[Python-Dev] latest bsddb3 test problems

[Jeremy Hylton]

On Tue, 2003-09-23 at 05:23, Anthony Baxter wrote:
> On a friend's debian testing machine, the bsddb test suite crashes with 
> a deadlock on _some_ occasions. On a Mac OS X machine, the test suite
> hangs :-(
> 
> On the Mac OS X machine, the bsddb3 stuff doesn't get built under 2.3.
> 
> On the other hand, code using bsddb seems to work fine, so it's almost
> certainly just that the test suite needs to be fixed to not be broken.

I think the question is whether the tests are doing something completely
artificial that fails, or if they are simply provoking failures that
would be particularly difficult to provoke in real life applications.

In Zope's BDBStorage, we've definitely seen some weird behavior.  For
example, we had a bug in our code where we tried to delete an entry from
a BTree that did not exist.  That operation failed with a
DBNotFoundError.  We caught that error can called txn.abort(); the abort
failed with a DBRunRecoveryError.  We didn't have any idea what the
abort would fail in that case.

It may be that we misunderstood how to use the Berkeley storage, but it
leaves me with the uneasy feeling that something isn't right.  The
failing tests make me feel the same way.

Jeremy