Re: [Patch, Fortran] PR 44213: ICE when extending abstract type

[Tobias Schlüter <tobias dot schlueter at physik dot uni-muenchen dot de>]

On 2010-05-21 19:32, Janus Weil wrote:
> > First, can you fix your mailer? It regards both attachments as binaries:
> >
> > Content-Type: application/octet-stream; name="pr44213.diff"
> >
> > Content-Type: application/octet-stream; name="abstract_type_7.f03"
>
> Oops, sorry. Gmail is to blame. But I'm not sure how to fix it ...

By brute force: add .txt to the file extensions :)

Cheers,
- Tobi

> > > Regtested on x86_64-unknown-linux-gnu. Ok for trunk?
> >
> > Secondly, regarding the patch: It looks OK. However, I wonder whether
> > one cannot short-cut the recursion if the parent-type of an abstract
> > type is not abstract. In this case, the checks should have been
> > performed already for the parent type, shouldn't they?
>
> Ah, yes. Actually I think you're right. The attached new version does
> this. I will commit after another regtest.
>
> Cheers,
> Janus
>
>
>
> > > 2010-05-21  Janus Weil<janus@gcc.gnu.org>
> > >
> > >        PR fortran/44213
> > >        * resolve.c (ensure_not_abstract): Allow abstract types with
> > >        non-abstract ancestors.
> > >
> > >
> > > 2010-05-21  Janus Weil<janus@gcc.gnu.org>
> > >
> > >        PR fortran/44213
> > >        * gfortran.dg/abstract_type_7.f03: New.