[Patch, Fortran] PR fortran/35840: asynchronous specifier for a data transfert statement shall be an initialization expression

[Patch, Fortran] PR fortran/35840: asynchronous specifier for a data transfert statement shall be an initialization expression

[Mikael Morin]

[-- Attachment #1: Type: text/plain, Size: 633 bytes --]

Hi,

this patch fixes PR35840.
It takes checking code from gfc_match_init_expr to a separated function
and uses it to check exotic (non-constant, but reducible) asynchronous
specifier ("y"//"e"//trim("s  ") which reduces to "yes" was the one
reported).

This was first posted on the bugzilla, and pre-approved there.
It is regression tested.

Is this Ok ?*
Mikael.

PS : My copyright assignment should reach the FSF by the end of this
week or more probably next week.

(*) This is my first patch proposal so I'm awaiting comments about
everything (patch, Changelog, message, syntactic conventions, testcase,
...). Be very strict.




[-- Attachment #2: Changes --]
[-- Type: text/plain, Size: 535 bytes --]

2008-10-13  Mikael Morin  <mikael.morin@tele2.fr>

	PR fortran/35840
	* expr.c (gfc_reduce_init_expr): New function, containing checking code
	from gfc_match_init_expr, so that checking can be deferred. 
	(gfc_match_init_expr): Use gfc_reduce_init_expr.
	* io.c (check_io_constraints): Use gfc_reduce_init_expr instead of 
	checking that the expression is a constant. 
	* match.h (gfc_reduce_init_expr): Prototype added. 

2008-10-13  Mikael Morin  <mikael.morin@tele2.fr>

	PR fortran/35840
	* gfortran.dg/write_check4.f90: New test.

[-- Attachment #3: 35840.diff --]
[-- Type: text/plain, Size: 3947 bytes --]

Index: gcc/fortran/expr.c
===================================================================
--- gcc/fortran/expr.c	(révision 141074)
+++ gcc/fortran/expr.c	(copie de travail)
@@ -2378,21 +2378,15 @@
   return t;
 }
 
+/* Reduces a general expression to an initialization expression (a constant).
+   This used to be part of gfc_match_init_expr.
+   Note that this function doesn't free the given expression on FAILURE.  */
 
-/* Match an initialization expression.  We work by first matching an
-   expression, then reducing it to a constant.  */
-
-match
-gfc_match_init_expr (gfc_expr **result)
+gfc_try
+gfc_reduce_init_expr (gfc_expr *expr)
 {
-  gfc_expr *expr;
-  match m;
   gfc_try t;
 
-  m = gfc_match_expr (&expr);
-  if (m != MATCH_YES)
-    return m;
-
   gfc_init_expr = 1;
   t = gfc_resolve_expr (expr);
   if (t == SUCCESS)
@@ -2400,18 +2394,12 @@
   gfc_init_expr = 0;
 
   if (t == FAILURE)
-    {
-      gfc_free_expr (expr);
-      return MATCH_ERROR;
-    }
+    return FAILURE;
 
   if (expr->expr_type == EXPR_ARRAY
       && (gfc_check_constructor_type (expr) == FAILURE
-	  || gfc_expand_constructor (expr) == FAILURE))
-    {
-      gfc_free_expr (expr);
-      return MATCH_ERROR;
-    }
+      || gfc_expand_constructor (expr) == FAILURE))
+    return FAILURE;
 
   /* Not all inquiry functions are simplified to constant expressions
      so it is necessary to call check_inquiry again.  */ 
@@ -2419,6 +2407,33 @@
       && !gfc_in_match_data ())
     {
       gfc_error ("Initialization expression didn't reduce %C");
+      return FAILURE;
+    }
+
+  return SUCCESS;
+}
+
+
+/* Match an initialization expression.  We work by first matching an
+   expression, then reducing it to a constant.  */
+
+match
+gfc_match_init_expr (gfc_expr **result)
+{
+  gfc_expr *expr;
+  match m;
+  gfc_try t;
+
+  expr = NULL;
+
+  m = gfc_match_expr (&expr);
+  if (m != MATCH_YES)
+    return m;
+
+  t = gfc_reduce_init_expr (expr);
+  if (t != SUCCESS)
+    {
+      gfc_free_expr (expr);
       return MATCH_ERROR;
     }
 
