------- Comment #3 from rguenth at gcc dot gnu dot org 2010-04-13 09:02 -------
Micha - compared to 4.4 we arrive with
(gdb) call debug_rtx (target)
(mem/c/i:SI (plus:DI (reg/f:DI 54 virtual-stack-vars)
(const_int -4 [0xfffffffffffffffc])) [0 result+0 S4 A32])
instead of
(gdb) call debug_rtx (target)
(reg:SI 59 [ D.1598 ])
which isn't valid for
(define_expand "isinfxf2"
[(use (match_operand:SI 0 "register_operand" ""))
(use (match_operand:XF 1 "register_operand" ""))]
"TARGET_USE_FANCY_MATH_387
&& TARGET_C99_FUNCTIONS"
but we assert that it is. Do we never expect a MEM for target or why were
we lucky before?
I can obviously "fix" this by
Index: builtins.c
===================================================================
--- builtins.c (revision 158225)
+++ builtins.c (working copy)
@@ -2316,6 +2316,7 @@ expand_builtin_interclass_mathfn (tree e
tree orig_arg = arg;
/* Make a suitable register to place result in. */
if (!target
+ || !REG_P (target)
|| GET_MODE (target) != TYPE_MODE (TREE_TYPE (exp)))
target = gen_reg_rtx (TYPE_MODE (TREE_TYPE (exp)));
--
rguenth at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |matz at gcc dot gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43730
