Hi!
On the following testcase we ICE during error recovery.
Setting DECL_INITIAL of a FUNCTION_DECL to error_mark_node can't
be possibly helpful, the error must have been diagnosed already;
DECL_INITIAL on the FUNCTION_DECL should be kept as BLOCK.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2015-01-26 Jakub Jelinek <ja...@redhat.com>
PR c/64766
* c-typeck.c (store_init_value): Don't overwrite DECL_INITIAL
of FUNCTION_DECLs with error_mark_node.
* gcc.dg/pr64766.c: New test.
--- gcc/c/c-typeck.c.jj 2015-01-26 12:09:57.000000000 +0100
+++ gcc/c/c-typeck.c 2015-01-26 14:13:08.142325547 +0100
@@ -6422,7 +6422,8 @@ store_init_value (location_t init_loc, t
warning (OPT_Wtraditional, "traditional C rejects automatic "
"aggregate initialization");
- DECL_INITIAL (decl) = value;
+ if (value != error_mark_node || TREE_CODE (decl) != FUNCTION_DECL)
+ DECL_INITIAL (decl) = value;
/* ANSI wants warnings about out-of-range constant initializers. */
STRIP_TYPE_NOPS (value);
--- gcc/testsuite/gcc.dg/pr64766.c.jj 2015-01-26 14:20:25.023822836 +0100
+++ gcc/testsuite/gcc.dg/pr64766.c 2015-01-26 14:20:07.000000000 +0100
@@ -0,0 +1,9 @@
+/* PR c/64766 */
+/* { dg-do compile } */
+
+void
+foo ()
+{
+}
+
+void foo () = 0; /* { dg-error "is initialized like a variable|invalid
initializer" } */
Jakub
