On Fri, Aug 14, 2009 at 2:13 PM, Jim Fulton<jim@zope.com> wrote:
On Fri, Aug 14, 2009 at 1:58 PM, Benji York<benji@zope.com> wrote:
By "buildbot output" do you mean simply the fact that all tests passed on all platforms?
I'm being intentionally vague, but we need to know that the tests apply to the versions we're thinking of merging. At a minimum, the individual run outputs should contain the svn revision #. So, when merging changes, we make sure that the test branch changes being merged have been tested.
I hesitate to suggest this (because it sounds like more work when we don't have enough volunteer's as-is), but how about if each branch had its tests run each time there was a commit made to it. That way we don't have to worry about mixing up changes in a single "test" branch and we'd also be able to tell -- by looking at the buildbot output for a particular branch -- if a branch that was being considered for merging to the trunk is actually passing on all platforms (OS and Python versions). I imagine that functionality would also be helpful to people developing on a branch. They could know if their changes cause tests to fail on platforms they might not have access to. -- Benji York Senior Software Engineer Zope Corporation