Freeze exception for matplotlib 0.98.3-1 / -2
Hello again Luk & Release Team! :)
Petr just reported a bug[1] for a FTBFS on GNU/kFreeBSD; even if it's
not a release candidate arch[2], given the small change it requires, a
-2 revision will be uploaded this evening. I suppose that if you would
kindly approve 0.98.3-1 there would be no problem for 0.98.3-2 :)
Thanks for you patience,
Sandro
[1] http://bugs.debian.org/493858
[2] http://release.debian.org/lenny/arch_qualify.html
PS: reporting the previous email below due to subject change
> On Sat, Aug 2, 2008 at 16:52, Luk Claes <luk@debian.org> wrote:
>> Sandro Tosi wrote:
>>> Currently the archive has 0.98.1 mpl version, and we're working
>>> closely with upstream to let a new upstream release into Debian. The
>>> real goal for the new release is bugfixing and mainly documentation.
>>>
>>> We missed 0.98.2 version because of some glitches in doc generation,
>>> and we have reached yesterday the consensus on releasing 0.98.3 (they
>>> wait for my confirmation the debian package is working fine).
>>>
>>> I got the package almost ready in svn, and the great improvement is we
>>> are now able to generate matplotlib documentation at build time, and
>>> ship it in -doc package along with the binary one. It's a lot of doc,
>>> and given the possibility to users to have it on their system is
>>> really a plus (given the lots of API mpl has), and I think lenny users
>>> deserve this.
>>>
>>> In order to build such doc we need Sphinx 0.4.2 (due to changes in
>>> last version of sphinx, used in mpl docbase), and it's really
>>> impossible to backport the changes from 0.98.3 to 0.98.1 (the one in
>>> the archive).
>>>
>>> I'd like to know if I'm allowed to upload 0.98.3 version in the
>>> archive (when the sphinx situation will be fixed). Please note that
>>> upstreams are really responsive and they have waited for Debian to be
>>> comfortable with a new release before actually announce it in the last
>>> months.
>>
>> Uploading to unstable doesn't really harm and it gives us a better idea
>> about how invasive the changes are you are talking about...
>
> As suggested, matplotlib 0.98.3 has just been uploaded in unstable
> (thanks to Piotr!).
>
> I'd just like to recap what this release represent: since about 1
> month and a half, I've worked closely with upstream to be able to
> build and ship mpl documentation along with binary package for Debian
> (you can check mailing list threads[1][2][3][4][5] we exchanged with
> upstream) and we are now finally able to archive this great goal (16M
> of docs lets you understand how wide and complex this package can be,
> so having doc easily available is really a plus).
>
> As usually for upstream, a point-release (0.98.X) is a bug-fixing
> release, trying to reduce enhancements concentrating to close
> regressions and bugs, and the work hard to do it, always keeping in
> mind Debian as a main objective; the main code change could be
> summarized in:
>
> * fixed memory licking (as reported by upstream author in #492735)
>
> but not to exclude the many bugfixes for visualization, corner-cases
> and error-handling.
>
> We were unable to upload 0.98.2 version in Debian, because there was
> some problems with the doc generation, so we decided with upstream to
> skip that version and concentrate on 0.98.3.
>
> Sadly, due to changes in Sphinx 0.42.2 adopted in matplotlib, the
> 0.98.3 mpl version was incompatible with sphinx 0.41.1 APIs, so
> additional delay (we needed that version in unstable and so on), so we
> missed the freeze begin.
>
> I'm attaching the debdiff from 0.98.1 to 0.98.3, generated with the command
>
> $ debdiff matplotlib_0.98.1-1.dsc matplotlib_0.98.3-1.dsc | filterdiff
> -x '*/doc/*' -x '*example*' -x '*font*' > matplotlib_0.98.3-1.debdiff
>
> I know, it's big, but there are a lot of changes in docstring & so
> (doc is good :) ).
>
> I made a first run of test on direct Depends from python-matplotlib,
> gastables model-builder nulog, taking the latest source from unstable,
> recompiling and testing with 0.98.3 and all 3 works fine, and Piotr
> tested pondus and ipython, both work fine too.
>
> After this long preparation, I'd like to kindly request to please
> unblock this new version :D As a cautelative solution, we can think
> about a longer transition time (15days?) just to avoid any hidden
> regressions.
>
> Thanks in advance for your attention,
> Sandro
>
> [1] http://sourceforge.net/search/index.php?words=subject%3A(%2B%22matplotlib+0.98.3+release+candidate%2C+please+test%22)+AND+ml_name%3A(matplotlib-devel+matplotlib-users)&sort=posted_date&sortdir=desc&offset=0&group_id=80706&type_of_search=mlists&pmode=0
> [2] http://sourceforge.net/search/index.php?words=subject%3A(%2B%22preparing+for+0.98.3+and+0.91.5%22)+AND+ml_name%3A(matplotlib-devel)&sort=posted_date&sortdir=desc&offset=0&group_id=80706&type_of_search=mlists&pmode=0
> [3] http://sourceforge.net/mailarchive/message.php?msg_id=8b2d7b4d0807020446h567a703x86e38c2d8ed6ef3e%40mail.gmail.com
> [4] http://sourceforge.net/mailarchive/message.php?msg_id=20080701014046.GA11155%40chiark.greenend.org.uk
> [5] http://sourceforge.net/search/index.php?group_id=80706&form_submit=Search&search_subject=1&type_of_search=mlists&all_words=&exact_phrase=reverting+changes+to+contour.py&some_word=&without_words=&ml_name[]=matplotlib-devel&ml_name[]=matplotlib-users&posted_date_start=&posted_date_end=
--
Sandro Tosi (aka morph, Morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi
Reply to: