[gfortran,patch] Fix wrong-code for ASSOCIATED

[FX Coudert]

> I noticed that in the ELSE portion of outer ithe IF statement, there
> is an "if (ss1 == gfc_ss_terminator)" block (see line 2858).  Do
> we need to insert the same two lines above the "tmp = " (line 2866)?

Right you are. Applying the same fix (see patch at the end of the mail) 
allows the following testcase to work:

   integer :: i
   integer,target :: u
   logical :: l
   type dt
     integer, pointer :: a => NULL()
   end type dt
   type(dt) :: obj(1)

   i = 2
   l = associated(obj(i)%a)
   print *, l
   l = associated(obj(i)%a,u)
   print *, l
   end

I intend to commit the whole patch in 24 hours, unless someone objects.

FX



PS: The complete patch is

Index: trans-intrinsic.c
===================================================================
--- trans-intrinsic.c   (revision 114721)
+++ trans-intrinsic.c   (working copy)
@@ -2840,6 +2840,8 @@
            gfc_conv_expr_lhs (&arg1se, arg1->expr);
            tmp2 = gfc_conv_descriptor_data_get (arg1se.expr);
          }
+      gfc_add_block_to_block (&se->pre, &arg1se.pre);
+      gfc_add_block_to_block (&se->post, &arg1se.post);
        tmp = build2 (NE_EXPR, boolean_type_node, tmp2,
                     fold_convert (TREE_TYPE (tmp2), null_pointer_node));
        se->expr = tmp;
@@ -2863,6 +2865,8 @@
            gfc_conv_expr (&arg1se, arg1->expr);
            arg2se.want_pointer = 1;
            gfc_conv_expr (&arg2se, arg2->expr);
+         gfc_add_block_to_block (&se->pre, &arg1se.pre);
+         gfc_add_block_to_block (&se->post, &arg1se.post);
            tmp = build2 (EQ_EXPR, boolean_type_node, arg1se.expr, 
arg2se.expr);
            se->expr = tmp;
          }