Fix PR fortran/44693, take 2

Fix PR fortran/44693, take 2

[Thomas Koenig]

[-- Attachment #1: Type: text/plain, Size: 793 bytes --]

Hello world,

here is a correct version of a fix for PR 44693.  This contains the
necessary special case for SPREAD (where dim can be one larger than the
rank of the array), an expanded test case and a correction to an old
test case.

I think this is finally correct.

Regression-tested, no new regressions.

OK for trunk?

	Thomas

2010-07-06  Thomas Koenig  <tkoenig@gcc.gnu.org>

	PR fortran/PR44693
	* check.c (dim_rank_check):  Also check intrinsic functions.
	Adjust permissible rank for functions which reduce the rank of
	their argument.  Spread is an exception, where DIM can
	be one larger than the rank of array.

2010-07-06  Thomas Koenig  <tkoenig@gcc.gnu.org>

	PR fortran/PR44693
	* gfortran.dg/dim_range_1.f90:  New test.
	* gfortran.dg/minmaxloc_4.f90:  Remove invalid test.


[-- Attachment #2: pr44693-3.diff --]
[-- Type: text/x-patch, Size: 1379 bytes --]

Index: testsuite/gfortran.dg/minmaxloc_4.f90
===================================================================
--- testsuite/gfortran.dg/minmaxloc_4.f90	(Revision 161784)
+++ testsuite/gfortran.dg/minmaxloc_4.f90	(Arbeitskopie)
@@ -3,7 +3,6 @@
 PROGRAM TST
   IMPLICIT NONE
   REAL :: A(1,3)
-  REAL :: B(3,1)
   A(:,1) = 10
   A(:,2) = 20
   A(:,3) = 30
@@ -13,9 +12,4 @@
   if (minloc(sum(a(:,1:3),1),1) .ne. 1) call abort()
   if (maxloc(sum(a(:,1:3),1),1) .ne. 3) call abort()
 
-  B(1,:) = 10
-  B(2,:) = 20
-  B(3,:) = 30
-  if (minloc(sum(b(1:3,:),2),2) .ne. 1) call abort()
-  if (maxloc(sum(b(1:3,:),2),2) .ne. 3) call abort()
 END PROGRAM TST
Index: fortran/check.c
===================================================================
--- fortran/check.c	(Revision 161784)
+++ fortran/check.c	(Arbeitskopie)
@@ -473,12 +473,15 @@
   if (dim == NULL)
     return SUCCESS;
 
-  if (dim->expr_type != EXPR_CONSTANT
-      || (array->expr_type != EXPR_VARIABLE
-	  && array->expr_type != EXPR_ARRAY))
+  if (dim->expr_type != EXPR_CONSTANT)
     return SUCCESS;
 
-  rank = array->rank;
+  if (array->expr_type == EXPR_FUNCTION && array->value.function.isym
+      && array->value.function.isym->id == GFC_ISYM_SPREAD)
+    rank = array->rank + 1;
+  else
+    rank = array->rank;
+
   if (array->expr_type == EXPR_VARIABLE)
     {
       ar = gfc_find_array_ref (array);

[-- Attachment #3: dim_range_1.f90 --]
[-- Type: text/x-fortran, Size: 781 bytes --]

! { dg-do compile }
! PR 44693 - check for invalid dim even in functions.
! Based on a test case by Dominique d'Humieres.
subroutine test1(esss,Ix,Iyz, n)
  real(kind=kind(1.0d0)), dimension(n), intent(out) :: esss
  real(kind=kind(1.0d0)), dimension(n,n,n) :: sp
  real(kind=kind(1.0d0)), dimension(n,n) :: Ix,Iyz
  esss = sum(Ix * Iyz, 0) ! { dg-error "is not a valid dimension index" }
  esss = sum(Ix * Iyz, 1)
  esss = sum(Ix * Iyz, 2)
  esss = sum(Ix * Iyz, 3) ! { dg-error "is not a valid dimension index" }
  sp = spread (ix * iyz, 0, n) ! { dg-error "is not a valid dimension index" }
  sp = spread (ix * iyz, 1, n)
  sp = spread (ix * iyz, 2, n)
  sp = spread (ix * iyz, 3, n)
  sp = spread (ix * iyz, 4, n) ! { dg-error "is not a valid dimension index" }
end subroutine

Re: Fix PR fortran/44693, take 2

[Tobias Burnus]

Thomas Koenig wrote:
> here is a correct version of a fix for PR 44693.  This contains the
> necessary special case for SPREAD (where dim can be one larger than the
> rank of the array), an expanded test case and a correction to an old
> test case.
>
> I think this is finally correct.
> Regression-tested, no new regressions.
> OK for trunk?
>   

OK. Thanks for the patch - and great that you have spotted (possibly
with the test suite) that SPREAD as only intrinsic can have "1 <= dim <=
n +1".

Tobias

> 2010-07-06  Thomas Koenig  <tkoenig@gcc.gnu.org>
>
> 	PR fortran/PR44693
> 	* check.c (dim_rank_check):  Also check intrinsic functions.
> 	Adjust permissible rank for functions which reduce the rank of
> 	their argument.  Spread is an exception, where DIM can
> 	be one larger than the rank of array.
>
> 2010-07-06  Thomas Koenig  <tkoenig@gcc.gnu.org>
>
> 	PR fortran/PR44693
> 	* gfortran.dg/dim_range_1.f90:  New test.
> 	* gfortran.dg/minmaxloc_4.f90:  Remove invalid test.
>
>   

Re: Fix PR fortran/44693, take 2

[Thomas Koenig]

Hi Tobias,

> > Regression-tested, no new regressions.
> > OK for trunk?
> >   
> 
> OK. Thanks for the patch - and great that you have spotted (possibly
> with the test suite) that SPREAD as only intrinsic can have "1 <= dim
> <=
> n +1".


Thanks for the quick review.  Committed, as rev. 161884.

As a matter of fact, I suddenly remembered not understanding this point
once in the library :-)

	Thomas