[PATCH][C fam] Fix PR80593, bogus strict-aliasing warnings

[PATCH][C fam] Fix PR80593, bogus strict-aliasing warnings

[Richard Biener]

The following reverts an earlier change of mine (2008) to explicitely
warn about accessing alias-set zero memory with alias-set non-zero.
That was supposed to catch the case in g++.dg/warn/Wstrict-aliasing-6.C
which is

int foo ()
{
  char buf[8];
  return *((int *)buf); /* { dg-warning "strict-aliasing" } */
}

but at least since the typeless storage work this is considered valid
and thus this warning is really bogus.

Bootstrapped on x86_64-unknown-linux-gnu, testing in progress.

Ok for trunk and GCC 7 branch?

Thanks,
Richard.

2017-05-18  Richard Biener  <rguenther@suse.de>

	PR c++/80593
	* c-warn.c (strict_aliasing_warning): Do not warn for accesses
	to alias-set zero memory.

	* g++.dg/warn/Wstrict-aliasing-bogus-char-2.C: New testcase.
	* g++.dg/warn/Wstrict-aliasing-6.C: Adjust expected outcome.

Index: gcc/c-family/c-warn.c
===================================================================
--- gcc/c-family/c-warn.c	(revision 248179)
+++ gcc/c-family/c-warn.c	(working copy)
@@ -537,10 +537,10 @@ strict_aliasing_warning (tree otype, tre
 	    = get_alias_set (TREE_TYPE (TREE_OPERAND (expr, 0)));
 	  alias_set_type set2 = get_alias_set (TREE_TYPE (type));
 
-	  if (set1 != set2 && set2 != 0
-	      && (set1 == 0
-		  || (!alias_set_subset_of (set2, set1)
-		      && !alias_sets_conflict_p (set1, set2))))
+	  if (set2 != 0
+	      && set1 != set2
+	      && !alias_set_subset_of (set2, set1)
+	      && !alias_sets_conflict_p (set1, set2))
 	    {
 	      warning (OPT_Wstrict_aliasing, "dereferencing type-punned "
 		       "pointer will break strict-aliasing rules");
Index: gcc/testsuite/g++.dg/warn/Wstrict-aliasing-bogus-char-2.C
===================================================================
--- gcc/testsuite/g++.dg/warn/Wstrict-aliasing-bogus-char-2.C	(nonexistent)
+++ gcc/testsuite/g++.dg/warn/Wstrict-aliasing-bogus-char-2.C	(working copy)
@@ -0,0 +1,19 @@
+// { dg-do compile }
+// { dg-options "-O2 -Wstrict-aliasing" }
+
+template<unsigned _Len, unsigned _Align>
+struct aligned_storage
+{
+  union type
+    {
+      unsigned char __data[_Len];
+      struct __attribute__((__aligned__((_Align)))) { } __align;
+    };
+};
+
+aligned_storage<sizeof(int), __alignof__(int)>::type storage;
+
+int main()
+{
+  *reinterpret_cast<int*>(&storage) = 42; // { dg-bogus "break strict-aliasing" }
+}
Index: gcc/testsuite/g++.dg/warn/Wstrict-aliasing-6.C
===================================================================
--- gcc/testsuite/g++.dg/warn/Wstrict-aliasing-6.C	(revision 248179)
+++ gcc/testsuite/g++.dg/warn/Wstrict-aliasing-6.C	(working copy)
@@ -4,6 +4,6 @@
 int foo ()
 {
   char buf[8];
-  return *((int *)buf); /* { dg-warning "strict-aliasing" } */
+  return *((int *)buf); /* { dg-bogus "strict-aliasing" } */
 }
 

Re: [PATCH][C fam] Fix PR80593, bogus strict-aliasing warnings

[Marek Polacek]

On Thu, May 18, 2017 at 01:00:39PM +0200, Richard Biener wrote:
> 
> The following reverts an earlier change of mine (2008) to explicitely
> warn about accessing alias-set zero memory with alias-set non-zero.
> That was supposed to catch the case in g++.dg/warn/Wstrict-aliasing-6.C
> which is
> 
> int foo ()
> {
>   char buf[8];
>   return *((int *)buf); /* { dg-warning "strict-aliasing" } */
> }
> 
> but at least since the typeless storage work this is considered valid
> and thus this warning is really bogus.
> 
> Bootstrapped on x86_64-unknown-linux-gnu, testing in progress.
> 
> Ok for trunk and GCC 7 branch?

I think this is fine.

	Marek

Re: [PATCH][C fam] Fix PR80593, bogus strict-aliasing warnings

[Richard Biener]

On Fri, 19 May 2017, Marek Polacek wrote:

> On Thu, May 18, 2017 at 01:00:39PM +0200, Richard Biener wrote:
> > 
> > The following reverts an earlier change of mine (2008) to explicitely
> > warn about accessing alias-set zero memory with alias-set non-zero.
> > That was supposed to catch the case in g++.dg/warn/Wstrict-aliasing-6.C
> > which is
> > 
> > int foo ()
> > {
> >   char buf[8];
> >   return *((int *)buf); /* { dg-warning "strict-aliasing" } */
> > }
> > 
> > but at least since the typeless storage work this is considered valid
> > and thus this warning is really bogus.
> > 
> > Bootstrapped on x86_64-unknown-linux-gnu, testing in progress.
> > 
> > Ok for trunk and GCC 7 branch?
> 
> I think this is fine.

Thanks.

Committed to trunk sofar.

Richard.