[petsc-dev] Bad fortran interface generations

Stefano Zampini stefano.zampini at gmail.com
Tue Sep 18 13:37:34 CDT 2012


externalpackages/sowing* is empty in my broken source. Bgq has bfort in
/usr/local/bin. actual petsc-dev version recognized it and thus use it.

I switched to require --download-sowing in configure and this solves the
problem. There was any change in config related to don't use sowing by
default recently?

Thanks,
Stefano

2012/9/18 Satish Balay <balay at mcs.anl.gov>

> Can you do rm -rf externalpackages/sowing* and rerun configure - and
> see if the problem persists?
>
> Satish
>
> On Tue, 18 Sep 2012, Stefano Zampini wrote:
>
> > Something is changed between HG revision
> > ce7508606e824d026ea20d05da900e98f37b3864 and the current tip which
> affected
> > the automatic generation of Fortran Interfaces.
> > In particular, PETSc-dev compilation is now broken on some .h90 headers
> > created in include/finclude/ftn-auto/.
> >
> > For example, in include/finclude/ftn-auto/petscvec.h90 for the current
> tip
> >
> >         subroutine VecGhostUpdateBegin(g, insertmode, scattermode ,ierr)&
> >      &
> >
> > whereas for ce7508606e824d026ea20d05da900e98f37b3864
> >
> >         subroutine VecGhostUpdateBegin(g, insertmode, scattermode ,ierr)
> >
> > The form with ampersand is a sintax error for bgxlf90. Removing the two
> > ampersands solves the problem.
> >
> > ......
> >
> "/fermi/home/userprod/propro01/szampini/petsc-dev/include/finclude/petscvec.h90",
> > line 9.7: 1515-019 (S) Syntax is incorrect.
> >
> "/fermi/home/userprod/propro01/szampini/petsc-dev/src/vec/f90-mod/petscvecmod.F",
> > line 25.8: 1513-085 (S) Improper END statement specified. This END
> > statement will be matched with the program unit or procedure named _main.
> > ** _main   === End of Compilation 138 ===
> > ......
> >
> >
>
>


-- 
Stefano
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20120918/666cb772/attachment.html>


More information about the petsc-dev mailing list