On Jun 12, 2013, at 9:18 PM, Miłosz Pigłas <mil...@archeocs.com> wrote:
> thank you for your help. Adapter is almost ready, Great! > but there is still one test case: > org.apache.cayenne.access.NestedDataContext_DeadlockTest, which fails. Test > passes only if checking threads is delayed few second - probably firebird > performance is to low. Yeah, don't worry about this. This test is not adapter-specific. > Test org.apache.cayenne.query.SelectQueryTest#testSelectLongIn(), ends with > error because of generated statement size and firebird limitations > (http://www.firebirdfaq.org/faq299/). Are such errors aceptable? For other DBs we used UnitDbAdapter as a means to suppress certain unit tests on some DBs. Although looking at this test, I am not quite sure what it tests :) I'd just remove this particular test actually. There are cases when *Cayenne* generates IN clauses, e.g. for prefetching or paginated queries. And the max size of IN is controlled by the following property: http://cayenne.apache.org/docs/3.1/api/org/apache/cayenne/configuration/Constants.html#SERVER_MAX_ID_QUALIFIER_SIZE_PROPERTY But we are not dealing with this here. So yeah, let's remove it. > Some tests, which tries to delete all records from table where one record has > reference to other record in _the same table_ also ends in error. Usually in > method setUpAfterInjection(), which prepares database to test. > Firebird doesn't recognise that all records are deleted and denies breaking > foreign key constraint. > Example of such case is org.apache.cayenne.access.JoinPrefetchTest. > > My question is how to deal with such errors? In this case solution might be > removing references before clearing table, but I'm not sure it is good idea > to change code of test cases. I am +1 on doing that if it solves the problem. How we prepare test cases is up to us, and if we come up with a strategy that works with more DBs, I am all for it. Thanks, Andrus
