public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [C PATCH] Fix up diagnostics (PR c/25801)
@ 2014-04-18 20:30 Marek Polacek
  2014-05-01 22:54 ` Joseph S. Myers
  0 siblings, 1 reply; 4+ messages in thread
From: Marek Polacek @ 2014-04-18 20:30 UTC (permalink / raw)
  To: GCC Patches; +Cc: Joseph S. Myers

Here's a patch that tidies up diagnostics given when there's
an arithmetic on pointer to an incomplete type.  Firstly, we ought not
to say "unknown structure" when the type is not a structure; secondly,
we shouldn't report identical error twice.
(It might make sense to instead "increment/decrement of" just say
"arithmetic on", but I'm not changing that now.)

Regtested on x86_64-unknown-linux-gnu, bootstrapped on
powerpc64-unknown-linux-gnu, ok for trunk?

2014-04-18  Marek Polacek  <polacek@redhat.com>

	PR c/25801
	* c-typeck.c (c_size_in_bytes): Don't call error.
	(pointer_diff): Remove comment.
	(build_unary_op): Improve error messages.

	* gcc.dg/pr25801.c: New test.

diff --git gcc/c/c-typeck.c gcc/c/c-typeck.c
index 65aad45..bc977a7 100644
--- gcc/c/c-typeck.c
+++ gcc/c/c-typeck.c
@@ -1760,15 +1760,10 @@ c_size_in_bytes (const_tree type)
 {
   enum tree_code code = TREE_CODE (type);
 
-  if (code == FUNCTION_TYPE || code == VOID_TYPE || code == ERROR_MARK)
+  if (code == FUNCTION_TYPE || code == VOID_TYPE || code == ERROR_MARK
+      || !COMPLETE_OR_VOID_TYPE_P (type))
     return size_one_node;
 
-  if (!COMPLETE_OR_VOID_TYPE_P (type))
-    {
-      error ("arithmetic on pointer to an incomplete type");
-      return size_one_node;
-    }
-
   /* Convert in case a char is more than one unit.  */
   return size_binop_loc (input_location, CEIL_DIV_EXPR, TYPE_SIZE_UNIT (type),
 			 size_int (TYPE_PRECISION (char_type_node)
@@ -3528,7 +3523,6 @@ pointer_diff (location_t loc, tree op0, tree op1)
   if (!COMPLETE_OR_VOID_TYPE_P (TREE_TYPE (TREE_TYPE (orig_op1))))
     error_at (loc, "arithmetic on pointer to an incomplete type");
 
-  /* This generates an error if op0 is pointer to incomplete type.  */
   op1 = c_size_in_bytes (target_type);
 
   if (pointer_to_zero_sized_aggr_p (TREE_TYPE (orig_op1)))
@@ -4002,16 +3996,18 @@ build_unary_op (location_t location,
 
 	if (typecode == POINTER_TYPE)
 	  {
-	    /* If pointer target is an undefined struct,
+	    /* If pointer target is an incomplete type,
 	       we just cannot know how to do the arithmetic.  */
 	    if (!COMPLETE_OR_VOID_TYPE_P (TREE_TYPE (argtype)))
 	      {
 		if (code == PREINCREMENT_EXPR || code == POSTINCREMENT_EXPR)
 		  error_at (location,
-			    "increment of pointer to unknown structure");
+			    "increment of pointer to an incomplete type %qT",
+			    TREE_TYPE (argtype));
 		else
 		  error_at (location,
-			    "decrement of pointer to unknown structure");
+			    "decrement of pointer to an incomplete type %qT",
+			    TREE_TYPE (argtype));
 	      }
 	    else if (TREE_CODE (TREE_TYPE (argtype)) == FUNCTION_TYPE
 		     || TREE_CODE (TREE_TYPE (argtype)) == VOID_TYPE)
diff --git gcc/testsuite/gcc.dg/pr25801.c gcc/testsuite/gcc.dg/pr25801.c
index e69de29..70b4ef8 100644
--- gcc/testsuite/gcc.dg/pr25801.c
+++ gcc/testsuite/gcc.dg/pr25801.c
@@ -0,0 +1,19 @@
+/* PR c/25801 */
+/* { dg-do compile } */
+
+int (*a)[];
+struct S *s;
+
+void
+f (void)
+{
+  a++; /* { dg-error "increment of pointer to an incomplete type" } */
+  ++a; /* { dg-error "increment of pointer to an incomplete type" } */
+  a--; /* { dg-error "decrement of pointer to an incomplete type" } */
+  --a; /* { dg-error "decrement of pointer to an incomplete type" } */
+
+  s++; /* { dg-error "increment of pointer to an incomplete type" } */
+  ++s; /* { dg-error "increment of pointer to an incomplete type" } */
+  s--; /* { dg-error "decrement of pointer to an incomplete type" } */
+  --s; /* { dg-error "decrement of pointer to an incomplete type" } */
+}

	Marek

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [C PATCH] Fix up diagnostics (PR c/25801)
  2014-04-18 20:30 [C PATCH] Fix up diagnostics (PR c/25801) Marek Polacek
@ 2014-05-01 22:54 ` Joseph S. Myers
  2014-05-02  7:44   ` Marek Polacek
  0 siblings, 1 reply; 4+ messages in thread
From: Joseph S. Myers @ 2014-05-01 22:54 UTC (permalink / raw)
  To: Marek Polacek; +Cc: GCC Patches

On Fri, 18 Apr 2014, Marek Polacek wrote:

> @@ -1760,15 +1760,10 @@ c_size_in_bytes (const_tree type)

I think the comment on this function should be updated to explain the 
interface contract when an incomplete (or function) type is passed (i.e. 
return size_one_node, caller is responsible for any diagnostics).

> diff --git gcc/testsuite/gcc.dg/pr25801.c gcc/testsuite/gcc.dg/pr25801.c
> index e69de29..70b4ef8 100644
> --- gcc/testsuite/gcc.dg/pr25801.c
> +++ gcc/testsuite/gcc.dg/pr25801.c
> @@ -0,0 +1,19 @@
> +/* PR c/25801 */
> +/* { dg-do compile } */
> +
> +int (*a)[];
> +struct S *s;
> +
> +void
> +f (void)
> +{
> +  a++; /* { dg-error "increment of pointer to an incomplete type" } */
> +  ++a; /* { dg-error "increment of pointer to an incomplete type" } */
> +  a--; /* { dg-error "decrement of pointer to an incomplete type" } */
> +  --a; /* { dg-error "decrement of pointer to an incomplete type" } */
> +
> +  s++; /* { dg-error "increment of pointer to an incomplete type" } */
> +  ++s; /* { dg-error "increment of pointer to an incomplete type" } */
> +  s--; /* { dg-error "decrement of pointer to an incomplete type" } */
> +  --s; /* { dg-error "decrement of pointer to an incomplete type" } */
> +}

The test should I think also cover incomplete union types, and pointer 
subtraction for all the different cases of incomplete types, and the += 
and -= operators (i.e. cover invalid arithmetic on pointers to incomplete 
types more thoroughly, even if not actually affected by this patch).

-- 
Joseph S. Myers
joseph@codesourcery.com

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [C PATCH] Fix up diagnostics (PR c/25801)
  2014-05-01 22:54 ` Joseph S. Myers
@ 2014-05-02  7:44   ` Marek Polacek
  2014-05-02 17:51     ` Joseph S. Myers
  0 siblings, 1 reply; 4+ messages in thread
From: Marek Polacek @ 2014-05-02  7:44 UTC (permalink / raw)
  To: Joseph S. Myers; +Cc: GCC Patches

On Thu, May 01, 2014 at 10:54:18PM +0000, Joseph S. Myers wrote:
> I think the comment on this function should be updated to explain the 
> interface contract when an incomplete (or function) type is passed (i.e. 
> return size_one_node, caller is responsible for any diagnostics).
 
Done.

> The test should I think also cover incomplete union types, and pointer 
> subtraction for all the different cases of incomplete types, and the += 
> and -= operators (i.e. cover invalid arithmetic on pointers to incomplete 
> types more thoroughly, even if not actually affected by this patch).

