[obish/dwarf2] Describe frame-base hack

[obish/dwarf2] Describe frame-base hack

[Andrew Cagney]

[-- Attachment #1: Type: text/plain, Size: 126 bytes --]

Hello,

Stumbled across this during testing of symbol_ops.  This adds a comment 
alerting the reader as to a problem.

Andrew

[-- Attachment #2: diffs --]
[-- Type: text/plain, Size: 1195 bytes --]

Index: ChangeLog
2004-01-26  Andrew Cagney  <cagney@redhat.com>

	* dwarf2read.c (read_func_scope): Document hack.

Index: dwarf2read.c
===================================================================
RCS file: /cvs/src/src/gdb/dwarf2read.c,v
retrieving revision 1.124
diff -u -r1.124 dwarf2read.c
--- dwarf2read.c	23 Jan 2004 22:41:28 -0000	1.124
+++ dwarf2read.c	26 Jan 2004 19:54:06 -0000
@@ -2211,6 +2211,15 @@
      it.  */
   attr = dwarf_attr (die, DW_AT_frame_base);
   if (attr)
+    /* FIXME: cagney/2004-01-26: The DW_AT_frame_base's location
+       expression is being recorded directly in the function's symbol
+       and not in a separate frame-base object.  I guess this hack is
+       to avoid adding some sort of frame-base adjunct/annex to the
+       function's symbol :-(.  The problem with doing this is that it
+       results in a function symbol with a location expression that
+       has nothing to do with the location of the function, ouch!  The
+       relationship should be: a function's symbol has-a frame base; a
+       frame-base has-a location expression.  */
     dwarf2_symbol_mark_computed (attr, new->name, cu);
 
   list_in_scope = &local_symbols;

Re: [obish/dwarf2] Describe frame-base hack

[Daniel Jacobowitz]

On Mon, Jan 26, 2004 at 02:57:18PM -0500, Andrew Cagney wrote:
> Hello,
> 
> Stumbled across this during testing of symbol_ops.  This adds a comment 
> alerting the reader as to a problem.

There is already a FIXME describing this in symtab.h which answers your
guess below.

> Index: ChangeLog
> 2004-01-26  Andrew Cagney  <cagney@redhat.com>
> 
> 	* dwarf2read.c (read_func_scope): Document hack.
> 
> Index: dwarf2read.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/dwarf2read.c,v
> retrieving revision 1.124
> diff -u -r1.124 dwarf2read.c
> --- dwarf2read.c	23 Jan 2004 22:41:28 -0000	1.124
> +++ dwarf2read.c	26 Jan 2004 19:54:06 -0000
> @@ -2211,6 +2211,15 @@
>       it.  */
>    attr = dwarf_attr (die, DW_AT_frame_base);
>    if (attr)
> +    /* FIXME: cagney/2004-01-26: The DW_AT_frame_base's location
> +       expression is being recorded directly in the function's symbol
> +       and not in a separate frame-base object.  I guess this hack is
> +       to avoid adding some sort of frame-base adjunct/annex to the
> +       function's symbol :-(.  The problem with doing this is that it
> +       results in a function symbol with a location expression that
> +       has nothing to do with the location of the function, ouch!  The
> +       relationship should be: a function's symbol has-a frame base; a
> +       frame-base has-a location expression.  */
>      dwarf2_symbol_mark_computed (attr, new->name, cu);
>  
>    list_in_scope = &local_symbols;


-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer