[PATCH] Don't add DW_AT_{main_subprogram,calling_convention} attributes more than once (PR debug/60152)

[PATCH] Don't add DW_AT_{main_subprogram,calling_convention} attributes more than once (PR debug/60152)

[Jakub Jelinek]

Hi!

gen_subprogram_die is often called more than once for the same decl
(e.g. the first time through force_decl_die etc.), but it always
unconditionally calls add_calling_convention_attribute which thus
may add the attributes several times.

Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
trunk?

Alternatively, we could avoid the add_calling_convention_attribute
call in gen_subprogram_die if subr_die == orig_die, i.e. if we have
already processed the decl once, hopefully the calling conventions would be
the same in all cases.

2014-02-12  Jakub Jelinek  <jakub@redhat.com>

	PR debug/60152
	* dwarf2out.c (add_calling_convention_attribute): Don't add
	DW_AT_main_subprogram or DW_AT_calling_convention attributes
	if they are already present.

--- gcc/dwarf2out.c.jj	2014-01-30 20:31:19.000000000 +0100
+++ gcc/dwarf2out.c	2014-02-12 18:26:41.098486306 +0100
@@ -16873,13 +16873,15 @@ add_calling_convention_attribute (dw_die
 	rely on the old way, which we thus keep.  */
       value = DW_CC_program;
 
-      if (dwarf_version >= 4 || !dwarf_strict)
+      if ((dwarf_version >= 4 || !dwarf_strict)
+	  && !get_AT (subr_die, DW_AT_main_subprogram))
 	add_AT_flag (subr_die, DW_AT_main_subprogram, 1);
     }
 
   /* Only add the attribute if the backend requests it, and
      is not DW_CC_normal.  */
-  if (value && (value != DW_CC_normal))
+  if (value && value != DW_CC_normal
+      && !get_AT (subr_die, DW_AT_calling_convention))
     add_AT_unsigned (subr_die, DW_AT_calling_convention, value);
 }
 

	Jakub

Re: [PATCH] Don't add DW_AT_{main_subprogram,calling_convention} attributes more than once (PR debug/60152)

[Cary Coutant]

> gen_subprogram_die is often called more than once for the same decl
> (e.g. the first time through force_decl_die etc.), but it always
> unconditionally calls add_calling_convention_attribute which thus
> may add the attributes several times.
>
> Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
> trunk?

Looks good to me.

> Alternatively, we could avoid the add_calling_convention_attribute
> call in gen_subprogram_die if subr_die == orig_die, i.e. if we have
> already processed the decl once, hopefully the calling conventions would be
> the same in all cases.

s/orig_die/old_die/

This sounds a little more efficient -- no calls to get_AT just to see
if the attribute has already been added.

I've got a slight preference for your alternate proposal (assuming it works).

-cary