Re: Upload of python-networkx 1.9
and where are the SVN commits for such changes? have you talked with
upstream about the huge tarball?
On Wed, Jun 25, 2014 at 9:41 AM, Thomas Goirand <zigo@debian.org> wrote:
> Hi,
>
> As per #750233, I have uploaded python-networkx 1.9 in the DELAYED/2
> queue. Considering the currently opened RC bug (ie: #750233), I couldn't
> wait for too long, otherwise risking AUTORM of some of my packages. The
> changelog is as follow:
>
> * Team upload.
> * New upstream release (Closes: #750233).
> * Ran wrap-and-sort.
> * debian/copyright in parseable format 1.0.
> * Added extend-diff-ignore = "^[^/]*[.]egg-info/" in d/source/options,
> and removed the clean of the egg-info folder.
> * Refreshed/rebased 10_doc_relocation, 20_example_dirs_remove and
> 30_use_local_objects.inv patches.
> * Removed 50_force_ordering_in_dict_to_numpy_array_functions.patch
> 55_fixups_to_numpy_array_functions_patch.patch
> 60_fix_tests_failure_due_to_py3.3_hash_randomization.patch applied upstream.
> * debian/control: removed python-support and added dh-python as build-dep.
> * Added python-3.4.patch (comes from the Ubuntu package).
> * Removed 40_no_setuptools_in_requires.txt (there's no setup_egg.py
> anymore in upstream source code).
> * Also cleans networkx/version.py, which is generated.
> * Also cleans test.*, doc/source/templates/gallery.html and fontList.cache.
> * Added do-not-use-sphinx_rtd_theme.patch.
> * debian/rules: do not use setup_egg.py, it's gone from upstream source.
> * Move all the doc generation stuff into the override_dh_sphinxdoc.
> * Removed obsolete XS-Python-Version: >= 2.6 X-Python3-Version: >= 3.2.
> * Adds using-local-mathjax.js.patch to avoid privacy breach in the doc.
> * Added missing --with python2.
> * Added debian/rules get-orig-src target to generate the orig.tar.xz file.
> * Added missing build-depends: python-decorator and python-numpydoc
> (needed for building docs).
> * Fixed watch file to handle the +dfsg1 mangling.
>
> I first tried to make the changes minimal, but it simply wasn't possible
> in this case. There was just too many things to do to have a working
> package. I tested building in a cowbuilder and it went fine. I also
> tested rebuilding some of the reverse dependencies that I knew were unit
> testing with networkx, and it went fine too.
>
> Note that since the pypi tarball was huge for no reason (it contained
> generated files from the documentation build process, leading to 35 MB
> of stuff instead of less than 1), so I used the git tag from github
> instead, to generate the orig.tar.xz. In the process I also cleaned some
> non-free files (minimized javascript). The process is now documented in
> debian/rules. [It would probably be best to have it done in a
> debian/repack file, using the watch file though, but I have no time for
> that yet.]
>
> If anyone opposes to this upload, it can be reverted/dcut. The diff of
> the debian folder is attached (missing the watch file fixes for the
> +dfsg1 handling).
>
> Cheers,
>
> Thomas Goirand (zigo)
--
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi
Reply to: