[Python-Dev] Snake-Pit: Cataloging Python builds on different platforms

[Christopher Blunck]

Hi all-

<Disclaimer>
Apologies in advance if this is off topic for python-dev.  I spoke with
Brett Cannon today at PyCON regarding this topic and he said to build something,
and bring it up with pydev when I had something.  I have something so I'm
bringing it up.
</Disclaimer>


Regarding snake-pit . . .

Python is a great cross-platform language.  It runs on all kinds of different
operating systems and architectures.  Staying on top of new operating systems
and architectures is challenging - just ask any of the people that managed
the snake farm.  However, I believe that it's crucially important for us to
know what platforms the latest Python runs on.  I'm specifically referring to
the HEAD tag in the python cvs module on sourceforge.

While at PyCON today I looked around and was really impressed by the variety
of architectures and operating systems that attendees ran on their laptops.  I
thought that if this many people ran this wide a variety on their laptops, 
there may be even more platforms that python enthusiasts run at home.  And
maybe they'd be willing to share some CPU cycles to help the Python 
development effort.

Although low, there are barriers to entry for building Python from source.
You have to know the CVSROOT, know how to check out the code, know which
directory the code resides in, and know how to build it.

Snake-pit is a simple shell script that does all of this for you - it 
checks out the Python source code anonymously, builds it, tests it, and
produces pretty output:
[chris at titan snake-pit]$ ./build.sh
Processing branch: main
Downloading code... done.
Building: clean, configure, compile.
Testing... done.
Results:
            Failed:  1
           Skipped:  32
  Skips Unexpected:  2
            Passed:  239
  
            Failed:  test_repr


What I'd like to do is begin catalog'ing this information on a nightly basis
on different platforms.  The goal is:  at any given time, we'll know which
platforms Python has been compiled and tested on.  After the information
has been catalog'ed, we can do a variety of transformations of it - we can
post it on a webpage or mail it to a listserve (if an error occurs).


I'm writing to solicit feedback for what information would be useful to capture
from a Python build.  So far I'm planning on capturing:
  Platform Category (Linux, BSD, Solaris)
  Distribution (RedHat, FreeBSD, Solaris)
  Version (9.0, 5.2.1, 9)
  Kernel (2.4.22-1.2174, ??, ??)
  SMP?

I'm interested in any feedback (positive and negative) that anybody would be
willing to provide.  Is this fools gold, or is it something that could be
useful?  If useful, how could it be *very* useful?


I'll be at PyCON tomorrow, but if you have comments and don't want to talk in
person the mailing list will be fine.


-c

-- 
 00:00:00  up 114 days, 13:44, 25 users,  load average: 0.85, 0.54, 0.35