(j3.2006) Submodule/module interface mismatch: is thisstandard-conforming?

[Rafik Zurob]

IBM XL Fortran also flags the error:

$ xlf2008 x.f -qsource
** one_module   === End of Compilation 1 ===
         12 |  integer module function one()
            ....a...........................
a - "x.f", line 12.4: 1513-277 (S) The function result of separate module 
procedure one must have the same type and type parameters as the 
corresponding function result in the module procedure interface body.
** one_submodule   === End of Compilation 2 ===
** _main   === End of Compilation 3 ===
1501-511  Compilation failed for file x.f.
$

I used to sometimes skip implicit none in procedure declarations in 
interface bodies even though I consistently put it in the specification 
section of the enclosing module/program/procedure.  I changed my coding 
style to always include it after I had a one character typo in a dummy 
argument name that caused it be implicitly typed instead of having the TKR 
info from the explicit declaration (with the correct name) I had.  I think 
this case is similar.

Out of curiosity, in the processors where compilation did not fail, do you 
get an error if one() is explicitly given real type in the interface 
block?

Regards

Rafik

j3-bounces at mailman.j3-fortran.org wrote on 07/10/2016 07:22:13 PM:

> From: Damian Rouson <damian at sourceryinstitute.org>
> To: j3 <j3 at j3-fortran.org>
> Cc: GCC-Fortran-ML <fortran at gcc.gnu.org>
> Date: 07/10/2016 07:25 PM
> Subject: (j3.2006) Submodule/module interface mismatch: is this 
> standard-conforming?
> Sent by: j3-bounces at mailman.j3-fortran.org
> 
> 
> All,
> 
> Is the program below standard-conforming?  The Cray compiler gives 
> an error message stating that the IMPLICIT NONE disallows the 
> implicitly typed interface body and and a second error message 
> reporting a conflict between the interface body and and the module 
> function in the submodule.  This is the behavior I would hope would 
> be the intention of the standard, but I?m not certain it is the 
> standard behavior.
> 
> The gfortran 6.1.0 compiler compiles the program without errors and 
> prints the following:
> 
> 9.18340949E-41  is not            1
> 
> The Intel 16.0 compiler compiles the program without errors and 
> prints the following:
> 
>  1.000000      is not            1
> 
> Despite my hopes, I suspect the interface body is standard-
> conforming on the grounds that the IMPLICIT statement does not apply
> to the interface block.  Nonetheless, I suspect that  having an 
> interface body that refers to a procedure with the same names as a 
> different procure in the a child submodule makes the program 
non-conforming.
> 
> Damian
> 
> 
> module one_module
>   implicit none
>   interface
>     module function one()
>     end function
>   end interface
> end module
> 
> submodule(one_module) one_submodule
>   implicit none
> contains
>   integer module function one()
>     one = 1 
>   end function
> end submodule 
> 
> use one_module
> print *,one()," is not ",1
> end 
> 
> 
> _______________________________________________
> J3 mailing list
> J3 at mailman.j3-fortran.org
> http://mailman.j3-fortran.org/mailman/listinfo/j3