[Patch, Fortran] PR38065 Fix checking of contained function

[Patch, Fortran] PR38065 Fix checking of contained function

[Tobias Burnus]

[-- Attachment #1: Type: text/plain, Size: 543 bytes --]

gfortran had two bugs:

a) A function, contained in a public module procedure, which returns a
private derived type, was rejected claiming that the function is public,
which is nonsense

b) A public module function, returning a private derived type was
rejected with -std=f2003, but they are only invalid in -std=f95.

The patch fixes both problems and should be save for the regression only
phase as it is simple and prints an error less often than before, which
makes it safe.

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

Tobias

[-- Attachment #2: privatetype.diff --]
[-- Type: text/plain, Size: 2582 bytes --]

2008-11-11  Tobias Burnus  <burnus@net-b.de>

	PR fortran/38065
	* resolve.c (resolve_fntype): Fix private derived type checking.

2008-11-11  Tobias Burnus  <burnus@net-b.de>

	PR fortran/38065
	* gfortran.dg/private_type_11.f90: New test.
	* gfortran.dg/private_type_12.f90: New test.

Index: gcc/fortran/resolve.c
===================================================================
--- gcc/fortran/resolve.c	(Revision 141763)
+++ gcc/fortran/resolve.c	(Arbeitskopie)
@@ -10179,12 +10179,14 @@ resolve_fntype (gfc_namespace *ns)
     }
 
   if (sym->ts.type == BT_DERIVED && !sym->ts.derived->attr.use_assoc
+      && !sym->attr.contained
       && !gfc_check_access (sym->ts.derived->attr.access,
 			    sym->ts.derived->ns->default_access)
       && gfc_check_access (sym->attr.access, sym->ns->default_access))
     {
-      gfc_error ("PUBLIC function '%s' at %L cannot be of PRIVATE type '%s'",
-		 sym->name, &sym->declared_at, sym->ts.derived->name);
+      gfc_notify_std (GFC_STD_F2003, "Fortran 2003: PUBLIC function '%s' at "
+		      "%L of PRIVATE type '%s'", sym->name,
+		      &sym->declared_at, sym->ts.derived->name);
     }
 
     if (ns->entries)
Index: gcc/testsuite/gfortran.dg/private_type_11.f90
===================================================================
--- gcc/testsuite/gfortran.dg/private_type_11.f90	(Revision 0)
+++ gcc/testsuite/gfortran.dg/private_type_11.f90	(Revision 0)
@@ -0,0 +1,23 @@
+! { dg-do compile }
+! { dg-options "-std=f2003" }
+! PR fortran/38065
+!
+! Reported by Norman S. Clerman
+! and reduced by Joost VandeVondele
+!
+MODULE M1
+  IMPLICIT NONE
+  PRIVATE
+  TYPE T1
+   INTEGER :: I1
+  END TYPE T1
+  PUBLIC :: S1,F2
+CONTAINS
+  SUBROUTINE S1
+  CONTAINS
+   TYPE(T1) FUNCTION F1()
+   END FUNCTION F1
+  END SUBROUTINE S1
+  TYPE(T1) FUNCTION F2()
+  END FUNCTION F2
+END MODULE M1
Index: gcc/testsuite/gfortran.dg/private_type_12.f90
===================================================================
--- gcc/testsuite/gfortran.dg/private_type_12.f90	(Revision 0)
+++ gcc/testsuite/gfortran.dg/private_type_12.f90	(Revision 0)
@@ -0,0 +1,23 @@
+! { dg-do compile }
+! { dg-options "-std=f95" }
+! PR fortran/38065
+!
+! Reported by Norman S. Clerman
+! and reduced by Joost VandeVondele
+!
+MODULE M1
+  IMPLICIT NONE
+  PRIVATE
+  TYPE T1
+   INTEGER :: I1
+  END TYPE T1
+  PUBLIC :: S1,F2
+CONTAINS
+  SUBROUTINE S1
+  CONTAINS
+   TYPE(T1) FUNCTION F1()
+   END FUNCTION F1
+  END SUBROUTINE S1
+  TYPE(T1) FUNCTION F2() ! { dg-error "Fortran 2003: PUBLIC variable 'f2'" }
+  END FUNCTION F2
+END MODULE M1

Re: [Patch, Fortran] PR38065 Fix checking of contained function

[Jerry DeLisle]

Tobias Burnus wrote:
> gfortran had two bugs:
> 
> a) A function, contained in a public module procedure, which returns a
> private derived type, was rejected claiming that the function is public,
> which is nonsense
> 
> b) A public module function, returning a private derived type was
> rejected with -std=f2003, but they are only invalid in -std=f95.
> 
> The patch fixes both problems and should be save for the regression only
> phase as it is simple and prints an error less often than before, which
> makes it safe.
> 
> Build and regtested on x86-64-linux.
> OK for the trunk?
> 
> Tobias
> 
OK, Thanks for patch.

