This is a regression present on mainline and 4.7 branch and a variant of:
http://gcc.gnu.org/ml/gcc-patches/2012-05/msg01076.html
The difference is that there is a barrier between the CALL_INSN and the
CALL_ARG_LOCATION note (delete_related_insns is called from reorg.c and the dbr
pass is run after the barriers pass, which puts the barrier there on purpose).
Tested on x86_64-suse-linux, applied on the mainline and 4.7 branch.
2012-08-25 Eric Botcazou <ebotca...@adacore.com>
PR rtl-optimization/54088
* jump.c (delete_related_insns): Robustify latest change.
--
Eric Botcazou
Index: jump.c
===================================================================
--- jump.c (revision 190665)
+++ jump.c (working copy)
@@ -1260,9 +1260,9 @@ delete_related_insns (rtx insn)
&& GET_CODE (PATTERN (insn)) == SEQUENCE
&& CALL_P (XVECEXP (PATTERN (insn), 0, 0))))
{
- rtx p = insn;
+ rtx p;
- for (p = NEXT_INSN (p);
+ for (p = next && INSN_DELETED_P (next) ? NEXT_INSN (next) : next;
p && NOTE_P (p);
p = NEXT_INSN (p))
if (NOTE_KIND (p) == NOTE_INSN_CALL_ARG_LOCATION)
