[PATCH 5/5] "Fix" the rest of the fallouts of new intra-SRA

[Richard Guenther]

On Tue, 28 Apr 2009, Martin Jambor wrote:

> The following patch amends testcases (rather than the compiler) so
> that they not fail with the new intra-SRA.
> 
> The FRE testcases rely on the fact that SRA does not scalarize unions.
> The new one,  however, does.  Therefore I simply  switched SRA off for
> them.  Hopefully that is the correct thing to do.
> 
> The gfortran.dg/pr25923.f90 expects a weird warning that I do not give
> out because the new SRA does not scalarize anything in that particular
> testcase  because the  individual fields  of the  structure  are never
> accessed individually.  An  unpatched compiler with -fno-tree-sra also
> does not give any warning.  I believe it was left there as potentially
> uselful  and not  misleading but  I tent  to believe  it's  not really
> required...

Hm.  The PR was about garbled warning messages and the warning itself
does look correct (and useful).  But it indeed does not make sense
for SRA to scalarize the return value.  Thus I think it is more 
appropriate to XFAIL the warning.
 
> With  the previous and  this patch,  there are  no regressions  in the
> testsuite on  x86_64-linux-gnu.  I  have even bootstrapped  and tested
> Ada.

The FRE testcase changes are ok.  XFAILing the fortran testcase is also
ok if there are no objections from the Fortran maintainers (CCed).

Thanks,
Richard.

> Thanks,
> 
> Martin
> 
> 
> 2009-04-27  Martin Jambor  <mjambor@suse.cz>
> 
> 	    * gfortran.dg/pr25923.f90: Remove warning expectation.
> 	    * gcc.dg/tree-ssa/ssa-fre-7.c: Compile with -fno-tree-sra.
> 	    * gcc.dg/tree-ssa/ssa-fre-8.c: Likewise.
> 	    * gcc.dg/tree-ssa/ssa-fre-9.c: Likewise.
> 
> 
> Index: mine/gcc/testsuite/gfortran.dg/pr25923.f90
> ===================================================================
> --- mine.orig/gcc/testsuite/gfortran.dg/pr25923.f90
> +++ mine/gcc/testsuite/gfortran.dg/pr25923.f90
> @@ -10,7 +10,7 @@ implicit none
>  
>  contains
>  
> -  function baz(arg) result(res) ! { dg-warning "res.yr' may be" }
> +  function baz(arg) result(res)
>      type(bar), intent(in) :: arg
>      type(bar) :: res
>      logical, external:: some_func
> Index: mine/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-7.c
> ===================================================================
> --- mine.orig/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-7.c
> +++ mine/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-7.c
> @@ -1,5 +1,5 @@
>  /* { dg-do compile } */
> -/* { dg-options "-O -fdump-tree-fre-details -fdump-tree-optimized" } */
> +/* { dg-options "-O -fno-tree-sra -fdump-tree-fre-details -fdump-tree-optimized" } */
>  #if (__SIZEOF_INT__ == __SIZEOF_FLOAT__)
>  typedef int intflt;
>  #elif (__SIZEOF_LONG__ == __SIZEOF_FLOAT__)
> Index: mine/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-8.c
> ===================================================================
> --- mine.orig/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-8.c
> +++ mine/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-8.c
> @@ -1,5 +1,5 @@
>  /* { dg-do compile } */
> -/* { dg-options "-O -fdump-tree-fre-details" } */
> +/* { dg-options "-O -fno-tree-sra -fdump-tree-fre-details" } */
>  #if (__SIZEOF_INT__ == __SIZEOF_FLOAT__)
>  typedef int intflt;
>  #elif (__SIZEOF_LONG__ == __SIZEOF_FLOAT__)
> Index: mine/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-9.c
> ===================================================================
> --- mine.orig/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-9.c
> +++ mine/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-9.c
> @@ -1,5 +1,5 @@
>  /* { dg-do compile } */
> -/* { dg-options "-O -fdump-tree-fre-stats" } */
> +/* { dg-options "-O -fno-tree-sra -fdump-tree-fre-stats" } */
>  
>  union loc {
>      unsigned reg;
> 
> 

-- 
Richard Guenther <rguenther@suse.de>
Novell / SUSE Labs
SUSE LINUX Products GmbH - Nuernberg - AG Nuernberg - HRB 16746 - GF: Markus Rex