[TIP] including tests in packages, or not

[C. Titus Brown]

On Sun, Sep 13, 2009 at 07:48:16PM -0700, Fernando Perez wrote:
-> On Sun, Sep 13, 2009 at 7:35 PM, C. Titus Brown <ctb at msu.edu> wrote:
-> > I plan to get it working with
-> >
-> > ? ? ? ?python -m tagnabbit.tests
-> >
-> > which seems nice and simply (especially if I can get it reporting back
-> > to my CI server, heh).
-> 
-> As a suggestion, if you set that up do make it ask for user
-> confirmation before sending anything.  People get (understandably)
-> antsy with 'phone home' codes, even if they are open source.  But
-> otherwise it's a great idea.

yeah, yeah ;)

-> Now that you mention it, I think it would be great to have in the test
-> machinery an optional function to send a nicely pickled,
-> information-rich test report back to the server:
-> 
-> import foo;foo.send_test_report()
-> 
-> With a server listening on the other end that could automatically load
->  and display on a webpage entries for test reports as they arrive,
-> this would be a great way to get 'human buildbots'.  Basically I think
-> a buildbot machinery is great to have, but this complementary to fully
-> automatic, always-on buildbots.  It's just a way of getting a user to
-> have a no-brainer, one-line way of contributing back test information.

Yep.

I wish I was smart enough to come up with this stuff myself, but
cmake/ctest/dart gave me the idea.

-> Given enough cpu/system data, this could also allow for the
-> aggregation of some performance data in the long run (noisy, but
-> useful...) that could help spot regressions, etc.

Yep.

--t
-- 
C. Titus Brown, ctb at msu.edu