dhis2-devs team mailing list archive

[Bob Jolliffe <bobjolliffe@xxxxxxxxx>]

2009/4/6 Bob Jolliffe <bobjolliffe@xxxxxxxxx>:
>> 1) Modify the getNativeIdentifierGeneratorClass() method in the DerbyDialect
>> to return IdentityGenerator.class. But this might break lots of things in
>> Hibernate.
>
> Which means I think that this should be ok.  Don't know why nobody has
> done it yet though.  I can have a go at patching it and see if the
> hibernate test suite all runs ok.  If nothing else I'll learn a bit
> more about hibernate :-)
>

OK this is interesting.  After downloading and compiling hibernate
(I'm warming slowly to maven ..) I found that hibernate is actually
quite a bit happier with the patched Dialect using
IdentityGenerator.class than with the original.

Results of running the extensive testsuite with the original DerbyDialect:
Tests run: 1417, Failures: 5, Errors: 58, Skipped: 0

Results of running the extensive testsuite with the patched DerbyDialect:
Tests run: 1417, Failures: 5, Errors: 40, Skipped:

And none of the errors with the patched dialect are new errors - ie.
no new errors are introduced by using IdentityGenerator.class.

All this suggests that whatever was causing the original problem with
Derby Identity has indeed been resolved so the hibernate guys should
probably revert.  And this will solve a problem for us.  I wonder
should I file a bug report with hibernate?

And this will probably be the least hacky solution for us.

Cheers
Bob

PS.  of course the 5 failures and 40 errors might still be
troublesome.  I haven't looked into the detail of how critical they
might be.  I'm just running the same suite against our h2 version now
to compare how it does.