[C PATCH] Fix ice-on-invalid with typedef declarations in for loops (PR c/68320)

[C PATCH] Fix ice-on-invalid with typedef declarations in for loops (PR c/68320)

[Marek Polacek]

While fixing c/67784 I noticed that the following testcase was breaking
because we ended up with a TYPENAME whose name is NULL.
In c_parser_declspecs:
 2301               /* For a typedef name, record the meaning, not the name.
 2302                  In case of 'foo foo, bar;'.  */
 2303               t.spec = lookup_name (value);
we pass T to declspecs_add_type and that segvs when t.spec is NULL.

I think this is best fixed by treating such tokens as IDs and let subsequent
code error on such an invalid code with proper diagnostics.

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

2015-11-13  Marek Polacek  <polacek@redhat.com>

	PR c/68320
	* c-parser.c (c_parser_for_statement): Treat unknown tokens as IDs.

	* gcc.dg/pr68320.c: New test.

diff --git gcc/c/c-parser.c gcc/c/c-parser.c
index c01d651..82d5ce5 100644
--- gcc/c/c-parser.c
+++ gcc/c/c-parser.c
@@ -5757,12 +5757,12 @@ c_parser_for_statement (c_parser *parser, bool ivdep)
     {
       c_token *token = c_parser_peek_token (parser);
       tree decl = lookup_name (token->value);
-      if (decl == NULL_TREE)
-	;
+      if (decl == NULL_TREE || VAR_P (decl))
+	/* If DECL is null, we don't know what this token might be.  Treat
+	   it as an ID for better diagnostics; we'll error later on.  */
+	token->id_kind = C_ID_ID;
       else if (TREE_CODE (decl) == TYPE_DECL)
 	token->id_kind = C_ID_TYPENAME;
-      else if (VAR_P (decl))
-	token->id_kind = C_ID_ID;
     }
 
   token_indent_info next_tinfo
diff --git gcc/testsuite/gcc.dg/pr68320.c gcc/testsuite/gcc.dg/pr68320.c
index e69de29..7060af8 100644
--- gcc/testsuite/gcc.dg/pr68320.c
+++ gcc/testsuite/gcc.dg/pr68320.c
@@ -0,0 +1,67 @@
+/* PR c/68320 */
+/* { dg-do compile } */
+/* { dg-options "" } */
+
+void
+fn1 (void)
+{
+  for (typedef int T;;) /* { dg-error "declaration of non-variable" } */
+    if (1)
+      ;
+  T x; /* { dg-error "unknown type name" } */
+}
+
+void
+fn2 (int i)
+{
+  for (typedef int T;;) /* { dg-error "declaration of non-variable" } */
+    if (1)
+      i = 5;
+  T x; /* { dg-error "unknown type name" } */
+}
+
+void
+fn3 (void)
+{
+  for (typedef int T;;) /* { dg-error "declaration of non-variable" } */
+    if (1)
+      {
+      }
+  T *x; /* { dg-error "unknown type name" } */
+}
+
+void
+fn4 (void)
+{
+  for (typedef int T;;) /* { dg-error "declaration of non-variable" } */
+    if (1)
+      ;
+  T, T; /* { dg-error "undeclared" } */
+}
+
+void
+fn5 (void)
+{
+  for (typedef int T;;) /* { dg-error "declaration of non-variable" } */
+    if (1)
+      ;
+  T = 10; /* { dg-error "undeclared" } */
+}
+
+void
+fn6 (void)
+{
+  for (typedef int T;;) /* { dg-error "declaration of non-variable" } */
+    if (1)
+      ;
+  T[0]; /* { dg-error "undeclared" } */
+}
+
+void
+fn7 (void)
+{
+  for (typedef int T;;) /* { dg-error "declaration of non-variable" } */
+    if (1)
+      ;
+  T (); /* { dg-warning "implicit declaration" } */
+}

	Marek

Re: [C PATCH] Fix ice-on-invalid with typedef declarations in for loops (PR c/68320)

[Joseph Myers]

On Fri, 13 Nov 2015, Marek Polacek wrote:

> 2015-11-13  Marek Polacek  <polacek@redhat.com>
> 
> 	PR c/68320
> 	* c-parser.c (c_parser_for_statement): Treat unknown tokens as IDs.
> 
> 	* gcc.dg/pr68320.c: New test.

OK.

-- 
Joseph S. Myers
joseph@codesourcery.com