Done.

Another change is that I used COMPLETE_TYPE_P instead of
COMPLETE_OR_VOID_TYPE_P, that would be a bug and before it didn't make
much sense: we checked VOID_TYPE before.

Tested x86_64-linux, ok now?

2014-05-02  Marek Polacek  <polacek@redhat.com>

	PR c/25801
	* c-typeck.c (c_size_in_bytes): Update comment.  Don't call error.
	Return size_one_node when the type is not complete.
	(pointer_diff): Remove comment.
	(build_unary_op): Improve error messages.

	* gcc.dg/pr25801.c: New test.

diff --git gcc/c/c-typeck.c gcc/c/c-typeck.c
index 98567b4..41770bc 100644
--- gcc/c/c-typeck.c
+++ gcc/c/c-typeck.c
@@ -1754,22 +1754,20 @@ type_lists_compatible_p (const_tree args1, const_tree args2,
     }
 }
 \f
-/* Compute the size to increment a pointer by.  */
+/* Compute the size to increment a pointer by.  When a function type or void
+   type or incomplete type is passed, size_one_node is returned.
+   This function does not emit any diagnostics; the caller is responsible
+   for that.  */
 
 static tree
 c_size_in_bytes (const_tree type)
 {
   enum tree_code code = TREE_CODE (type);
 
-  if (code == FUNCTION_TYPE || code == VOID_TYPE || code == ERROR_MARK)
+  if (code == FUNCTION_TYPE || code == VOID_TYPE || code == ERROR_MARK
+      || !COMPLETE_TYPE_P (type))
     return size_one_node;
 
-  if (!COMPLETE_OR_VOID_TYPE_P (type))
-    {
-      error ("arithmetic on pointer to an incomplete type");
-      return size_one_node;
-    }
-
   /* Convert in case a char is more than one unit.  */
   return size_binop_loc (input_location, CEIL_DIV_EXPR, TYPE_SIZE_UNIT (type),
 			 size_int (TYPE_PRECISION (char_type_node)
@@ -3530,7 +3528,6 @@ pointer_diff (location_t loc, tree op0, tree op1)
   if (!COMPLETE_OR_VOID_TYPE_P (TREE_TYPE (TREE_TYPE (orig_op1))))
     error_at (loc, "arithmetic on pointer to an incomplete type");
 
-  /* This generates an error if op0 is pointer to incomplete type.  */
   op1 = c_size_in_bytes (target_type);
 
   if (pointer_to_zero_sized_aggr_p (TREE_TYPE (orig_op1)))
@@ -4004,16 +4001,18 @@ build_unary_op (location_t location,
 
