On Thursday 27 September 2007 09:35, Jim Fulton wrote:
> On Sep 26, 2007, at 7:57 PM, Roger Ineichen wrote:
> > Hi
> > Can anybody tell me why we restrict our test setup
> > in zope eggs and only use a subset of package for
> > our test setup?
> It isn't practical, during development, to test all of the eggs that
> might be affected by a change, which is, BTW a *much* larger set than
> the old Zope 3 tree.
I am not saying that testing *all* packages is necessary, but a healthy set of
them. We did not consider this impractical when developing on the entire
trunk. Back then, it was part of our development responsibility. We always
ensured that some set of Zope 3 was always stable together. This is
completely gone now. I want a solution; buildbot is okay, but I would prefer
something closer to the time of the checkin/development.
> I personally don't see much value in that, especially considering the
> effort involved.
I really can't believe you are saying this. I think Zope 3 would not be as
stable as it is today, if we would not have always tested the trunk when we
> Depends on what "this" is. I think running all of the tests in the
> old Zope 3 tree whenever we change a component is a waste of time.
I beg to differ. It has something to do with responsibility to the community.
I am not saying you have to test all of Zope 3 when changing zc.* for
example. But I think if a package in a defined "core" set is changed,
requiring to run all the tests of the "core" set should be required. A really
good example is ``zope.component``. Not testing it across several packages is
very dangerous for our stability.
CBU Physics & Chemistry (B.S.) / Tufts Physics (Ph.D. student)
Web2k - Web Software Design, Development and Training
Zope3-dev mailing list