[PATCH 5 of 5] commands: always register special command properties in registrar (API)

[Yuya Nishihara]

On Sat, 27 Feb 2016 22:06:14 -0800, Gregory Szorc wrote:
> On Thu, Jan 14, 2016 at 6:36 AM, Yuya Nishihara <yuya at tcha.org> wrote:
> > On Wed, 13 Jan 2016 14:35:36 -0600, Matt Mackall wrote:
> > > On Mon, 2016-01-04 at 00:17 +0900, Yuya Nishihara wrote:
> > > > On Sat, 2 Jan 2016 18:11:07 -0800, Gregory Szorc wrote:
> > > > > On Sat, Jan 2, 2016 at 5:51 PM, Yuya Nishihara <yuya at tcha.org>
> > wrote:
> > > > > > > This is marked as API incompatible because extensions examining
> > > > > > > commands.{norepo,optionalrepo,inferrepo} will no longer work
> > since this
> > > > > > > variable is no longer populated at extension load time. This is
> > > > > > > demonstrated by the simple change to mq.py.
> > > > > >
> > > > > > If we can drop commands.{norepo,optionalrepo,inferrepo}, API
> > compatibility
> > > > > > issues will be more discoverable.
> > > > > >
> > > > >
> > > > > I'd like to drop them too. But I think it may have significant
> > implications
> > > > > on legacy extensions. Perhaps this is a mpm decision?
> > > >
> > > > CC-ing mpm. :-)
> > >
> > > I love to break extensions, but perhaps we shouldn't do it so wholesale
> > so close
> > > to the freeze.
> >
> > Ok, I'll patchbomb it next month. Thanks.
> 
> Yuya: What is the state of your patches? I was looking at Python 3
> compatibility again today and commands.py and cmdutil.py not using
> absolute_import is blocking some porting work, which a number of people
> have expressed interest in working on. I think addressing command
> registration is beyond the skills of a new contributor. So I'd like to get
> these 2 modules sorted out so people can start hacking on the long tail of
> Python 3 work in mercurial.*.

Flushed. Sorry, they were pushed near the bottom of my unsent stack as they
weren't a blocker of my work. :-)