[TIP] Idea for GSOC : a test infrastructure for PyPI

Michał Kwiatkowski constant.beta at gmail.com
Mon Mar 22 11:23:36 PDT 2010

On Mon, Mar 22, 2010 at 6:10 PM, Tarek Ziadé <ziade.tarek at gmail.com> wrote:
> On Mon, Mar 22, 2010 at 12:52 PM, Paul Hildebrandt <Paul.Hildebrandt at disneyanimation.com> wrote:
>> This seems similar to what Michal and Grig did in 2006.
>> http://mousebender.wordpress.com/application/
>> I wonder if there is any overlap that would be helpful.
> Indeed ! I understand now why there are some cheesecake fields in the PyPI Database.
> Thanks for this link, I am ccing them to know what's the exact status of the project,
> and what they have achieved so far.

So we had two projects with Grig, one in 2006 and another one in 2007.

The one from 2006 had two parts:
 1. extending the Cheesecake tool,
 2. building a Cheesecake service that would run the tool on new PyPI packages.

You can read about the project here:


Besides working on Cheesecake and the Cheesecake service, I also
contributed some patches to PyPI, signed off by Richard. The idea was
to not only have the Cheesecake service dashboard, but to actually
show Cheesecake scores on the PyPI itself. That's why you see
cheesecake_* tables in PyPI database. I have those patches lying
around somewhere, I can send you copies, but you should be able to
find them in the PyPI repository (look around August 2006).

The Cheesecake service went live in February 2007, here's a related
discussion on distutils-sig:


I have a checkout of the code pointing at
svn+ssh://mk@svn.idyll.org/svn/mk/cheesecake_service/trunk, but I'm
not sure this repository still works. Anyway, I can put that on
github/bitbucket if you want.

The second GSoC project was about checking patches against CPython
trunk. The system would automatically get the patch from Roundup,
apply it, build it and run the full test suite in a sandboxed
environment. Here's some more information:


We also had a Trac instance at http://pybots.org/pvs/ . I'm not sure
if Grig kept a backup, but after I migrated to bzr all docs were moved
to the repository was well. The code for the PVS is here:


That's the actual bazaar repo, so the history is preserved as well.
:-) Roundup and Subversion integration is probably not very useful,
but code dealing with VMware can certainly be reused.

I guess that sums it up. Let me know if you have more questions.


More information about the testing-in-python mailing list