On Tue, Mar 11, 2014 at 4:36 AM, Nick Coghlan <ncoghlan@gmail.com> wrote:


On 11 Mar 2014 09:10, "Antoine Pitrou" <antoine@python.org> wrote:
>
> On lun., 2014-03-10 at 16:02 -0700, Alex Gaynor wrote:
> > Hi all,
> >
> >
> > I'd like to propose Brian Kearns for commit. He's been a committer on
> > PyPy for about a year and a half now, and in particular he's done a
> > bunch of "Python version" works: things like upgrading us from the
> > 2.7.3 stdlib to the 2.7.6 stdlib, and py3k work. He's interested in
> > having commit for the purposes of doing interop work on the stdlib
> > tests: things like making sure tests aren't reliant on refcounting,
> > correctly marking tests as impl details, etc.
>
> I'd really prefer someone to have experience in contributing to CPython
> before they get commit rights. I might mistaken, but I can't find any
> contribution bearing Brain's name.
>
> Furthermore, giving arbitrary commit rights to core devs of third-party
> projects (such as PyPy and Twisted) doesn't seem to have produced any
> significant CPython contributions from them, IIRC.

Aye, our processes are currently arcane enough that posting a patch to the tracker is *less* work than doing the commit yourself (on the other hand, it depends on another human to get it committed).

On the other hand, if Brian has read PEP 462 and *still* wants to commit patches directly, then I wouldn't be opposed (given Alex's recommendation). Call it +0.


There's also precedent as Antoine alluded to; we previously gave two people on each major interpreter implementation commit rights to work on compatibility issues (Alex and Maciej represent PyPy directly, although obviously people like Armin had commit rights predating PyPy). But as Antoine also pointed out, the experiment never really went anywhere.

I still think it's a good idea, though, to get the other interpreters involved with helping with compatibility issues. I would be fine if Brian started out restricted to Lib/test to fix tests that weren't properly marked as implementation details, and if that went well then allowed to branch out to actually fixing compatibility bugs. If people are uncomfortable with even that restriction of abilities then we can do this the old fashioned way of Brian submitting patches through bugs.python.org before requesting commit privileges again.