 	if (typecode == POINTER_TYPE)
 	  {
-	    /* If pointer target is an undefined struct,
+	    /* If pointer target is an incomplete type,
 	       we just cannot know how to do the arithmetic.  */
 	    if (!COMPLETE_OR_VOID_TYPE_P (TREE_TYPE (argtype)))
 	      {
 		if (code == PREINCREMENT_EXPR || code == POSTINCREMENT_EXPR)
 		  error_at (location,
-			    "increment of pointer to unknown structure");
+			    "increment of pointer to an incomplete type %qT",
+			    TREE_TYPE (argtype));
 		else
 		  error_at (location,
-			    "decrement of pointer to unknown structure");
+			    "decrement of pointer to an incomplete type %qT",
+			    TREE_TYPE (argtype));
 	      }
 	    else if (TREE_CODE (TREE_TYPE (argtype)) == FUNCTION_TYPE
 		     || TREE_CODE (TREE_TYPE (argtype)) == VOID_TYPE)
diff --git gcc/testsuite/gcc.dg/pr25801.c gcc/testsuite/gcc.dg/pr25801.c
index e69de29..10b53d3 100644
--- gcc/testsuite/gcc.dg/pr25801.c
+++ gcc/testsuite/gcc.dg/pr25801.c
@@ -0,0 +1,44 @@
+/* PR c/25801 */
+/* { dg-do compile } */
+/* { dg-options "-std=c99" } */
+
+int (*a)[];
+struct S *s;
+union U *u;
+enum E *e;
+
+void
+f (void)
+{
+  a++; /* { dg-error "increment of pointer to an incomplete type" } */
+  ++a; /* { dg-error "increment of pointer to an incomplete type" } */
+  a--; /* { dg-error "decrement of pointer to an incomplete type" } */
+  --a; /* { dg-error "decrement of pointer to an incomplete type" } */
+  a += 1; /* { dg-error "invalid use of array with unspecified bounds" } */
+  a -= 1; /* { dg-error "invalid use of array with unspecified bounds" } */
+  a - a; /* { dg-error "arithmetic on pointer to an incomplete type" } */
+
+  s++; /* { dg-error "increment of pointer to an incomplete type" } */
+  ++s; /* { dg-error "increment of pointer to an incomplete type" } */
+  s--; /* { dg-error "decrement of pointer to an incomplete type" } */
+  --s; /* { dg-error "decrement of pointer to an incomplete type" } */
+  s += 1; /* { dg-error "invalid use of undefined type" } */
+  s -= 1; /* { dg-error "invalid use of undefined type" } */
+  s - s; /* { dg-error "arithmetic on pointer to an incomplete type" } */
+
+  u++; /* { dg-error "increment of pointer to an incomplete type" } */
+  ++u; /* { dg-error "increment of pointer to an incomplete type" } */
+  u--; /* { dg-error "decrement of pointer to an incomplete type" } */
+  --u; /* { dg-error "decrement of pointer to an incomplete type" } */
+  u += 1; /* { dg-error "invalid use of undefined type" } */
+  u -= 1; /* { dg-error "invalid use of undefined type" } */
+  u - u; /* { dg-error "arithmetic on pointer to an incomplete type" } */
+
+  e++; /* { dg-error "increment of pointer to an incomplete type" } */
+  ++e; /* { dg-error "increment of pointer to an incomplete type" } */
+  e--; /* { dg-error "decrement of pointer to an incomplete type" } */
+  --e; /* { dg-error "decrement of pointer to an incomplete type" } */
+  e += 1; /* { dg-error "invalid use of undefined type" } */
+  e -= 1; /* { dg-error "invalid use of undefined type" } */
+  e - e; /* { dg-error "arithmetic on pointer to an incomplete type" } */
+}

	Marek

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [C PATCH] Fix up diagnostics (PR c/25801)
  2014-05-02  7:44   ` Marek Polacek
@ 2014-05-02 17:51     ` Joseph S. Myers
  0 siblings, 0 replies; 4+ messages in thread
From: Joseph S. Myers @ 2014-05-02 17:51 UTC (permalink / raw)
  To: Marek Polacek; +Cc: GCC Patches

On Fri, 2 May 2014, Marek Polacek wrote:

> On Thu, May 01, 2014 at 10:54:18PM +0000, Joseph S. Myers wrote:
> > I think the comment on this function should be updated to explain the 
> > interface contract when an incomplete (or function) type is passed (i.e. 
> > return size_one_node, caller is responsible for any diagnostics).
>  
> Done.
> 
> > The test should I think also cover incomplete union types, and pointer 
> > subtraction for all the different cases of incomplete types, and the += 
> > and -= operators (i.e. cover invalid arithmetic on pointers to incomplete 
> > types more thoroughly, even if not actually affected by this patch).
> 
> Done.
> 
> Another change is that I used COMPLETE_TYPE_P instead of
> COMPLETE_OR_VOID_TYPE_P, that would be a bug and before it didn't make
> much sense: we checked VOID_TYPE before.
> 
> Tested x86_64-linux, ok now?

OK.

-- 
Joseph S. Myers
joseph@codesourcery.com

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2014-05-02 17:51 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-04-18 20:30 [C PATCH] Fix up diagnostics (PR c/25801) Marek Polacek
2014-05-01 22:54 ` Joseph S. Myers
2014-05-02  7:44   ` Marek Polacek
2014-05-02 17:51     ` Joseph S. Myers

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).