[committed] Fix OpenMP parsing of the specification part in functions (PR fortran/71704)
Jakub Jelinek
jakub@redhat.com
Fri Jul 8 16:18:00 GMT 2016
On Fri, Jul 08, 2016 at 09:13:50AM -0700, Cesar Philippidis wrote:
> On 06/30/2016 10:47 AM, Jakub Jelinek wrote:
>
> > The Fortran parser apparently relies in functions that have still undecided
> > kind of the result that ST_GET_FCN_CHARACTERISTICS artificial statement is
> > returned before any executable statements in the function.
> > In normal statements that is ensured through decode_statement calling
> > decode_specification_statement, which parses just a subset of statements,
> > but for OpenMP we need to do something similar. If we figure out we want
> > only the case_omp_decl statements, for any other we just try to gfc_match
> > the keyword and if we match it, it means we'd be about to return an OpenMP
> > executable statement, so instead return ST_GET_FCN_CHARACTERISTICS.
> >
> > Bootstrapped/regtested on x86_64-linux and i686-linux, committed to trunk,
> > queued for 6.2 backport.
> >
> > Cesar, note OpenACC will need something similar (though,
> > decode_acc_statement uses just the match macro, so you'll need another one
> > for the executable statements).
>
> Here's the OpenACC followup for this patch. Is it OK for trunk and gcc6?
ENOPATCH
Jakub
More information about the Gcc-patches
mailing list