[patch, fortran] PR41807 [4.5/4.4 Regression] data statement with nested type constructors

[Jerry DeLisle <jvdelisle at verizon dot net>]

Hi,

This patch fixes this bug by removing an error check in resolve.c that uses 
gfc_is_constant_expr.  This function performs an expansion of the constructors, 
evidently too soon, causing the regression by modifying the expressions.

Obviously, these constructors do get properly expanded later.  Perhaps it should 
be renamed a bit to reflect this, maybe gfc_expand_is_constant_expr.

Regardless, after removing this function from next_data_value and running the 
testsuite I observed that the only failure was in data_value_1.f90.  The failure 
is an assert in trans-const.c (gfc_convert_const).  This seemed like the natural 
place to verify that each array element is actually a constant.

Thus I replaced the assert with the check for BT_CONSTANT, rewording the error 
message to be more general. The test case data_value_1.f90 is revised to reflect 
the new error message. A new test case, originally from Steve Kargl is also 
attached.

Regression tested on x86-64-gnu-linux.

OK for trunk?

Regards,

Jerry

2009-11-21  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

	* trans-const.c (gfc_conv_const): Fix typo in comment. Replace assert
	with error message if not constant.
	* resolve.c (next_data_value): Delete check for constant.

Index: testsuite/gfortran.dg/data_value_1.f90
===================================================================
--- testsuite/gfortran.dg/data_value_1.f90	(revision 154170)
+++ testsuite/gfortran.dg/data_value_1.f90	(working copy)
@@ -3,12 +3,14 @@
 ! is not a constant and so the DATA statement did not have
 ! a constant value expression.
 !
+! Modified dg-error for PR41807
+!
 ! Contributed by Philippe Marguinaud <philippe.marguinaud@meteo.fr>
 !
       TYPE POINT
         REAL :: X 
       ENDTYPE
       TYPE(POINT) :: P
-      DATA P / POINT(1.+X) / ! { dg-error "non-constant DATA value" }
+      DATA P / POINT(1.+X) / ! { dg-error "non-constant initialization" }
       print *, p
       END
Index: fortran/trans-const.c
===================================================================
--- fortran/trans-const.c	(revision 154411)
+++ fortran/trans-const.c	(working copy)
@@ -340,7 +340,7 @@ void
 gfc_conv_constant (gfc_se * se, gfc_expr * expr)
 {
   /* We may be receiving an expression for C_NULL_PTR or C_NULL_FUNPTR.  If
-     so, they expr_type will not yet be an EXPR_CONSTANT.  We need to make
+     so, the expr_type will not yet be an EXPR_CONSTANT.  We need to make
      it so here.  */
   if (expr->ts.type == BT_DERIVED && expr->ts.u.derived
       && expr->ts.u.derived->attr.is_iso_c)
@@ -353,7 +353,11 @@ gfc_conv_constant (gfc_se * se, gfc_expr * expr)
         }
     }
 
-  gcc_assert (expr->expr_type == EXPR_CONSTANT);
+  if (expr->expr_type != EXPR_CONSTANT)
+    {
+      gfc_error ("non-constant initialization expression at %L", &expr->where);
+      return;
+    }
 
   if (se->ss != NULL)
     {
Index: fortran/resolve.c
===================================================================
--- fortran/resolve.c	(revision 154411)
+++ fortran/resolve.c	(working copy)
@@ -11083,9 +11083,6 @@ next_data_value (void)
 {
   while (mpz_cmp_ui (values.left, 0) == 0)
     {
-      if (!gfc_is_constant_expr (values.vnode->expr))
-	gfc_error ("non-constant DATA value at %L",
-		   &values.vnode->expr->where);
 
       if (values.vnode->next == NULL)
 	return FAILURE;

! { dg-do run }
! PR41807  data statement with nested type constructors
! Test case provided by Steve Kargl
  implicit none

  type :: a
     real :: x(3)
  end type a

  integer, parameter :: n = 3

  type(a) :: b(n)

  real, parameter :: d1(3) = (/1., 2., 3./)
  real, parameter :: d2(3) = (/4., 5., 6./)
  real, parameter :: d3(3) = (/7., 8., 9./)

  integer :: i, z(n)
 
  data (b(i), i = 1, n) /a(d1), a(d2), a(d3)/
  data (z(i), i = 1, n) / 1, 2, 3/

  if (any(z.ne.[1, 2, 3])) call abort
  if (any(b(1)%x.ne.[1, 2, 3]) .or. &
      any(b(2)%x.ne.[4, 5, 6]) .or. &
      any(b(3)%x.ne.[7, 8, 9])) call abort
end