/gcc/2000-12/msg00870.html
> 
> However, it seems to me that GDB should not be silently discarding
> local symbols, hence the reason for the new complaint.
> 
> Okay to commit?
> 
> 	* dbxread.c (discarding_local_symbols_complaint): New complaint.
> 	(process_one_symbol): Complain about discarding local symbols
> 	due to a misplaced N_LBRAC entry.
> 
> Index: dbxread.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/dbxread.c,v
> retrieving revision 1.32
> diff -u -p -r1.32 dbxread.c
> --- dbxread.c	24 Apr 2002 08:00:54 -0000	1.32
> +++ dbxread.c	8 May 2002 23:54:30 -0000
> @@ -203,6 +203,9 @@ struct complaint repeated_header_complai
>  
>  struct complaint unclaimed_bincl_complaint =
>  {"N_BINCL %s not in entries for any file, at symtab pos %d", 0, 0};
> +
> +struct complaint discarding_local_symbols_complaint =
> +{"misplaced N_LBRAC entry; discarding local symbols which have no enclosing block", 0, 0};
>  \f
>  /* find_text_range --- find start and end of loadable code sections
>  
> @@ -2881,7 +2884,21 @@ process_one_symbol (int type, int desc, 
>        /* Can only use new->locals as local symbols here if we're in
>           gcc or on a machine that puts them before the lbrack.  */
>        if (!VARIABLES_INSIDE_BLOCK (desc, processing_gcc_compilation))
> -	local_symbols = new->locals;
> +	{
> +	  if (local_symbols != NULL)
> +	    {
> +	      /* GCC development snapshots from March to December of
> +		 2000 would output N_LSYM entries after N_LBRAC
> +		 entries.  As a consequence, these symbols are simply
> +		 discarded.  Complain if this is the case.  Note that
> +		 there are some compilers which legitimately put local
> +		 symbols within an LBRAC/RBRAC block; this complaint
> +		 might also help sort out problems in which
> +		 VARIABLES_INSIDE_BLOCK is incorrectly defined.  */
> +	      complain (&discarding_local_symbols_complaint);
> +	    }
> +	  local_symbols = new->locals;
> +	}
>  
>        if (context_stack_depth
>  	  > !VARIABLES_INSIDE_BLOCK (desc, processing_gcc_compilation))