[ZODB-Dev]Re:SVN: ZODB/branches/anguenot-ordering-beforecommitsubscribers/src/transaction/Implementsthe ordering of before commit subscribers :

[Tim Peters]

[Julien Anguenot]
> ok I changed it as you mentionned above + the use of a global index on
> the transaction for the bisect. Like this the complexity is removed.

Thank you again, Julien!  It's "obviously right" now, and that's always a
nice place to declare victory <wink>.

> For the deprecations, I checked a little bit and I saw you're having lots
> of custom warnings utils so I'll let you cope with this one ;)

Right, I'll take it from here.  It's not just arranging to do the
deprecation, it's also fooling the test suite into verifying that
DeprecationWarning gets raised, but without spraying warnings to the
console.  Some dark magic is involved, and I don't want to discourage you so
soon from working on ZODB again ;-)

I intend to get your branch merged into the trunk by the end of tomorrow.  I
hoped to do it today, but Fred's working on another long-standing irritation
today:  the code in ZODB 3.4/3.5's setup.py has nothing to do with the code
that creates a ZODB release anymore, so those two ways of building+testing
ZODB 3.4/3.5 have nothing in common anymore, and they do get out of synch.
Fred is adding more intelligence to zpkg, primarily for Zope3, and we've
been using ZODB trunk today as a simpler warmup exercise.  (This shouldn't
make it any harder to merge your branch in the end, it's just making it
harder to find quality time to work on it today.)