Jerry

Re: [Patch, Fortran] PR38065 Fix checking of contained function

[Hans-Peter Nilsson]

(I don't see it reported anywhere, ignore if duplicate.)

> Date: Tue, 11 Nov 2008 16:09:44 +0100
> From: Tobias Burnus <burnus@net-b.de>

> 2008-11-11  Tobias Burnus  <burnus@net-b.de>
> 
> 	PR fortran/38065
> 	* resolve.c (resolve_fntype): Fix private derived type checking.

It looks like your patch caused a regression on all targets, see
also gcc-testresults:

FAIL: gfortran.dg/private_type_4.f90  -O   (test for errors, line 14)

For my autotester for cris-axis-elf, revision 141775 worked,
while it fails since at least 141781.

brgds, H-P

Re: [Patch, Fortran] PR38065 Fix checking of contained function

[H.J. Lu]

On Wed, Nov 12, 2008 at 7:19 AM, Hans-Peter Nilsson
<hans-peter.nilsson@axis.com> wrote:
> (I don't see it reported anywhere, ignore if duplicate.)
>
>> Date: Tue, 11 Nov 2008 16:09:44 +0100
>> From: Tobias Burnus <burnus@net-b.de>
>
>> 2008-11-11  Tobias Burnus  <burnus@net-b.de>
>>
>>       PR fortran/38065
>>       * resolve.c (resolve_fntype): Fix private derived type checking.
>
> It looks like your patch caused a regression on all targets, see
> also gcc-testresults:
>
> FAIL: gfortran.dg/private_type_4.f90  -O   (test for errors, line 14)
>
> For my autotester for cris-axis-elf, revision 141775 worked,
> while it fails since at least 141781.
>

Yes, this caused:

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38094

-- 
H.J.

Re: [Patch, Fortran] PR38065 Fix checking of contained function

[Tobias Burnus]

Hans-Peter Nilsson wrote:
>> 2008-11-11  Tobias Burnus  <burnus@net-b.de>
>>
>> 	PR fortran/38065
>> 	* resolve.c (resolve_fntype): Fix private derived type checking.
>>     
> It looks like your patch caused a regression on all targets, see
> also gcc-testresults:
>
> FAIL: gfortran.dg/private_type_4.f90  -O   (test for errors, line 14)
>   
Indeed. I wonder why I did not see it with:

   make -j2 check-gfortran//unix/{-m64,-m32}

(One answer could be that I might have misread a part result as final
result.) I have now patched gfortran.dg/private_type_4.f90 and verified
using a simple "make check-gfortran".


The patch itself is OK, the test merely needed a "-std=f95" as the
program is valid in Fortran 2003.
Committed the attached patch as obvious (Rev. 141798).

Thanks for the report.

Tobias

Index: testsuite/ChangeLog
===================================================================
--- testsuite/ChangeLog (Revision 141797)
+++ testsuite/ChangeLog (Arbeitskopie)
@@ -1,3 +1,9 @@
+2008-11-12  Tobias Burnus  <burnus@net-b.de>
+
+       PR fortran/38065
+       PR fortran/38094
+       * gfortran.dg/private_type_4.f90: Add -std=f95 option.
+
 2008-11-12  Andreas Krebbel  <krebbel1@de.ibm.com>

        * gcc.target/s390/frame-addr1.c: New testcase.
Index: testsuite/gfortran.dg/private_type_4.f90
===================================================================
--- testsuite/gfortran.dg/private_type_4.f90    (Revision 141797)
+++ testsuite/gfortran.dg/private_type_4.f90    (Arbeitskopie)
@@ -1,5 +1,8 @@
 ! { dg-do compile }
+! { dg-options "-std=f95" }
 ! PR 25093: Check that a PUBLIC function can't be of PRIVATE type
+! in Fortran 95; in Fortran 2003 it is allowed (cf. PR fortran/38065)
+!
 module m1

     type :: t1