M.-A. Lemburg writes:
Wouldn't it be better to explicitly name the test script using a keyword parameter to setup() ?
I think that would be a good elaboration, but something simple like this would be useful for the typical case.
- test (if available) should be run prior to install
Agreed; I didn't want to mess with the other commands until we determine whether it should go in.
- installing should not be possible with failing tests (errors) unless an override option is used
- the test command should have an option to turn warnings into errors
Are they treated differently? What's missing is that I don't know offhand how to raise the error condition for distutils, otherwise I would.