fix get_expr_operands thinko
Richard Henderson
rth@redhat.com
Sat Jul 17 11:54:00 GMT 2004
Blah. I'd seen simd-[23].c fail yesterday, but because of sloppiness
on my part I assumed it was DannyB's, just because the same update/test
cycle did in fact bring in a (different) new failure from him.
Oh well, fixed now.
r~
* tree-ssa-operands.c (get_expr_operands): Fix 2004-07-15
switchification wrt CONSTRUCTOR. Document the reason.
Index: tree-ssa-operands.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/tree-ssa-operands.c,v
retrieving revision 2.19
diff -c -p -d -r2.19 tree-ssa-operands.c
*** tree-ssa-operands.c 16 Jul 2004 21:13:03 -0000 2.19
--- tree-ssa-operands.c 16 Jul 2004 22:11:35 -0000
*************** get_expr_operands (tree stmt, tree *expr
*** 971,976 ****
--- 971,988 ----
stmt_ann (stmt)->has_volatile_ops = true;
return;
+ case CONSTRUCTOR:
+ {
+ /* General aggregate CONSTRUCTORs have been decomposed, but they
+ are still in use as the COMPLEX_EXPR equivalent for vectors. */
+
+ tree t;
+ for (t = TREE_OPERAND (expr, 0); t ; t = TREE_CHAIN (t))
+ get_expr_operands (stmt, &TREE_VALUE (t), opf_none, prev_vops);
+
+ return;
+ }
+
case TRUTH_NOT_EXPR:
case BIT_FIELD_REF:
do_unary:
*************** get_expr_operands (tree stmt, tree *expr
*** 1025,1031 ****
case EXC_PTR_EXPR:
case FILTER_EXPR:
case LABEL_DECL:
- case CONSTRUCTOR:
/* Expressions that make no memory references. */
return;
--- 1037,1042 ----
More information about the Gcc-patches
mailing list