On Feb 22, 2013, at 11:51 PM, Stefano Rivera wrote: >1. Status quo: Provide a nosetests-3.X script for the default version at > build time. > Pros: None > Cons: > - This potentially breaks unit tests if there are two supported 3.x > versions. >2. Drop all nosetsts-3.X scripts. > Pros: > - Maintainers who were aware of the problems with 1 had to manually > call python3.X /usr/bin/nostests3 anyway, so this doesn't cause > them any harm. > - Don't accidentally end up with dependencies on all python3.Xs > Cons: > - Maintainers who weren't needed their packages patched. >3. Apply a messy patch to generate scripts based on py3verions -s at > build time. > Pros: > - Neat > Cons: > - It's ugly as hell > - Have to do a sourceful upload for each python3 supported versions > change > - Will accidentally end up with dependencies on all python3.Xs >4. Use .rtinstall, .rtremove, postinst, and prerm scripts to maintain > all the nosetsets-3.X scripts (pytest does this) > Pros: > - Neat > - No accidental dependencies on all python3.Xs > Cons: > - You are creating and deleting things in /usr/bin in maintainer > scripts - this made some people cringe.
I wish we would do #4. I suppose it's a little cringe worthy, especially because (as you later point out) you'd probably also want to add versions for the -dbg flavors too. But that bothers me less than not having those scripts available, since I think users will expect them to be there. For example, the tox documentation example suggests calling nosetests (albeit, for Python 2) directly. http://tox.readthedocs.org/en/latest/example/nose.html?highlight=nose Is #4 really that horrible? -Barry
signature.asc
Description: PGP signature
