Re: zope-tests - FAILED: 4, OK: 12

2013/4/25 Marius Gedminas <[email protected]>:
> On Thu, Apr 25, 2013 at 09:49:28AM -0400, Tres Seaver wrote:
>> On 04/25/2013 04:36 AM, Marius Gedminas wrote:
>> > So latest stable ZODB3 (3.10.5) is incompatible with the two latest
>> > stable persistent releases (4.0.5 and 4.0.6).
>> >
>> > What should we do?  Release ZODB3 3.10.6 with a backport of these
>> > patches?
>>
>> ZODB 3.10 bundles its own version of persistent:  the standalone one
>> shouldn't be used at all if ZODB 3.10 is in play.
>
> Yet somehow we're getting both in a z3c.contents buildout.  Note that
> z3c.contents doesn't directly require either one.
>
> It looks like we're getting persistent via BTrees 4.0.5, and we're
> getting BTrees via zope.annotation 4.2.0, and we're getting
> zope.annotation via zope.copypastemove 3.8.0, which is a direct
> requirement of z3c.contents.
>
> Are BTrees bundled with ZODB3 as well?  (Yes.)
>
> Can setuptools dependencies express complicated Debian-package-like
> concepts like "ZODB3 provides BTrees, persistent" and "ZODB3 conflicts
> with BTrees, persistent"?
>
> Are there other ways of avoiding a flag day?  I don't think we can
> reasonably require that no final-version PyPI package may depend on
> BTrees or persistent until ZODB 4.0.0 final is finally out.

I have removed some deprecated ftest dependencies like
zope.app.securitypolicy that pulls in the whole Zope 3. Tests passed.

Then I pinned some versions until nothing required ZODB3 any more.

Let's see how winbot likes it.

--
Gediminas
_______________________________________________
Zope-Dev maillist  -  [email protected]
https://mail.zope.org/mailman/listinfo/zope-dev
**  No cross posts or HTML encoding!  **
(Related lists -
 https://mail.zope.org/mailman/listinfo/zope-announce
 https://mail.zope.org/mailman/listinfo/zope )