[Bug rtl-optimization/56131] [4.8 regression] gcc.dg/pr56035.c ICEs gcc on sparc-linux
vries at gcc dot gnu.org
gcc-bugzilla@gcc.gnu.org
Fri Feb 22 10:30:00 GMT 2013
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56131
--- Comment #14 from vries at gcc dot gnu.org 2013-02-22 10:30:06 UTC ---
Steven,
thanks for the comments.
> Nothing even trying to look at the CFG after freeing it, so the looks at
> BLOCK_FOR_INSN in delete_insn are non-sense.
AFAIU now from
http://gcc.gnu.org/onlinedocs/gccint/Maintaining-the-CFG.html#Maintaining-the-CFG,
BLOCK_FOR_INSN == NULL shows whether the CFG has been freed or not, so I'd say
it makes sense to test for equality of BLOCK_FOR_INSN when non-NULL.
> Looking for the basic block
> anywhere at all at this point makes no sense,
If I understand you correctly with 'this point' you mean 'after the CFG has
been freed'?
> basic block contents and
> boundaries are not maintained and may be scrambled enough to make even
> the basic block notes unreliable.
OK. I've added a comment in insn-notes.def in the patch below to make that more
explicit.
> Also, "If the label is not marked with a bb, assume it's the same bb"
> is wrong if the label is a marker for a constant pool or a jump table.
OK, fixed in patch below.
Does this patch address all your concerns?
...
Index: gcc/insn-notes.def
===================================================================
--- gcc/insn-notes.def (revision 195874)
+++ gcc/insn-notes.def (working copy)
@@ -70,7 +70,9 @@ INSN_NOTE (CALL_ARG_LOCATION)
/* Record the struct for the following basic block. Uses
NOTE_BASIC_BLOCK. FIXME: Redundant with the basic block pointer
- now included in every insn. */
+ now included in every insn. NOTE: If there's no CFG anymore, in other
words,
+ if BLOCK_FOR_INSN () == NULL, NOTE_BASIC_BLOCK cannot be considered
reliable
+ anymore. */
INSN_NOTE (BASIC_BLOCK)
/* Mark the inflection point in the instruction stream where we switch
Index: gcc/cfgrtl.c
===================================================================
--- gcc/cfgrtl.c (revision 195874)
+++ gcc/cfgrtl.c (working copy)
@@ -135,7 +135,7 @@ delete_insn (rtx insn)
if (! can_delete_label_p (insn))
{
const char *name = LABEL_NAME (insn);
- basic_block bb, label_bb = BLOCK_FOR_INSN (insn);
+ basic_block bb = BLOCK_FOR_INSN (insn);
rtx bb_note = NEXT_INSN (insn);
really_delete = false;
@@ -144,15 +144,13 @@ delete_insn (rtx insn)
NOTE_DELETED_LABEL_NAME (insn) = name;
/* If the note following the label starts a basic block, and the
- label is a member of the same basic block, interchange the two.
- If the label is not marked with a bb, assume it's the same bb. */
+ label is a member of the same basic block, interchange the two. */
if (bb_note != NULL_RTX
&& NOTE_INSN_BASIC_BLOCK_P (bb_note)
- && (label_bb == NOTE_BASIC_BLOCK (bb_note)
- || label_bb == NULL))
+ && bb != NULL
+ && bb == BLOCK_FOR_INSN (bb_note))
{
reorder_insns_nobb (insn, insn, bb_note);
- bb = NOTE_BASIC_BLOCK (bb_note);
BB_HEAD (bb) = bb_note;
if (BB_END (bb) == bb_note)
BB_END (bb) = insn;
...
More information about the Gcc-bugs
mailing list