Re: [gentoo-dev] Re: Re: RFC: Make 10.0 profiles EAPI-2 'compliant'

[Ciaran McCreesh <ciaran.mccreesh@googlemail.com>]

[-- Attachment #1: Type: text/plain, Size: 2616 bytes --]

On Thu, 13 Aug 2009 19:22:16 +0100
Steven J Long <slong@rathaus.eclipse.co.uk> wrote:
> > PMS accurately reflects the Portage documentation and the commit
> > message that introduced the feature -- it's purely for use
> > in /etc/portage/, which is beyond the scope of PMS.
> >
> If it's pre-EAPI it's part of EAPI '0'. That you neglected to
> document it, for whatever reason, is irrelevant.

No, it's not part of EAPI 0. It's an accident. If you'd like another
example of an accident, Portage won't complain if you stick garbage in
certain metadata keys; this does not mean PMS should say that it's
legal to put garbage in metadata keys.

> > It is not the business of PMS to enforce undocumented features
> It's not undocumented, as you just pointed out above.

Using it in the tree is undocumented. Using it in user configuration is
beyond the scope of PMS.

> > that Portage supports only by accident
> Oh, so now you know the minds of the portage developers?

Yes. I know that they said when adding the directory feature that it
was for use in user configuration files. That's what the commit message
said, and that's what the documentation patch said. The documentation
change explicitly only allowed the feature in user configuration, not
the tree.

Had the feature intended to be tree-usable, the documentation change
would have said so.

> I'd like to present an alternative viewpoint: portage developers are
> happy to work to PMS, since it has utility for users. But ultimately,
> they're not that bothered about pushing for new things, since the
> process means dealing with you; adding features for portage only and
> leaving it up to the wider community to push for them in EAPIs is an
> awful lot less hassle.

Even a casual look at EAPI 3 will show that that is nonsense. But then,
you already know that from the previous times that claim has been made
and refuted.

> > and that aren't used in the tree. 
>
> Circular argument, don't you think? It's not in-tree so we won't put
> it in PMS. It's not in PMS, so it's not allowed in-tree.

That's only a circular argument if you snip out the rest of the
sentence.

> I'd like to ask the Council to consider whether EAPI development has
> not in fact supplanted a large part of the GLEP process (specifically
> the technical aspects to do with ebuild development.) As such,
> insisting on all discussion on bugzilla is in fact a subversion of
> the original process that people have agreed upon.

Moving the discussion to bugzilla was the Council's decision, not mine.

-- 
Ciaran McCreesh

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]