Index: gcc/fortran/io.c
===================================================================
--- gcc/fortran/io.c	(révision 141074)
+++ gcc/fortran/io.c	(copie de travail)
@@ -2973,7 +2973,7 @@
     {
       static const char * asynchronous[] = { "YES", "NO", NULL };
 
-      if (dt->asynchronous->expr_type != EXPR_CONSTANT)
+      if (gfc_reduce_init_expr (dt->asynchronous) != SUCCESS)
 	{
 	  gfc_error ("ASYNCHRONOUS= specifier at %L must be an initialization "
 		     "expression", &dt->asynchronous->where);
Index: gcc/fortran/match.h
===================================================================
--- gcc/fortran/match.h	(révision 141074)
+++ gcc/fortran/match.h	(copie de travail)
@@ -199,6 +199,7 @@
 /* expr.c -- FIXME: this one should be eliminated by moving the
    matcher to matchexp.c and a call to a new function in expr.c that
    only makes sure the init expr. is valid.  */
+gfc_try gfc_reduce_init_expr (gfc_expr *expr);
 match gfc_match_init_expr (gfc_expr **);
 
 /* array.c.  */
Index: gcc/testsuite/gfortran.dg/write_check4.f90
===================================================================
--- gcc/testsuite/gfortran.dg/write_check4.f90	(révision 0)
+++ gcc/testsuite/gfortran.dg/write_check4.f90	(révision 0)
@@ -0,0 +1,17 @@
+! { dg-do compile }
+!
+! PR fortran/35840 
+!
+! The asynchronous specifier for a data transfer statement shall be 
+! an initialization expression
+!
+! Contributed by Tobias Burnus <burnus@gcc.gnu.org>
+!
+  character(2) :: no
+  no = "no"
+  open (unit=10, asynchronous = no)              ! Ok, it isn't a transfer stmt
+  write(*,*, asynchronous="Y"//"E"//trim("S  ")) ! Ok, it is an init expr
+  write(*,*, asynchronous=no)  ! { dg-error "must be an initialization expression" } 
+  read (*,*, asynchronous="Y"//"e"//trim("S  "))
+  read (*,*, asynchronous=no)  ! { dg-error "must be an initialization expression" }
+end

Re: [Patch, Fortran] PR fortran/35840: asynchronous specifier for a data transfert statement shall be an initialization expression

[Mikael Morin]

Hello,

same here:
ping and blablabla commit blablabla.

Mikael


Mikael Morin wrote:
> Hi,
> 
> this patch fixes PR35840.
> It takes checking code from gfc_match_init_expr to a separated function
> and uses it to check exotic (non-constant, but reducible) asynchronous
> specifier ("y"//"e"//trim("s  ") which reduces to "yes" was the one
> reported).
> 
> This was first posted on the bugzilla, and pre-approved there.
> It is regression tested.
> 
> Is this Ok ?*
> Mikael.
> 
> PS : My copyright assignment should reach the FSF by the end of this
> week or more probably next week.
> 
> (*) This is my first patch proposal so I'm awaiting comments about
> everything (patch, Changelog, message, syntactic conventions, testcase,
> ...). Be very strict.
> 
> 
> 
> 

Re: [Patch, Fortran] PR fortran/35840: asynchronous specifier for a data transfert statement shall be an initialization expression

[Tobias Burnus]

Mikael Morin wrote:
>> This was first posted on the bugzilla, and pre-approved there.
>> It is regression tested.
>>
>> Is this Ok ?*
>>     
A belate OK (since already checked in as Rev. 141497), but granted it
was pre-approved by Jerry in the PR.

Tobias

Re: [Patch, Fortran] PR fortran/35840: asynchronous specifier for a data transfert statement shall be an initialization expression

[Mikael Morin]

Patch committed as revision 141497.
I also fixed a wrong date in the ChangeLog for my previous commit.
For Tobias, sorry I have just got your e-mail.
This patch was reviewed on bugzilla only, I didn't get any feedback
after posting the patch on the list. But it is almost unchanged, with
just a testcase and a ChangeLog entry added.


Mikael