[PATCH,testsuite] fix PR 19232

[PATCH,testsuite] fix PR 19232

[Nathan Froyd]

[-- Attachment #1: Type: text/plain, Size: 700 bytes --]

The attached patch fixes PR 19232, which is a problem with -fPIC and
inlining.  The compiler will not inline single-use functions that are
not declared inline with -fPIC, as those functions could be overridden
in other translation units.

The simplest fix seems to be to declare the relevant functions as
inline, since that is what the test intended.  The compiler is then
perfectly happy to inline the functions, which triggers the intended
warnings.

Tested on powerpc-wrs-vxworks with and without -fPIC with no
regressions.  OK to commit?

-Nathan

gcc/testsuite/
2007-07-25  Nathan Froyd  <froydnj@codesourcery.com>

	PR/19232
	* gcc.dg/assign-warn-3.c (f0): Declare as inline.
	(f1): Likewise.


[-- Attachment #2: assign-warn.patch --]
[-- Type: text/plain, Size: 829 bytes --]

Index: gcc/testsuite/gcc.dg/assign-warn-3.c
===================================================================
--- gcc/testsuite/gcc.dg/assign-warn-3.c	(revision 177249)
+++ gcc/testsuite/gcc.dg/assign-warn-3.c	(working copy)
@@ -5,9 +5,9 @@
 /* { dg-options "-O3 -std=c99 -pedantic-errors" } */
 
 /* This is valid to execute, so maybe shouldn't warn at all.  */
-void f0(x) signed char *x; { }
+inline void f0(x) signed char *x; { }
 void g0(unsigned char *x) { f0(x); } /* { dg-warning "warning: pointer targets in passing argument 1 of 'f0' differ in signedness" } */
 
 /* This is undefined on execution but still must compile.  */
-void f1(x) int *x; { }
+inline void f1(x) int *x; { }
 void g1(unsigned int *x) { f1(x); } /* { dg-warning "warning: pointer targets in passing argument 1 of 'f1' differ in signedness" } */

Re: [PATCH,testsuite] fix PR 19232

[Janis Johnson]

On Wed, 2007-07-25 at 13:52 -0700, Nathan Froyd wrote:
> The attached patch fixes PR 19232, which is a problem with -fPIC and
> inlining.  The compiler will not inline single-use functions that are
> not declared inline with -fPIC, as those functions could be overridden
> in other translation units.
> 
> The simplest fix seems to be to declare the relevant functions as
> inline, since that is what the test intended.  The compiler is then
> perfectly happy to inline the functions, which triggers the intended
> warnings.
> 
> Tested on powerpc-wrs-vxworks with and without -fPIC with no
> regressions.  OK to commit?

OK.

Janis