On Fri, Aug 8, 2008 at 6:28 PM, Phillip J. Eby <[EMAIL PROTECTED]>wrote:
> ... >> >> >> >> https://code.launchpad.net/~setuptools/setuptools-test/main<https://code.launchpad.net/%7Esetuptools/setuptools-test/main> >> > > My initial reaction is that it's off to a good start, but the tests > themselves seem rather shallow; more like "smoke tests" (i.e., turn it on > and see if smoke comes out) than functional tests. Right, they are not detailed at all at this point > > > I'm thinking it might help to use the setuptools.sandbox facility to log > files created, deleted, modified, etc. by the process. That would probably > be a better test of what has/hasn't been done than using ellipses on the > logs, which is order-dependent as well as having the ability to skip lines > where the wrong thing is being done, etc. The way things are being done > now, they probably won't be able to test some of the things that are most > likely to break (i.e., the complexities of easy_install). > > (Probably in order to do that I'll need to add a new sandboxing class that > creates a "mock" filesystem and allows before/after expectations to be set.) Here's a proposal that could be done in a new 'test' sprint I guess : - add something in the sandbox to record what is being done underneath (using for example what Ian has suggested - ScriptTest) - change our doctests so they actually use the recorded info : files created, removed, modified + return code maybe (but not the stdout which vary too much from one system to another) At least Chris Galvan and I are interested in helping in this. Tarek
_______________________________________________ Distutils-SIG maillist - [email protected] http://mail.python.org/mailman/listinfo/distutils-sig
