[TIP] Everybody wants a pony!

C. Titus Brown ctb at msu.edu
Tue Mar 31 21:07:00 PDT 2009

On Tue, Mar 31, 2009 at 11:06:48AM -0600, m h wrote:
-> On Tue, Mar 31, 2009 at 10:57 AM, Kumar McMillan
-> <kumar.mcmillan at gmail.com> wrote:
-> > On Tue, Mar 31, 2009 at 11:37 AM, C. Titus Brown <ctb at msu.edu> wrote:
-> >>
-> >> To use Hudson, you need the JVM. ?Now you have two problems...
-> >>
-> >> (Think Snakebite, Kumar.)

[ ... ]

-> > In fact, you may find that Hudson requires no hacking at all for
-> > snakebite. ?From everything you posted in the feature-want list it
-> > would not require hacking, anyway. ?Those items are all supported
-> > afaik. ?But of course ymmv. ?It seems worth experimenting with.
-> I think Titus was referring to non-helpful platforms such as
-> hpux/aix/cray/weird mainframes that are now stacked up in his office,
-> that don't have such nice package management.

Yep.  (They're actually nicely unpacked now, just not all functioning ;)

-> I would imagine that getting python on said packages might be easier than java?

Yes.  Python has been rathe carefully crafted to be that way, AFAICT.

-> My concerns for snakebite would be:
-> * Who really cares about those platforms?  From what Guido said, he
-> doesn't.  I don't think students do.

Scientists care.  Snakebite is for Python packages as well as core; I
suspect the SciPy people will be overjoyed to be able to test.  I should
also say that a few core developers are enthusiastic about finally
having access to weird/esoteric types of dev boxes (bigmem, non-intel,
etc.)  So I think it's a 5%-er kind of issue, but nonetheless

-> * security.  You really want virtualization to solve that problem.
-> Unless you want to throw students at it.  (poor students....)

Virtualization is not available on every platform, but we'll certainly
be looking at it.

Another big concern with relying on Java is maintenance.  Spreading
ourselves thinner maintaining that in addition to everything else sounds
ugly to me (although admittedly rewriting something that already exists
is stupid).

We will have to look at Hudson.

C. Titus Brown, ctb at msu.edu

More information about the testing-in-python mailing list