|
|
On Mon, Sep 11, 2006 at 10:50:46PM +0200, Bernhard Fischer wrote:
> On Wed, Jul 05, 2006 at 03:41:21PM -0700, Steve Kargl wrote:
> >On Thu, Jul 06, 2006 at 12:35:10AM +0200, Paul Thomas wrote:
> >> Steve and Bernhard,
> >> >
> >> >>>
> >> >>>Prove me wrong!
> >> >>>
> >> >>
> >> >I just ran this patch against my large colllection of legal
> >> >Fortran (with extensions :-), and the only failure was in the
> >> >Moware package. The failure occurs with and without the patch. It
> >> >involves a *.F90 that needs to be preprocessed an Moware program,
> >> >and may have screwed up the Makefile sufficiently to get the
> >> >preprocessing mangled.
> >> >
> >> Sounds of humble pie being consumed.... munch, munch, munch!
> >>
> >
> >Well, in truth you are correct that gfc_match_name is only
> >one matcher that gets called often in a large set of possible
> >matchers. One thing to keep in mind with the matchers is
> >that errors are queued. If a match is eventually found, then the
> >queue is tossed out; otherwise, the queue is unwound and
> >errors are emitted.
>
> ping; The comments sounded ok, but I don't think that snipped ever got
> an explicit approval. Don't hesitate to apply if i managed to
> misunderstand any implied ACK :)
>
Unfortunately, this patch is causing regressions in the
testsuite. These regressions are occurring on invalid
code where your patch is triggering a false positive.
We'll need to think about the solution.
--
Steve
|
|