Re: [gentoo-dev] Python 3.1: Stabilization and news item

[William Hubbs]

On Wed, Mar 24, 2010 at 08:57:20PM +0100, Thomas Sachau wrote:
> Am 24.03.2010 19:03, schrieb William Hubbs:
> > On Wed, Mar 24, 2010 at 05:47:18PM +0000, Jeremy Olexa wrote:
> >>
> >> On Wed, 24 Mar 2010 10:32:37 -0700, Joshua Saddler <nightmorph@g.o>
> >> wrote:
> >>>>> But everyone else in Gentoo does, so . . .
> >
> > Really?  I've seen a few people object, but not everyone in gentoo.
> >
> >>>>
> >>>> Some Gentoo developers/users, who aren't Python maintainers, said that
> >>>> they didn't object to have Python 3 installed.
> >>>
> >>> They're in the minority, judging by the replies in this thread.
> >>
> >> I hate to get into the mix of this, but I suggest researching on "vocal
> >> minority" and/or "silent majority" - the most vocal ones on this thread are
> >> the minority of the population. I'm not attacking anyone, mind you.
> >>
> >> I haven't expressed anything on this thread but I'm ok with marking it
> >> stable and having concerned users mask it. The stages might get kinda funky
> >> with both python-2 and 3 on them, but..if they are not BROKEN, I don't
> >> care.
> >
> > I tend to agree with this.  I don't think it is right to force everyone
> > to wait until most of the tree works with python3 before it goes stable.
> > That is why python is slotted; it is possible to have both versions
> > installed at the same time.  If we have packages in the tree that are
> > pulling in both versions of python but are not compatible with them,
> > their dependencies need to be fixed.  If users do not want python-3 on
> > their systems, that is what /etc/portage/package.mask is for.
> >
> > If we are going to make everyone wait until python-3 works with most
> > packages in the tree, let's un-slot all versionf of python and hard mask
> > python-3.
> >
> > William
> >
>
> Who said, that we are against a stable python-3 version?
>
> The main point (as already pointed out in my previous thread about python-3) is, that it is not in
> any way required or used. But there are still wrong dependencies (where Arfrever just closes bugs as
> invalid) and packages like the mentioned "setuptools", which will always pull in python-3.

That is because setuptools works with both versions of python, and if a
user wants both versions of python on their system they will need
setuptools installed for both versions.

You say there are "wrong dependencies".  How are they wrong?  I mean, do
the packages with dev-lang/python in their deps not work with both
versions of python?  If they don't, they need to be fixed.  If they do,
they are correct.

> Why should we pull in python-3 for ever user, force the usual user to install a useless python-3 and
> additional files in python-3 path for many python packages? The minimum would be to tell them, that
> this python version is currently useless and they have the option to mask it locally. And i really
> dont think, that the default stable user knows, that python-3 is not really needed and can be
> masked, usually the pulled in dependencies are required, so he will expect the same for python-3.

If we make it clear in the news item that python-3 cannot be used as the
default python, so if users do not want it they should mask it, we have
done our job imho.  In other words, this is just a matter of informing
users.

William