[Patch, fortran] PR98458 - ICE in gfc_conv_array_initializer due to array expressions within implied do-loop

[Patch, fortran] PR98458 - ICE in gfc_conv_array_initializer due to array expressions within implied do-loop

[Paul Richard Thomas]

[-- Attachment #1: Type: text/plain, Size: 816 bytes --]

This is another obvious patch. The first version of the patch just expanded
the constructor. Steve used gfc_reduce_init_expr to fix the immediate
problem within simplify_reshape. Since this has the advantage of doing the
type check as well and so I incorporated it in this inclusive patch.

Regtested on FC33/x86_64 - OK for master (and 9-/10-branches)?

Paul

Fortran:Fix simplification of constructors with implied-do [PR98458]

2021-01-05  Paul Thomas  <pault@gcc.gnu.org>

gcc/fortran
PR fortran/98458
* simplify.c (is_constant_array_expr): If an array constructor
expression has elements other than constants or structures, try
fixing the expression with gfc_reduce_init_expr. Also, if shape
is NULL, obtain the array size and set it.

gcc/testsuite/
PR fortran/98458
* gfortran.dg/implied_do_3.f90 : New test.

[-- Attachment #2: submit.diff --]
[-- Type: text/x-patch, Size: 1609 bytes --]

diff --git a/gcc/fortran/simplify.c b/gcc/fortran/simplify.c
index bfd410cf31d..06bde5209dd 100644
--- a/gcc/fortran/simplify.c
+++ b/gcc/fortran/simplify.c
@@ -220,6 +220,8 @@ static bool
 is_constant_array_expr (gfc_expr *e)
 {
   gfc_constructor *c;
+  bool array_OK = true;
+  mpz_t size;
 
   if (e == NULL)
     return true;
@@ -231,13 +233,43 @@ is_constant_array_expr (gfc_expr *e)
   if (e->expr_type != EXPR_ARRAY || !gfc_is_constant_expr (e))
     return false;
 
+  for (c = gfc_constructor_first (e->value.constructor);
+       c; c = gfc_constructor_next (c))
+    if (c->expr->expr_type != EXPR_CONSTANT
+	  && c->expr->expr_type != EXPR_STRUCTURE)
+      {
+	array_OK = false;
+	break;
+      }
+
+  /* Check and expand the constructor.  */
+  if (!array_OK && gfc_init_expr_flag && e->rank == 1)
+    {
+      array_OK = gfc_reduce_init_expr (e);
+      /* gfc_reduce_init_expr resets the flag.  */
+      gfc_init_expr_flag = true;
+    }
+  else
+    return array_OK;
+
+  /* Recheck to make sure that any EXPR_ARRAYs have gone.  */
   for (c = gfc_constructor_first (e->value.constructor);
        c; c = gfc_constructor_next (c))
     if (c->expr->expr_type != EXPR_CONSTANT
 	  && c->expr->expr_type != EXPR_STRUCTURE)
       return false;
 
-  return true;
+  /* Make sure that the array has a valid shape.  */
+  if (e->shape == NULL && e->rank == 1)
+    {
+      if (!gfc_array_size(e, &size))
+	return false;
+      e->shape = gfc_get_shape (1);
+      mpz_init_set (e->shape[0], size);
+      mpz_clear (size);
+    }
+
+  return array_OK;
 }
 
 /* Test for a size zero array.  */

[-- Attachment #3: implied_do_3.f90 --]
[-- Type: text/x-fortran, Size: 429 bytes --]

! { dg-do compile }
!
! Test the fix for PR98458 in which array expressions within the implied-do
! array constructor caused an ICE in trans-array.c(gfc_conv_array_initializer).
!
! Contributed by Xiao Liu  <xiao.liu@compiler-dev.com>
!
program test
  implicit none
  integer :: i
  integer, parameter :: t(6) = [1,2,3,4,5,6]
  integer, parameter :: tmp(3,2) = reshape([(t(i:i+1),i=1,3)],[3,2])
  print *, tmp  ! Used to ICE
end

Re: [Patch, fortran] PR98458 - ICE in gfc_conv_array_initializer due to array expressions within implied do-loop

[Thomas Koenig]

Hi Paul,

> Regtested on FC33/x86_64 - OK for master (and 9-/10-branches)?

OK.

Thanks for the patch!

Best regards

	Thomas