Bug 71967

Summary: Protected and private specified at the same time and the symbol is accessible
Product: gcc Reporter: Vladimir Fuka <vladimir.fuka>
Component: fortranAssignee: Not yet assigned to anyone <unassigned>
Status: RESOLVED INVALID    
Severity: normal Keywords: accepts-invalid
Priority: P3    
Version: 5.3.1   
Target Milestone: 4.9.4   
Host: Target:
Build: Known to work:
Known to fail: Last reconfirmed: 2016-07-22 00:00:00

Description Vladimir Fuka 2016-07-22 10:17:46 UTC 
This program compiles

module m
 protected :: k
 private :: k
 
 integer k
end module

 use m
 print *, k
end

but I believe it should not.
Comment 1 Martin Liška 2016-07-22 11:16:38 UTC 
Confirmed, all GCC releases are affected (I've got 4.5+).
ICC compiler rejects the snippet:

/home/marxin/Programming/testcases/pr71967.f90(2): error #6406: Conflicting attributes or multiple declaration of name.   [K]
 protected :: k
--------------^
/home/marxin/Programming/testcases/pr71967.f90(8): error #7013: This module file was not generated by any release of this compiler.   [M]
 use m
-----^
compilation aborted for /home/marxin/Programming/testcases/pr71967.f90 (code 1)
Comment 2 Dominique d'Humieres 2016-07-22 11:31:45 UTC 
> but I believe it should not.

Why? If you think it is because you have

 protected :: k
 private :: k

I did not find such a restriction in the standard.
Comment 3 Vladimir Fuka 2016-07-22 11:40:22 UTC 
OK, protected is not an access-stmt.

But in that case, why is the private access-stmt ignored?
Comment 4 Dominique d'Humieres 2016-07-22 12:04:38 UTC 
> But in that case, why is the private access-stmt ignored?

It is not: consider

module m
 protected :: k
 private :: k
 
 integer :: k = 42
end module

 use m
 integer :: k = 3
 print *, k
end

'k' is set to 42 in module m, but is not visible in the main program where it is set to 3.
Comment 5 Vladimir Fuka 2016-07-22 12:16:07 UTC 
OMG, I forgot implicit none and the only clause. Sorry. So is ifort wrong
in treating protected like an access statement?
Dne 22. 7. 2016 13:04 napsal uživatel "dominiq at lps dot ens.fr" <
gcc-bugzilla@gcc.gnu.org>:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71967
>
> --- Comment #4 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
> > But in that case, why is the private access-stmt ignored?
>
> It is not: consider
>
> module m
>  protected :: k
>  private :: k
>
>  integer :: k = 42
> end module
>
>  use m
>  integer :: k = 3
>  print *, k
> end
>
> 'k' is set to 42 in module m, but is not visible in the main program where
> it
> is set to 3.
>
> --
> You are receiving this mail because:
> You reported the bug.
Comment 6 Dominique d'Humieres 2016-07-22 12:23:20 UTC 
> OMG, I forgot implicit none and the only clause.

module m
 protected :: k
 private :: k
 
 integer k
end module

 use m, only : k
 implicit none
 print *, k
end

is rejected by gfortran with

pr71967_db_1.f90:8:14:

  use m, only : k
              1
Error: Symbol 'k' referenced at (1) not found in module 'm'
pr71967_db_1.f90:10:11:

  print *, k
           1
Error: Symbol 'k' at (1) has no IMPLICIT type

>  So is ifort wrong in treating protected like an access statement?

No idea, but IMO

 protected :: k
 public :: k

should be valid.

Note that if ifort is right, I'ld like to see where it is said so in the standard.
Comment 7 Vladimir Fuka 2016-07-22 12:39:21 UTC 
I think we can mark this as INVALID then. The protected attribute is indeed not an access attribute.
Comment 8 Dominique d'Humieres 2016-07-22 15:43:31 UTC 
> I think we can mark this as INVALID then. ...

Doing so.