Python 2.6 transition update

[Kumar Appaiah-2]

Dear Debian Python,

As you may be aware, I have been making quite a bit of hue and cry
about the Python 2.6 transition blocking bugs[1]. I have managed to
submit patches for most of the packages to _make them build_ with
Python 2.6. However, as Paul Wise righly points out[2], my patches
merely make the packages build; I haven't really tested any of these
packages with Python 2.6 (In fact, I observe that fonttools is
dependent on python-numpy, which is not yet built with python 2.6, so
I shouldn't have sent that patch at all!). So, in my opinion, it would
be best if the maintainer, or someone who can invest time into
verifying that these packages do work with Python 2.6, performs
uploads to fix the blocking bugs.

Thanks!

Kumar

[1]: http://bugs.debian.org/cgi-bin/pkgreport.cgi?users=debian-python@...;tag=python2.6
[2]: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=547821#20
--
Kumar Appaiah


--
To UNSUBSCRIBE, email to debian-python-REQUEST@...
with a subject of "unsubscribe". Trouble? Contact listmaster@...

[Yaroslav Halchenko]

Wow... no replies in this imho important thread...

Hi Kumar,

Thanks for starting the roll.  Imho numpy/scipy and then any other
python- with binary modules and not absent rdepends should be indeed
rebuilt with 2.6 and installed in the (ch)root to verify if any other
2.6 package dependent on them not only builds but at least imports.

Module name should be easy to figure out in many cases - for most of
them since due to policy's "should" iirc rule module name should be
actual suffix of the "python-" package name.  If it does imports fine I
would look for module.test and if present -- run it.  That would imho be
very valuable information for package maintainers... it might even be
worth to setup some temporary repository with builds of the packages
with binary extensions (e.g numpy) for 2.6 so that anyone willing to
troubleshoot his dependent package for 2.6 compatibility could use
prebuilt packages from there instead of manual redoing what you
are doing ;-)

Cheers,
Yarik

On Wed, 14 Oct 2009, Kumar Appaiah wrote:

> Dear Debian Python,

> As you may be aware, I have been making quite a bit of hue and cry
> about the Python 2.6 transition blocking bugs[1]. I have managed to
> submit patches for most of the packages to _make them build_ with
> Python 2.6. However, as Paul Wise righly points out[2], my patches
> merely make the packages build; I haven't really tested any of these
> packages with Python 2.6 (In fact, I observe that fonttools is
> dependent on python-numpy, which is not yet built with python 2.6, so
> I shouldn't have sent that patch at all!). So, in my opinion, it would
> be best if the maintainer, or someone who can invest time into
> verifying that these packages do work with Python 2.6, performs
> uploads to fix the blocking bugs.

> Thanks!

> Kumar

> [1]: http://bugs.debian.org/cgi-bin/pkgreport.cgi?users=debian-python@...;tag=python2.6
> [2]: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=547821#20
--
                                  .-.
=------------------------------   /v\  ----------------------------=
Keep in touch                    // \\     (yoh@|www.)onerussian.com
Yaroslav Halchenko              /(   )\               ICQ#: 60653192
                   Linux User    ^^-^^    [175555]



--
To UNSUBSCRIBE, email to debian-python-REQUEST@...
with a subject of "unsubscribe". Trouble? Contact listmaster@...

[Kumar Appaiah-2]

Dear Yaroslav,

On Tue, Oct 20, 2009 at 10:01:05PM -0400, Yaroslav Halchenko wrote:
> Wow... no replies in this imho important thread...

Yes, even I was a tad surprised. Thanks for the message.

> Thanks for starting the roll.  Imho numpy/scipy and then any other
> python- with binary modules and not absent rdepends should be indeed
> rebuilt with 2.6 and installed in the (ch)root to verify if any other
> 2.6 package dependent on them not only builds but at least imports.

I did that. I rebuilt these packages with python 2.6, and used them in
my chroot for the test. numpy.test() works fine using Python 2.6.

> Module name should be easy to figure out in many cases - for most of
> them since due to policy's "should" iirc rule module name should be
> actual suffix of the "python-" package name.  If it does imports fine I
> would look for module.test and if present -- run it.  That would imho be
> very valuable information for package maintainers... it might even be
> worth to setup some temporary repository with builds of the packages
> with binary extensions (e.g numpy) for 2.6 so that anyone willing to
> troubleshoot his dependent package for 2.6 compatibility could use
> prebuilt packages from there instead of manual redoing what you
> are doing ;-)

If needed, I can offer builds of python-numpy (or whatever package)
with python 2.6. Unfortunately, I am using a custom python-central as
well in my chroot, since python-numpy and python-imaging seem to be
affected by bug #547565; some other workarounds might work, I am not
sure.

In any case, there are only 9 packages remaining with patches, and 4
without patches (of which one has been uploaded and actually supports
python 2.6), and one pending upload. I think it's time we uploaded
python 2.6 to main, and binNMU the important packages mentioned above
and proceed to fix the remaining issues.

So, Debian Python, could you please upload python2.6 to unstable?

Thanks!

Kumar
--
Kumar Appaiah


--
To UNSUBSCRIBE, email to debian-python-REQUEST@...
with a subject of "unsubscribe". Trouble? Contact listmaster@...

[Kumar Appaiah-2]

On Tue, Oct 20, 2009 at 09:31:30PM -0500, Kumar Appaiah wrote:
> On Tue, Oct 20, 2009 at 10:01:05PM -0400, Yaroslav Halchenko wrote:
> > Thanks for starting the roll.  Imho numpy/scipy and then any other
> > python- with binary modules and not absent rdepends should be indeed
> > rebuilt with 2.6 and installed in the (ch)root to verify if any other
> > 2.6 package dependent on them not only builds but at least imports.
>
> I did that. I rebuilt these packages with python 2.6, and used them in
> my chroot for the test. numpy.test() works fine using Python 2.6.

Oh, and just to be sure, the rdepends also were able to import numpy
and use it... just in case I wasn't clear.

Kumar
--
Kumar Appaiah


--
To UNSUBSCRIBE, email to debian-python-REQUEST@...
with a subject of "unsubscribe". Trouble? Contact listmaster@...