[C PATCH] Fix error-recovery with variable-sized compound literals (PR c/68090)

Marek Polacek polacek@redhat.com
Thu Nov 5 17:30:00 GMT 2015


This fixes an error-recovery problem, where we were trying to build
a C_MAYBE_CONST_EXPR even though the type is invalid.  I think let's
just skip this for error_mark_nodes.

Bootstrapped/regtested on x86_64-linux, ok for trunk?

2015-11-05  Marek Polacek  <polacek@redhat.com>

	PR c/68090
	* c-parser.c (c_parser_postfix_expression_after_paren_type): Don't
	deal with pre-evaluation on invalid types.

	* gcc.dg/pr68090.c: New test.

diff --git gcc/c/c-parser.c gcc/c/c-parser.c
index ab324d3..e2f55b3 100644
--- gcc/c/c-parser.c
+++ gcc/c/c-parser.c
@@ -7855,7 +7855,7 @@ c_parser_postfix_expression_after_paren_type (c_parser *parser,
   expr.value = build_compound_literal (start_loc, type, init.value, non_const);
   expr.original_code = ERROR_MARK;
   expr.original_type = NULL;
-  if (type_expr)
+  if (type != error_mark_node && type_expr)
     {
       if (TREE_CODE (expr.value) == C_MAYBE_CONST_EXPR)
 	{
diff --git gcc/testsuite/gcc.dg/pr68090.c gcc/testsuite/gcc.dg/pr68090.c
index e69de29..87b3b93 100644
--- gcc/testsuite/gcc.dg/pr68090.c
+++ gcc/testsuite/gcc.dg/pr68090.c
@@ -0,0 +1,13 @@
+/* PR c/68090 */
+/* { dg-do compile } */
+/* { dg-options "" } */
+
+void
+fn (int i)
+{
+  (int[(0, 1)]) { 0 }; /* { dg-error "compound literal has variable size" } */
+  (int[i]) { 0 }; /* { dg-error "compound literal has variable size" } */
+  (int[(0, i)]) { 0 }; /* { dg-error "compound literal has variable size" } */
+  (int [][i]){ 0 }; /* { dg-error "compound literal has variable size" } */
+  (int [][(1, 2)]){ 0 }; /* { dg-error "compound literal has variable size" } */
+}

	Marek



More information about the Gcc-patches mailing list