[Patch, Fortran] PR 43362 - PURE contraint - fix ICE, add missing check

[Patch, Fortran] PR 43362 - PURE contraint - fix ICE, add missing check

[Tobias Burnus]

[-- Attachment #1: Type: text/plain, Size: 1044 bytes --]

Found when adding the equivalent check for coarrays. There are two
issues, one NULL pointer access for a valid case (item (4) does not
apply) as accessing "rhs->symtree->n.sym"  causes a segfault if RHS is
not an EXPR_VARIABLE but, e.g., a structure constructor.

The other issue is that gfortran did not have a check for constraint (3).

Quoting Fortran 2003:

"C1272 In a pure subprogram any designator with a base object that is in
common or accessed by host or use association, is a dummy argument of a
pure function, is a dummy argument with INTENT (IN) of a pure
subroutine, or an object that is storage associated with any such
variable, shall not be used in the following contexts:
[...]
(3) As the expr corresponding to a component with the POINTER attribute
in a structure constructor.
(4) As the expr of an intrinsic assignment statement in which the
variable is of a derived type if the derived type has a pointer
component at any level of component selection; [...]"

Build and regtested on x86-64-linux.
OK for the trunk?

Tobias

[-- Attachment #2: impure.diff --]
[-- Type: text/x-patch, Size: 2285 bytes --]

2010-03-14  Tobias Burnus  <burnus@net-b.de>

	PR fortran/43362
	* resolve.c (resolve_structure_cons): Add missing PURE constraint.
	(resolve_ordinary_assign): Add check to avoid segfault.

2010-03-14  Tobias Burnus  <burnus@net-b.de>

	PR fortran/43362
	* gfortran.dg/impure_constructor_1.f90: New test.

Index: gcc/fortran/resolve.c
===================================================================
--- gcc/fortran/resolve.c	(Revision 157445)
+++ gcc/fortran/resolve.c	(Arbeitskopie)
@@ -921,6 +921,16 @@ resolve_structure_cons (gfc_expr *expr)
 		     "for pointer component '%s' should be a POINTER or "
 		     "a TARGET", &cons->expr->where, comp->name);
 	}
+
+      /* F2003, C1272 (3).  */
+      if (gfc_pure (NULL) && cons->expr->expr_type == EXPR_VARIABLE
+	  && gfc_impure_variable (cons->expr->symtree->n.sym))
+	{
+	  t = FAILURE;
+	  gfc_error ("Invalid expression in the derived type constructor for pointer "
+		     "component '%s' at %L in PURE procedure", comp->name,
+		     &cons->expr->where);
+	}
     }
 
   return t;
@@ -7947,6 +7957,7 @@ resolve_ordinary_assign (gfc_code *code,
       if (lhs->ts.type == BT_DERIVED
 	    && lhs->expr_type == EXPR_VARIABLE
 	    && lhs->ts.u.derived->attr.pointer_comp
+	    && rhs->expr_type == EXPR_VARIABLE
 	    && gfc_impure_variable (rhs->symtree->n.sym))
 	{
 	  gfc_error ("The impure variable at %L is assigned to "
Index: gcc/testsuite/gfortran.dg/impure_constructor_1.f90
===================================================================
--- gcc/testsuite/gfortran.dg/impure_constructor_1.f90	(Revision 0)
+++ gcc/testsuite/gfortran.dg/impure_constructor_1.f90	(Revision 0)
@@ -0,0 +1,30 @@
+! { dg-do compile }
+!
+! PR fortran/43362
+!
+module m
+  implicit none
+  type t
+    integer, pointer :: a
+  end type t
+  type t2
+    type(t) :: b
+  end type t2
+  type t3
+    type(t), pointer :: b
+  end type t3
+contains
+ pure subroutine foo(x)
+   type(t), target, intent(in) :: x
+   type(t2) :: y
+   type(t3) :: z
+
+   ! The following gave an ICE but is valid:
+   y = t2(x) ! Note: F2003, C1272 (3) and (4) do not apply
+   
+   ! Variant which is invalid as C1272 (3) applies
+   z = t3(x) ! { dg-error "Invalid expression in the derived type constructor" }
+ end subroutine foo
+end module m
+
+

Re: [Patch, Fortran] PR 43362 - PURE contraint - fix ICE, add missing check

[Jerry DeLisle]

Tobias Burnus wrote:
> Found when adding the equivalent check for coarrays. There are two
> issues, one NULL pointer access for a valid case (item (4) does not
> apply) as accessing "rhs->symtree->n.sym"  causes a segfault if RHS is
> not an EXPR_VARIABLE but, e.g., a structure constructor.
> 
> The other issue is that gfortran did not have a check for constraint (3).
> 
> Quoting Fortran 2003:
> 
> "C1272 In a pure subprogram any designator with a base object that is in
> common or accessed by host or use association, is a dummy argument of a
> pure function, is a dummy argument with INTENT (IN) of a pure
> subroutine, or an object that is storage associated with any such
> variable, shall not be used in the following contexts:
> [...]
> (3) As the expr corresponding to a component with the POINTER attribute
> in a structure constructor.
> (4) As the expr of an intrinsic assignment statement in which the
> variable is of a derived type if the derived type has a pointer
> component at any level of component selection; [...]"
> 
> Build and regtested on x86-64-linux.
> OK for the trunk?
> 
> Tobias
> 
OK, thanks

Jerry