Re: [Patch]:PR38166

[Mike Stump <mrs at apple dot com>]

On Feb 18, 2009, at 9:28 AM, Jack Howarth wrote:
> The attached patch skips g++.dg/ext/visibility/class1.C
> on darwin as Apple does in their trees. Tested on
> powerpc-apple-darwin9. Okay for gcc trunk?
>                          Jack
>
> 2009-02-18  Jack Howarth <howarth@bromo.med.uc.edu>
>
> 	PR testsuite/38166
> 	* g++.dg/ext/visibility/class1.C: Skip on Darwin.
>
> Index: gcc/testsuite/g++.dg/ext/visibility/class1.C
> ===================================================================
> --- gcc/testsuite/g++.dg/ext/visibility/class1.C	(revision 144262)
> +++ gcc/testsuite/g++.dg/ext/visibility/class1.C	(working copy)
> @@ -2,7 +2,8 @@
> // Init should not be hidden, so calling it should use the PLT.
>
> // { dg-options "-fpic" }
> -// { dg-do compile { target i?86-*-* x86_64-*-* *-*-darwin* } }
> +// { dg-do compile { target i?86-*-* x86_64-*-* } }
> +// { dg-skip-if "" { *-*-darwin* } { "*" } { "" } }
> // { dg-require-visibility "" }
> // { dg-final { scan-assembler "InitEv@PLT| 
> indirect_symbol.*InitEv" } }

Please modify that line back to:

// { dg-final { scan-assembler "InitEv@PLT" } }

since we are removing darwin mods to the file.  The reason why we're  
doing this is because the linker can synthesize stubs and no @PLT or  
other mechanism is needed to make the code work.  If one used the  
address of the function in a significant way, one would notice @PLT or  
other mechanism.

Ok with that mod.  As to why the skip is still necessary, the target  
line would still match i386-apple-darwin, and we still want to skip  
the testcase in that case.  The *-*-*-darwin was to pick up powerpc.