Bug 46991 - [OOP] polymorphic assumed-size actual arguments
Summary: [OOP] polymorphic assumed-size actual arguments
Status: NEW
Alias: None
Product: gcc
Classification: Unclassified
Component: fortran (show other bugs)
Version: 4.6.0
: P3 normal
Target Milestone: ---
Assignee: Not yet assigned to anyone
URL:
Keywords: rejects-valid
Depends on:
Blocks:
 
Reported: 2010-12-17 10:41 UTC by Tobias Burnus
Modified: 2013-12-19 08:29 UTC (History)
3 users (show)

See Also:
Host:
Target:
Build:
Known to work:
Known to fail:
Last reconfirmed: 2013-12-19 00:00:00


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Tobias Burnus 2010-12-17 10:41:38 UTC
Found at http://j3-fortran.org/pipermail/j3/2010-December/004084.html

Currently, it is unclear whether the following program is valid, but Robert Corbett of Oracle (formerly Sun) thinks it is (but does not seem to like it.)

[100% guessing, 90% wrongly]: The main issue he sees is seemingly that SUB2 does not need to have an explicit interface - which only works if the $data of a CLASS and a TYPE object start at the same offset of the passed pointer address.

I think that part is OK in gfortran, but gfortran has other issues:


Gfortran rejects the program with:

        CALL SUB2(A, N)
                  1
Warning: Type mismatch in argument 'a' at (1); passed CLASS(rec) to TYPE(rec)

[That's PR 46990.] I also tried SELECT TYPE but with limited success.


Using
        CLASS(REC) A(*)
one gets:
        PRINT *, A(:N)%A
                   1
Error: Syntax error in argument list at (1)


 * * *

       MODULE TYPES
         PRIVATE
         PUBLIC REC, REC2

         TYPE REC
           INTEGER A
         END TYPE

         TYPE, EXTENDS(REC) :: REC2
           INTEGER B
         END TYPE
       END

       SUBROUTINE SUB1(A, N)
         USE TYPES
         CLASS(REC), INTENT(IN) :: A(*)

         CALL SUB2(A, N)
       END

       SUBROUTINE SUB2(A, N)
         USE TYPES
         TYPE(REC) A(*)

         PRINT *, A(:N)%A
       END

       PROGRAM MAIN
         USE TYPES
         CLASS(REC), ALLOCATABLE :: A(:)
         INTERFACE
           SUBROUTINE SUB1(A, N)
             USE TYPES
             CLASS(REC), INTENT(IN) :: A(*)
           END SUBROUTINE
         END INTERFACE

         A = [ (REC2(I, I+1), I = 1, 10) ]
         CALL SUB1(A, 10)
       END
Comment 1 Harald Anlauf 2012-12-23 22:17:25 UTC
The code is rejected by nagfor:

NAG Fortran Compiler Release 5.3.1(907)
Error: pr46991.f90, line 16: CLASS array A must be a pointer, allocatable or assumed-shape
Errors in declarations, no further processing for SUB1
Error: pr46991.f90, line 34: CLASS array A must be a pointer, allocatable or assumed-shape

Changing from assumed size to assumed shape, however, make the
code compilable.

The code modified this way still does not compile with current trunk:

         A = [ (REC2(I, I+1), I = 1, 10) ]
         1
Error: Variable must not be polymorphic in intrinsic assignment at (1) - check that there is a matching specific subroutine for '=' operator

It is funny to see that ifort 13 generates a similar message:

pr46991.f90(38): error #6197: An assignment of different structure types is invalid.
         A = [ (REC2(I, I+1), I = 1, 10) ]
----------------^
pr46991.f90(38): error #8304: In an intrinsic assignment statement, variable shall not be polymorphic.   [A]
         A = [ (REC2(I, I+1), I = 1, 10) ]
---------^
Comment 2 Tobias Burnus 2012-12-23 22:33:24 UTC
(In reply to comment #1)
> Error: pr46991.f90, line 16: CLASS array A must be a pointer, allocatable or
> assumed-shape

I think that's a bug in NAG - at least, I cannot find anything in the standard which prohibits explicit-size or assumed-size polymorphic arrays.


> The code modified this way still does not compile with current trunk:
>          A = [ (REC2(I, I+1), I = 1, 10) ]
>          1
> Error: Variable must not be polymorphic in intrinsic assignment at (1) - check
> that there is a matching specific subroutine for '=' operator

See PR43366. That's a Fortran 2008 feature. Assignment to a polymorphic LHS variable is allowed, if and only if it is allocatable; it makes use of the realloc-on-assignment feature.
Comment 3 Dominique d'Humieres 2013-12-19 08:29:06 UTC
Compiling the test in comment 0 with gfortran 4.8.2 or trunk r206033 gives

pr46991.f90:16.39:

         CLASS(REC), INTENT(IN) :: A(*)
                                       1
Error: Assumed size polymorphic objects or components, such as that at (1), have not yet been implemented
pr46991.f90:34.43:

             CLASS(REC), INTENT(IN) :: A(*)
                                           1
Error: Assumed size polymorphic objects or components, such as that at (1), have not yet been implemented
pr46991.f90:18.19:

         CALL SUB2(A, N)
                   1
Warning: Type mismatch in argument 'a' at (1); passed REAL(4) to TYPE(rec)
pr46991.f90:38.9:

         A = [ (REC2(I, I+1), I = 1, 10) ]
         1
Error: Variable must not be polymorphic in intrinsic assignment at (1) - check that there is a matching specific subroutine for '=' operator
pr46991.f90:39.19:

         CALL SUB1(A, 10)
                   1
Error: Type mismatch in argument 'a' at (1); passed CLASS(rec) to REAL(4)