[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#573745: Please decide on Python interpreter packages maintainership

[Andreas Barth 2010-03-18]
> I agree. I tried to get one such meeting done some time ago (see
> below)

Here's my reply from that thread (Matthias or DPL never
responsed, or at least I wasn't in the loop)

[Piotr Ożarowski, 2009-09-27]
> Adding Raphael Hertzog to Cc as he knows the beginning of py{central,support}
> saga much better that I do and IMO always tries to find a consensus
> instead of arguing till death :-)
> > as part of the recent release mails, we got some - well, "complaints" is
> > the wrong word, but definitly in that direction - mails about python2.6 (or
> > rather its inexistence in unstable).
> Debian's Python situation is not perfect indeed. I don't know if it was
> better in the past - when I joined Debian few years ago it was already
> one big minefield[1] which doesn't help maintaining Python packages /
> recruiting new developers.
> [1] outdated Debian Python Policy (no consensus about what it should
>     contain), unwritten rules, 3 different (mostly undocumented) helper
>     tools, python maintainer not really active (yes, even then, see
>     reasons for founding DPMT), cdbs not supporting these new tools at
>     all (dh didn't exist at that time), maintainers forced to do lots of
>     work in debian/rules in order to satisfy not existing policy, ...
> Let me shortly describe the situation from my perspective (i.e. from a
> person who don't know much about what happened in Mexico).
> I started with dh_pysupport (which was the only tool available in
> unstable, except dh_python) but switched most of my packages (not all)
> to dh_pycentral really soon after that as I thought it was better
> designed - it supported Python extensions and didn't provide additional
> problems with __file__ and namespace issues (i.e. it stored .pyc files
> where everybody expected them to be). After a while, dh_pysupport
> started to be more and more mature (Python extensions support, bugs
> fixed really fast, additional features like Egg renaming, namespace
> (re)generation[2], etc.) and pycentral didn't change much and when it
> did change, it broke lots of packages by some internal changes never
> announced to anybody.
> [2] IIRC, this was the only change not really well announced/tested and
>     with wrong (see #459468) defaults in pysupport
> So I took advantage of the fact that nobody really sponsored Python
> packages[3] and that many DDs (who were my sponsorees in the past or
> whom I helped with some Python related problems) trusted me and decided
> to get rid of one of the tools: python-central. And it worked fine, lots
> of packages were converted in the last few months.
> [3] see [1] for possible reasons
> Then, after I talked with Matthias (on my first DebConf) and realized
> that he will not upload python2.6 until we'll fix the helpers issue once
> and for all - I tried to pick up one of his ideas - symlinks (it's
> almost perfect solution - the only real problem is the fact that
> arch:all packages needs to be rebuilt once list of supported Python
> version changes), try to adjust it a little bit (f.e. remove his pet -
> "current" keyword), and if it will be accepted - update official Debian
> Python Policy...
> That didn't work that well. Surprisingly many developers didn't like the
> idea and didn't even want to discuss it[4] and Joss is obviously not
> interested in changing status quo as status quo means removing
> python-central at some point (so far only Scott didn't want to convert
> his packages to python-support, it was very easy to convince other
> maintainers - the longest conversation I had was the one with Martin
> Krafft - and he just wanted to know why I want to convert his packages,
> didn't object much about dropping pycentral).
> [4] I thought we could find a workaround for arch:all binNMUs issue, maybe
>     some kind of semi-automatic rebuilds that do not suffer from binNMUs
>     problems...
> Here's what we plan to do in the near future and what can eventually be
> discussed in such meeting (I'm not really sure that such meeting will
> change much if both Joss and Matthias will not tell us they're willing
> to hear others arguments and make a compromise).
> We are fixing current tools and packages to support changes introduced
> in python2.6 which is currently in experimental:
> * http://bugs.debian.org/cgi-bin/pkgreport.cgi?users=debian-python@lists.debian.org;tag=python2.6
> * cdbs will go to unstable today (I uploaded NMU to DELAYED/7 a week ago)
> * I will attach a fix to python-central today or tomorrow
> and then we will add 2.6 to supported versions in NMUs (I will upload
> it, I'm so sick of the current situation that I will rather let Matthias
> remove my key from Debian keyring than release Squeeze without 2.6 or
> force us to do the transition 1 month before the release, like in
> Ubuntu). 
> If my NMU will be rejected, I plan to hijack python and try to maintain
> it in a team of few DDs (don't worry, I'll ask before the hijack, not
> after - if nobody will be interested, I will not maintain it alone -
> that would make it only worse) - I really don't want this to happen as
> Matthias is probably the one who knows more than anyone else about
> maintaining Python. I already took a quick look at python bugs to see if
> I'm able to fix them. I saw many easy to fix bugs (time consuming, but
> easy to fix) or even bugs that are not bugs at all (Python developers
> would call them "features" or user broke his system with local
> modifications). So...
> Matthias *needs* co-maintainers! Looks like he tries to respond to
> serious problems (f.e. he dealt with recent libdb issue) but doesn't
> have time at all to process other, less urgent ones.
> I will start proposing NMUs to these bugs (patches are not enough, see
> BTS) and if Matthias will like them, at some point I will propose to
> accept me as co-maintainer, if not, he'll have proves that he needs one
> and I'll try to force him to choose another one.
> Anyway, IMHO what can be discussed in such meeting is what can we do
> after the Python 2.6 transition:
> * try to fix my dh_python proposal and implement it, or
> * remove python-central completely and let python-support use
>   /usr/lib/python*/*-packages (i.e. official places, without the need to
>   use .pth files) and keep using /usr/{share,lib}/pyshared to ship
>   .py/.so files, or:
> * 3rd solution, perfect one this time (its author is waiting for a
>   perfect timing to hit us with it when we'll not expect it ;)
Piotr Ożarowski                         Debian GNU/Linux Developer
www.ozarowski.pl          www.griffith.cc           www.debian.org
GPG Fingerprint: 1D2F A898 58DA AF62 1786 2DF7 AEF6 F1A2 A745 7645

Attachment: signature.asc
Description: Digital signature

Reply to: