http://sourceware.org/ml/gdb-patches/2012-02/msg00673.html
Subject: [patch] Fix regression by me for gcc-4.0...gcc-4.4 i386 -O2 -g parameters (PR 13777)

Hi,

http://sourceware.org/bugzilla/show_bug.cgi?id=13777

for CentOS-5 i386 ls:

$ gdb ls
(gdb) start
Temporary breakpoint 1, main (argc=Cannot access memory at address 0x81b7c7cd

instead of GDB before my PR 12573 fix/change:

Temporary breakpoint 1, main (argc=1, ...

I asked before on #gcc since which versions GCC produced DW_AT_location which
is for any PC either optimized-out or it has valid value.  I was told since
gcc-4.0.  But that is not true.

I have bisected gcc and I found 4.4.0 was still broken, 4.5.0 was correct,
thanks to:
commit 25e880b1917bd6bbf07e86b5574c698f3e9472d9
Author: rth <rth@138bc75d-0d04-0410-961f-82ee72b054a4>
Date:   Sat May 30 00:33:46 2009 +0000
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@147995 138bc75d-0d04-0410-961f-82ee72b054a4
unwind info for epilogues

Curiously 4.4.x branch got later also fixed in this case by:
commit 61db8bd232daeed3751b43570fab16146145e096
Author: jakub <jakub@138bc75d-0d04-0410-961f-82ee72b054a4>
Date:   Tue Jun 2 07:18:16 2009 +0000
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/branches/gcc-4_4-branch@148070 138bc75d-0d04-0410-961f-82ee72b054a4
(but that missed 4.4.0)

As 4.5.0 was first FSF GCC with VTA I was trusting more the validity only
after VTA and these heuristic results seems to confirm that.

I find the change below definitely safe.

Someone may object the original PR 12573 (do not try to skip prologue for -O2
-g code as it may cause more confusion than anything else) as while it fixed
some GDB crashes there is no such purpose anymore with Tom's
ambiguous-linespec patch.  Still I believe PR 12573 was right to do.

No regressions on {x86_64,x86_64-m32,i686}-fedora17-linux-gnu.

I will check it in.


Thanks,
Jan


gdb/
2012-02-28  Jan Kratochvil  <jan.kratochvil@redhat.com>

	PR symtab/13777
	* dwarf2read.c (process_full_comp_unit): Set LOCATIONS_VALID only for
	GCC >=4.5.

gdb/testsuite/
2012-02-28  Jan Kratochvil  <jan.kratochvil@redhat.com>

	PR symtab/13777
	* gdb.dwarf2/dw2-skip-prologue.S (DW_AT_producer): Set it to 4.5.0.

--- a/gdb/dwarf2read.c
+++ b/gdb/dwarf2read.c
@@ -4841,7 +4841,9 @@ process_full_comp_unit (struct dwarf2_per_cu_data *per_cu)
 
       /* GCC-4.0 has started to support -fvar-tracking.  GCC-3.x still can
 	 produce DW_AT_location with location lists but it can be possibly
-	 invalid without -fvar-tracking.
+	 invalid without -fvar-tracking.  Still up to GCC-4.4.x incl. 4.4.0
+	 there were bugs in prologue debug info, fixed later in GCC-4.5
+	 by "unwind info for epilogues" patch (which is not directly related).
 
 	 For -gdwarf-4 type units LOCATIONS_VALID indication is fortunately not
 	 needed, it would be wrong due to missing DW_AT_producer there.
@@ -4849,7 +4851,7 @@ process_full_comp_unit (struct dwarf2_per_cu_data *per_cu)
 	 Still one can confuse GDB by using non-standard GCC compilation
 	 options - this waits on GCC PR other/32998 (-frecord-gcc-switches).
 	 */ 
-      if (cu->has_loclist && gcc_4_minor >= 0)
+      if (cu->has_loclist && gcc_4_minor >= 5)
 	symtab->locations_valid = 1;
 
       if (gcc_4_minor >= 5)
--- a/gdb/testsuite/gdb.dwarf2/dw2-skip-prologue.S
+++ b/gdb/testsuite/gdb.dwarf2/dw2-skip-prologue.S
@@ -30,7 +30,7 @@
 	.4byte	func_start			/* DW_AT_low_pc */
 	.4byte	func_end			/* DW_AT_high_pc */
 	.ascii	"main.c\0"			/* DW_AT_name */
-	.ascii	"GNU C 4.0.0\0"			/* DW_AT_producer must be >= 4.0  */
+	.ascii	"GNU C 4.5.0\0"			/* DW_AT_producer must be >= 4.5  */
 	.byte	2				/* DW_AT_language (DW_LANG_C) */
 
 	.uleb128	2			/* Abbrev: DW_TAG_subprogram */