Re: [PATCH] various: disallow --no-no-OPT for --no-opt options
- From
Jeff King <peff@peff.net>
- Date
- Apr 19, 2017, 15:05 UTC
- Message-ID
- <20170419150556.ycydwyoi27j7veva@sigill.intra.peff.net>
- In-Reply-To
- <CACBZZX5wHvWgqjBaiD4pJsVQug=+-4xRwkvfZ0uZ=meE3xR6NQ@mail.gmail.com>
On Wed, Apr 19, 2017 at 09:02:52AM +0200, Ævar Arnfjörð Bjarmason wrote:
Show 24 quoted lines
> On Wed, Apr 19, 2017 at 4:50 AM, Jeff King <peff@peff.net> wrote:
> > On Tue, Apr 18, 2017 at 07:40:37PM -0700, Junio C Hamano wrote:
> >
> >> > It might even be possible to detect the existing line and
> >> > have parse-options automatically respect "--foo" when "--no-foo" is
> >> > present. But that may run afoul of callers that add both "--foo" and
> >> > "--no-foo" manually.
> >>
> >> True but wouldn't that something we would want to avoid anyway?
> >> That is, "git cmd [--OPT | --no-OPT | --no-no-OPT]" from the end
> >> user's point of view should be an error because it is unclear what
> >> difference there are between --OPT and --no-no-OPT. And we should
> >> be able to add a rule to parse_options_check() to catch such an
> >> error.
> >
> > I meant that if you have something like this in your options array:
> >
> > { 0, "foo", OPTION_INTEGER, &foo, 1 },
> > { 0, "no-foo", OPTION_INTEGER, &foo, 2 },
>
> I may be missing something, but don't we already do exactly what
> you're describing here? See commit f1930c587 ("parse-options: allow
> positivation of options starting, with no-", 2012-02-25) from René
> Scharfe.Oh, so we do. I somehow totally missed that.
I think all is well, then. We do accept --no-no-foo still. IMHO it is not that big a deal (as long as we do not advertise it, then it does not hurt unless somebody actually tries it). But I do not mind if it goes away, as long as the positive --foo form still works (and it sounds from Jake's response that we'd need to do more than just NONEG there).
-Peff