diff mbox series

[fortran] PR92976 - [8/9/10 Regression][OOP] ICE in trans_associate_var, at fortran/trans-stmt.c:1963

Message ID CAGkQGiLV88scDhBmYpJ_WS6DNgguGJACD2f1T-fGVYAi0Cs1ZQ@mail.gmail.com
State New
Headers show
Series [fortran] PR92976 - [8/9/10 Regression][OOP] ICE in trans_associate_var, at fortran/trans-stmt.c:1963 | expand

Commit Message

Paul Richard Thomas Feb. 29, 2020, 5:30 p.m. UTC 
I am a tiny bit skeptical that this is a regression but I will check.
However, it has clearly been there from the early days of OOP without
being picked up.

The fix is to ensure that the temporary has the correct type of array spec.

Regtested on x86_64/FC31 - OK for trunk and 8-/9- branches ?

Cheers

Paul

2020-02-29  Paul Thomas  <pault@gcc.gnu.org>

    PR fortran/92976
    * match.c (select_type_set_tmp): If the selector array spec has
    explicit bounds, make the temporary's bounds deferred.

2020-02-29  Paul Thomas  <pault@gcc.gnu.org>

    PR fortran/92976
    * gfortran.dg/select_type_48.f90 : New test.

Comments

Thomas Koenig March 1, 2020, 1:43 p.m. UTC | #1 
Hi Paul,

> Regtested on x86_64/FC31 - OK for trunk and 8-/9- branches ?

OK, and thanks for the patch.

I think it makes sense to get this into gcc 9.3, which would need
a backport before the 5th of March.

Regards

	Thomas
Paul Richard Thomas March 1, 2020, 4:20 p.m. UTC | #2 
Committed to head as r10-6954-g957a1b14e99596610abb0777ca86a1c80dde24e0.

Thanks, Thomas

Paul

On Sun, 1 Mar 2020 at 13:43, Thomas Koenig <tkoenig@netcologne.de> wrote:
>
> Hi Paul,
>
> > Regtested on x86_64/FC31 - OK for trunk and 8-/9- branches ?
>
> OK, and thanks for the patch.
>
> I think it makes sense to get this into gcc 9.3, which would need
> a backport before the 5th of March.
>
> Regards
>
>         Thomas
diff mbox series

Patch

Index: gcc/fortran/match.c
===================================================================
*** gcc/fortran/match.c	(revision 279842)
--- gcc/fortran/match.c	(working copy)
*************** select_type_set_tmp (gfc_typespec *ts)
*** 6294,6301 ****
  		    = CLASS_DATA (selector)->attr.dimension;
  	      sym->attr.codimension
  		    = CLASS_DATA (selector)->attr.codimension;
! 	      sym->as
! 		    = gfc_copy_array_spec (CLASS_DATA (selector)->as);
  	    }
  	}
  
--- 6294,6307 ----
  		    = CLASS_DATA (selector)->attr.dimension;
  	      sym->attr.codimension
  		    = CLASS_DATA (selector)->attr.codimension;
! 	      if (CLASS_DATA (selector)->as->type != AS_EXPLICIT)
! 		sym->as = gfc_copy_array_spec (CLASS_DATA (selector)->as);
! 	      else
! 		{
! 		  sym->as = gfc_get_array_spec();
! 		  sym->as->rank = CLASS_DATA (selector)->as->rank;
! 		  sym->as->type = AS_DEFERRED;
! 		}
  	    }
  	}
  
Index: gcc/testsuite/gfortran.dg/select_type_48.f90
===================================================================
*** gcc/testsuite/gfortran.dg/select_type_48.f90	(nonexistent)
--- gcc/testsuite/gfortran.dg/select_type_48.f90	(working copy)
***************
*** 0 ****
--- 1,31 ----
+ ! { dg-do run }
+ !
+ ! Test the fix for PR92976, in which the TYPE IS statement caused an ICE
+ ! because of the explicit bounds of 'x'.
+ !
+ ! Contributed by Gerhard Steinmetz  <gscfq@t-online.de>
+ !
+ program p
+    type t
+       integer :: i
+    end type
+    class(t), allocatable :: c(:)
+    allocate (c, source = [t(1111),t(2222),t(3333)])
+    call s(c)
+    if (sum (c%i) .ne. 3333) stop 1
+ contains
+    subroutine s(x)
+       class(t) :: x(2)
+       select type (x)
+ ! ICE as compiler attempted to assign descriptor to an array
+          type is (t)
+             x%i = 0
+ ! Make sure that bounds are correctly translated.
+             call counter (x)
+       end select
+    end
+    subroutine counter (arg)
+      type(t) :: arg(:)
+      if (size (arg, 1) .ne. 2) stop 2
+    end
+ end