Re: SGI introduced a bug in 7.41 compiler?

From: Steve Siegfried (sos_at_zjod.net)
Date: 09/23/04 

Date: Thu, 23 Sep 2004 21:03:31 GMT

In msg <cipg5h$92j$1@news2.news.larc.nasa.gov>, Bil Kleb <Bil.Kleb@NASA.Gov>
wrote:
>
> According to ifc, ifort, lf95, nagf95, nasf95, pgf90, and the
> 7.3.1.2m version of SGI's f90 compiler, the follow code is ok:
>
> module cat
> integer :: a = 1
> end module cat
>
> module puppy
> contains
> subroutine one
> use cat, only : a_local=>a
> write(*,*) 'a_local = ', a_local
> end subroutine one
> subroutine two
> use cat, only : a_local=>a
> write(*,*) 'a_local = ', a_local
> end subroutine two
> end module puppy
>
> program main
> use puppy, only : one, two
> call one
> call two
> end program main
>
> However, version 7.41 of SGI's compiler does not think so:
>
> % f90 -version
> MIPSpro Compilers: Version 7.41
> % f90 local_rename.f90
>
> module puppy
> ^
> f90-855 f90: ERROR PUPPY, File = local_rename.f90, Line = 12, Column = 8
> The compiler has detected errors in module "PUPPY". No module information file will be crea
> ted for this module.
>
> use cat, only : a_local=>a
> ^
> f90-1015 f90: ERROR TWO, File = local_rename.f90, Line = 21, Column = 21
> Local-name "A_LOCAL" must only be referenced once in a rename-list in this scope.
>
> use puppy, only : one, two
> ^
> f90-894 f90: ERROR MAIN, File = local_rename.f90, Line = 31, Column = 7
> Module "PUPPY" has compile errors, therefore declarations obtained from the module via the U
> SE statement may be incomplete.
>
> f90: SGI MIPSpro Fortran 90 Version 7.42 (f14) Tue Sep 21, 2004 10:54:50
>
> Bug?
>
> --
> Bil Kleb, Hampton, Virginia
>

Yeah, and somebody caught it. In (the as-yet unreleased) 7.4.3m compilers,
this is back to compiling cleanly. If the brokenness of this interferes
with your Fortran programming pleasure, please consider submitting a bug
report to the usual suspects (who are usually more than happy to push fixes
back into previous releases). BTW: latest supported version is 7.4.2m, not
7.4.1m.

-Steve Siegfried
 sos@zjod.net (also one of the usual suspects at sosATsgiDOTcom, too ;-)