https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81192
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |tom at codesourcery dot com
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Ugh. Tail-merging does very stupid stuff here, commoning a loop header with an
inner loop latch.
Now it _seems_ to me we run into a pre-existing bug in same_succ_flush_bbs
that 'same' might be NULL. But it seems to be NULL from the start...
The following fixes it but not tail-mergings stupid (dangerous?) transform
that messes up loops. Where do we get a chance to "cancel" merging
candidates? We should cancel any that do not belong to the same loop father.
Index: gcc/tree-ssa-tail-merge.c
===================================================================
--- gcc/tree-ssa-tail-merge.c (revision 249645)
+++ gcc/tree-ssa-tail-merge.c (working copy)
@@ -810,7 +810,9 @@ same_succ_flush_bb (basic_block bb)
{
same_succ *same = BB_SAME_SUCC (bb);
BB_SAME_SUCC (bb) = NULL;
- if (bitmap_single_bit_set_p (same->bbs))
+ if (! same)
+ ;
+ else if (bitmap_single_bit_set_p (same->bbs))
same_succ_htab->remove_elt_with_hash (same, same->hashval);
else
bitmap_clear_bit (same->bbs, bb->index);
Tom?
