------- Comment #11 from sfilippone at uniroma2 dot it 2007-02-15 10:42 -------
(In reply to comment #10)
> > > I have still to re-read the test case to check whether TARGET is required
> >
> > However the accessed component is a POINTER to a derived type [...]
>
> Ok, I somehow didn't realize
> type field
> type(mesh), pointer :: msh => null()
> when I read
> msh_ => fld%msh
> One does not assign msh_ the variable fld%msh as target, but one assignes the
> target of the pointer fld%msh to the pointer msh_.
Yes, exactly: that's an object that has an independent life, is pointed to from
the field object, and I want to get access to it.
>
> But I don't understand why xlf does not like the TARGET attribute.
>
Beats me. But I guess I'd be happy enough to have gfortran working, for the
time being.
Actually I already have a workaround: instead of a function, put a subroutine
call get_mesh(mshp,fld)
which internally does
mshp => fld%msh
works perfectly. Why?
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30793
