This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

PATCH RFC: Fix PR debug/26964: Avoid duplicate debug info for enums


This is a patch for PR debug/26964, which I reported about generating
duplicate debug info for enum constants defined in a namespace.  From
the PR, a case like this

namespace { enum x { i = 1 }; }
x y;

generates duplicate information for the enum constant 'i'.  This type
of enum definition occurs quite a lot in libstdc++ header files.  If
you look at readelf --debug-dump of a file with libstdc++ debugging
info, you will see a lot of duplication in the DW_TAG_enumerator
lists.

I plan to check this in a couple of days to mainline.  I'm waiting to
see if anybody has any comments before I commit it.

I believe this is appropriate for the 4.1 and 4.2 branches as well.
It is arguably a regression in debug info size.  I will commit it to
those branches as well unless somebody objects.

Tested with bootstrap and testsuite run on i686-pc-linux-gnu.

Ian


2006-12-21  Ian Lance Taylor  <iant@google.com>

	PR debug/26964
	* dwarf2out.c (gen_type_die): Don't write out a DIE for
	ENUMERAL_TYPE if it was already written out.


Index: gcc/dwarf2out.c
===================================================================
--- gcc/dwarf2out.c	(revision 120114)
+++ gcc/dwarf2out.c	(working copy)
@@ -12727,7 +12727,12 @@ gen_type_die (tree type, dw_die_ref cont
 	}
 
       if (TREE_CODE (type) == ENUMERAL_TYPE)
-	gen_enumeration_type_die (type, context_die);
+	{
+	  /* This might have been written out by the call to
+	     declare_in_namespace.  */
+	  if (!TREE_ASM_WRITTEN (type))
+	    gen_enumeration_type_die (type, context_die);
+	}
       else
 	gen_struct_or_union_type_die (type, context_die);
 


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]