[Patch, fortran] PR29371 Coredump when using -fbounds-check with pointer & nullify

[Paul Thomas]

I intend to commit this, after it has regtested anew, under the 
"obvious" rule.  On
looking at it again, the patch is fixing manifestly wrong code and is so 
simple as
to be really obvious :-)

Paul

>:ADDPATCH fortran:
>
>This problem is quite specific to NULLIFY or the pointer assignment
>of NULL() when -fbounds-check is used.  The code shows that a temporary
>for the index expression, which is produced by the bounds check, is used
>before it is set.  The reason for this is that the expression is added
>to the block, followed by the se->pre, which carries the code associated
>with bounds checking. The patch is easy; change &block to &se->pre, so
>that everything happens in the right order.
>
>The test case is that of the reporter.
>
>Regtested on amd64/Cygwin_NT - OK for trunk?
>
>It is probable that this patch has gained some whitespace problems
>because of the way in which it was transmitted (copy and paste to a
>Citrix client - my ISP mail server is down.).
>
>Paul
>
>2006-10-12 Paul Thomas <pault@gcc.gnu.org>
>
>	PR fortran/29371
>	* trans-expr.c.c (gfc_trans_pointer_assignment): Add the expression
>	for the assignment of null to the data field to se->pre, rather
>	than block.	
>
>2006-10-12 Paul Thomas <pault@gcc.gnu.org>
>
>	PR fortran/29371
>	* gfortran.dg/nullify_3.f90: New test.
>
>
>Index: gcc/fortran/trans-expr.c
>===================================================================
>*** gcc/fortran/trans-expr.c	(revision 117614)
>--- gcc/fortran/trans-expr.c	(working copy)
>*************** gfc_trans_pointer_assignment (gfc_expr *
>*** 3149,3155 ****
>  	{
>  	case EXPR_NULL:
>  	  /* Just set the data pointer to null.  */
>! 	  gfc_conv_descriptor_data_set (&block, lse.expr, null_pointer_node);
>  	  break;
>  
>  	case EXPR_VARIABLE:
>--- 3149,3155 ----
>  	{
>  	case EXPR_NULL:
>  	  /* Just set the data pointer to null.  */
>! 	  gfc_conv_descriptor_data_set (&lse.pre, lse.expr, null_pointer_node);
>  	  break;
>  
>  	case EXPR_VARIABLE:
>Index: gcc/testsuite/gfortran.dg/nullify_3.f90
>===================================================================
>*** gcc/testsuite/gfortran.dg/nullify_3.f90	(revision 0)
>--- gcc/testsuite/gfortran.dg/nullify_3.f90	(revision 0)
>***************
>*** 0 ****
>--- 1,26 ----
>+ ! { dg-do run }
>+ ! { dg-options "-O0 -fbounds-check" }
>+ ! Tests patch for PR29371, in which the null pointer
>+ ! assignment would cause a segfault with the bounds
>+ ! check on.
>+ !
>+ ! Contributed by Tobias Burnus <tobias.burnus@physik.fu-berlin.de>
>+ !
>+ program test
>+   implicit none
>+   type projector_t
>+     real,   pointer :: ket(:, :), bra(:, :)
>+   end type projector_t
>+ 
>+   type(projector_t),pointer, dimension(:) :: p
>+   integer :: stat,i
>+   allocate(p(2),stat=stat)
>+   do i = 1, 2
>+         nullify(p(i)%bra)
>+         nullify(p(i)%ket)
>+   end do
>+   do i = 1, 2
>+         if (associated (p(i)%bra)) call abort ()
>+         if (associated (p(i)%ket)) call abort ()
>+   end do
>+ end program
>
>
>  
>