Re: [PATCH 4/9] Warnings win over strictness in AM_INIT_AUTOMAKE.

[Ralf Wildenhues]

* Stefano Lattarini wrote on Wed, Jan 05, 2011 at 08:22:13PM CET:
> On Wednesday 05 January 2011, Ralf Wildenhues wrote:
> > * Stefano Lattarini wrote on Tue, Jan 04, 2011 at 06:38:04PM CET:
> > 
> > > +  # will take precedence over warning settings defined implicitly by the
> > > +  # strictness.
> > 
> > Well, this works in the current code base, but only by accident: namely,
> > only because process_option_list is only ever called once, and with all
> > options at once.
> >
> Hmm... no, it's potentially called many times in `handle_options()'.
> But the later [PATCH 7/9] takes care of this.

Ah, ok.

> >  If some code later calls it like
> >   process_option_list (first-set-of-options);
> >   process_option_list (second-set-of-options);
> > 
> > then things will go wrong again.  I suspect that it will mean that
> >   AM_INIT_AUTOMAKE([foreign -Wno-portability])
> >   AUTOMAKE_OPTIONS = gnu
> > 
> > won't do what we want.  Hmm.  What exactly is it that we want to happen
> > in this case?  Should gnu override -Wno-portability if specified in a
> > (to-be) higher order place?
> >
> I assumed without saying that yes, this was to be the intended behaviour.
> And I still think it should be.  Sorry for not having been explicit about
> that before.
> 
> > I see two ways out: warnings are only switched after all options are
> > processed.
> >
> This is not good IMO, as it breaks usages like the the one in your
> example above.

Makes sense.

Thanks for explaining patiently, I think I now understand better.  I
hope to finish review (and approval) of this patch series this weekend.

Cheers,
Ralf