Bounds-checking for Fortran testsuite was: Re: [gfortran, testsuite] Fix bounds checking for substrings, enable bounds chekcing in dg testsuite, assorted fixes

[Andreas Tobler]

Janis Johnson wrote:
> On Wed, Dec 13, 2006 at 06:45:19PM +0100, Tobias Schlüter wrote:
>> Andrew Pinski <pinskia@gmail.com> wrote on Wed, 13 Dec 2006:
>>>> My Tcl-fu is nil, but does this patch fix it?
>>> Most likely but this patch is not enough to keep some failures away if
>>> the target GCC supports altivec but the machine does not.  Likewise for
>>> SSE2.  FX's testresults show those failures.  Maybe adding the testing
>>> of a trivial.f90 which contains a couple of vectorizable loops inside
>>> fortran-torture.exp will fix that issue.
>> Hm, so why does this not happen with the vect testsuite?  I was
>> precisely following its logic, which says "on i686, always _run_
>> tests".  I could add something like check_vmx_hw_available, of course,
>> I'm just wondering.
> 
> I think the initial error is a Tcl problem. I'm trying to improve my
> Tcl-fu so I'll take a look at it.
> 
> There are two effective targets for AltiVec tests; powerpc_altivec_ok
> says the test compiler can compile AltiVec code, and vmx_hw says that
> the test system can run the test.  I've got PowerPC systems that do
> and do not support VMX so I'll try these tests on both.

Ok, here is what helped me on darwin-ppc and linux-ppc, both 32-bit targets:

Index: gcc/testsuite/lib/gfortran.exp
===================================================================
--- gcc/testsuite/lib/gfortran.exp      (revision 119856)
+++ gcc/testsuite/lib/gfortran.exp      (working copy)
@@ -25,6 +25,7 @@
  load_lib prune.exp
  load_lib gcc-defs.exp
  load_lib target-libpath.exp
+load_lib target-supports.exp

  #
  # GFORTRAN_UNDER_TEST is the compiler under test.
Index: gcc/testsuite/lib/gfortran-dg.exp
===================================================================
--- gcc/testsuite/lib/gfortran-dg.exp   (revision 119856)
+++ gcc/testsuite/lib/gfortran-dg.exp   (working copy)
@@ -14,8 +14,8 @@
  # along with this program; if not, write to the Free Software
  # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 
02110-1301, USA.

+load_lib gcc-dg.exp
  load_lib fortran-torture.exp
-load_lib gcc-dg.exp

  # Define gfortran callbacks for dg.exp.


i?86-apple-darwin* is also happy with.

I try to find out why libgomp fails.

Andreas