[ZODB-Dev] ZEO cache bug? (again)

Tim Peters tim.peters at gmail.com
Thu Jun 1 19:21:14 EDT 2006


[Chris Withers]
>>   File "C:\Zope\2.9.2\lib\python\ZEO\cache.py", line 151, in setLastTid
>>     self.fc.settid(tid)
>>   File "C:\Zope\2.9.2\lib\python\ZEO\cache.py", line 1060, in settid
>>     raise ValueError("new last tid (%s) must be greater than "
>> ValueError: new last tid (244828509247418982) must be greater than
>> previous one (244828509247418982)

> I know we've been round this circle before, but can anyone (Tim?) see
> downsides of changing this test to be >= instead of >?

I have nothing new to say about this.  As noted multiple times before,
assorted code and internal comments contradict each other about the
intent here.  Truly resolving that still requires someone with
knowledge of cache verification invariants to determine the true
intent, and fix the code+comments accordingly.  Short of that, you're
still reduced to poke-and-hope.

> If not, what branches do I make this change on

Presumably all at and after ZODB 3.4 (i.e., post-MVCC branches; 3.3 is dead).

> and how do I run the tests?

test.py -vv --all

See ZODB's README.txt.  It's possible that some ZODB setup.py files
have changed again in ways that put them out of synch with their
corresponding README.txt files again.


More information about the ZODB-Dev mailing list