On Tue, Aug 24, 2010 at 10:31 AM, Adam GROSZER <agros...@gmail.com> wrote: > Hello Jim, > > Monday, August 23, 2010, 8:48:16 PM, you wrote: > > JF> On Mon, Aug 23, 2010 at 12:46 PM, Adam GROSZER <agros...@gmail.com> wrote: >>> Hello, >>> >>> I have here this version spec in setup.py: >>> 'zope.testing >=3.6.0,<3.10.0', >>> >>> but zc.buildout does not get it right. >>> Source is here: >>> svn://svn.zope.org/repos/main/zc.recipe.testrunner/branches/1.2.1 >>> Output is here: >>> http://paste.lisp.org/+2FSQ >>> >>> Why on earth does zc.buildout pick zope.testing = 3.10.0 with the >>> above spec?? > > JF> If you look earlier in the output you provide a link to, buildout does > JF> get it right. It picks 3.9.5. > > JF> Then, when zc.recipe.testrunner runs, it tries to load zope.testing > JF> explicitly: > > JF> eggs, ws = self.egg.working_set(('zope.testing', )) > > JF> It doesn't restrict the version here. This didn't cause a problem > JF> until you added the restriction in in setup.py. > > JF> Anyone not interested in the implementation of zc.recipe.testrunner should > JF> stop reading here, if you haven't already. :) > > JF> self.egg implies zc.recipe.testrunner, so the low-level buildout APIs > JF> are asked to create a working set for zc.recipe.testrunner and > JF> zope.testing. It happens to load packages in a breadth first order > JF> and gets the newest version of zope.testing before it tries to get the > JF> restricted dependency of zc.recipe.testrunner. > > JF> The easiest way to fix this is to remove the redundant and now > JF> incorrect reference to zope.testing in the call to > JF> self.egg.working_set: > > JF> eggs, ws = self.egg.working_set() > > Thanks for the pointer, but in my case it seems to work only with the > version spec added: > > eggs, ws = self.egg.working_set(('zope.testing <3.10.0', )) > > If I do > > eggs, ws = self.egg.working_set() > > zope.testing seems to be missed from the test script entirely.
Ah, that makes sense when you're testing something other than zc.recipe.testrunner itself. Jim -- Jim Fulton _______________________________________________ Distutils-SIG maillist - Distutils-SIG@python.org http://mail.python.org/mailman/listinfo/distutils-sig
