[ZODB-Dev] RE: [Zope-Annce] ZODB 3.2.4 release candidate 1released

[Chris McDonough]

On Tue, 2004-09-14 at 21:27, Tim Peters wrote:
> [Tim Peters]
> > ...
> > However, in the SVN setup, the ZODB code isn't shared by any Zope
> > checkout (2.8 and 3 and X3 all have distinct copies of ZODB, made at
> > varying times in the past), so there won't be a way to know how this
> > affects any of the Zopes before someone explicitly stitches a new ZODB
> > into them, one at a time.  I don't think the Zope 2.8 trunk has gotten a
> > new copy of ZODB since the day Jim created the SVN repository (I haven't
> > stitched one in there, and AFAICT 2.8 still lacks a project manager).
> 
> 3.3c1 has been released now (with the new "sticky failing commit" behavior),

Excellent!

> and I stitched it into the Zope 3 and X3 trees.  The latter went very easy
> -- at least the Zope 3 test suites didn't rely on a failing commit starting
> a new transaction by magic.  Well, *some* tests did.  A few tests provoked a
> commit failure on purpose, then just stopped.  That caused commits in
> *later* tests to raise the new TransactionFailedError.  

I imagine unit tests will probably be the thing to most likely run afoul
of this policy.

> But because that
> exception includes a full stack trace from the original failing commit, it
> was trivial to find the tests that didn't clean up after themselves.  It
> would have been much harder without that gimmick, so I'm happy you pushed
> for it, Chris.

That's me, always doing the hard work while you slack off. ;-)

- C