[Zope-PTK] Re: ZPatterns (sub-)transaction problems

Phillip J. Eby pje@telecommunity.com
Thu, 14 Sep 2000 10:40:22 -0500


At 03:56 PM 9/14/00 +0100, Steve Alexander wrote:
>
>Can we lose DynPersist altogether for this release? 

Unfortunately, no.  Zope's cPersistent class appears it may have a bug that
is roughly similar to the one we're fixing in DynPersist.  (i.e., not
calling __of__ or other binding operations on an object returned from
__getattr__)

I've written Jim to ask whether my interpretation of that code is correct,
but have not received a reply.  It may take a new Zope release before
DynPersist is able to go away.  :(


>Or are there still folks out there using ZPatterns with 2.1.6?

Support for 2.1.x was officially dropped with ZPatterns 0.4.2a1, but in
practice most of the support code is still there today.  It's just that
I've stopped testing on 2.1.x and no longer write code with 2.1.x
compatibility in mind.  The re-architecting of the transaction machinery
will *require* 2.2.x's transaction ordering semantics.