Mostly the errors were about incorrect object persistence state after
commit. This should be easy to fix. Also some drivers do not support
batches, so this should be turned off in default adapter (or rather batch
execution should be done as a set of queries), and adapters like Oracle
should turn it back on.
You can try switching your local copy back to ContextCommit and run unit
tests per Developer Guide instructions. You will see the full report. I
will also send you directly an archived report I got from Holger last
night.
Unrelated to that I wanted to organize the code a little better:
- we have too many sorters - old ones and new ones, need to merge them
together.
- I fill like using too many classes with default access is an indication
that some refactoring needs to be done, maybe with ressigning roles to the
participating classes. This will probably include ContextCommit, maybe
Craig's flattened relationship classes, and maybe DbAdapter
- ...
Andrus
> Andrei, is it just unsupported addBatch or there are more serious
> problems with the code? Is any logic broken? Because if it relates to
> the batching only customization of BatchInterpreter per DbAdapter is a
> piece of cake. Otherwise could you let me know what kinds of failures
> you observed? Maybe it will be easier for me to analyze and eliminate
> them.
>
> ----- Original Message -----
> From: "Andrus" <andru..bjectstyle.org>
> To: "cayenne-devel" <cayenne-deve..bjectstyle.org>
> Sent: Thursday, February 13, 2003 5:22 AM
> Subject: Re: Nightly build failure
>
>
> I looked at this some more, it will take me more than a few hours to
> clean up. What I am planning to do tonight is to revert Andriy's changes
> to DataContext, so that we have a working Cayenne again (I know its my
> fault - I recommended Andriy to switch Cayenne to his ContextCommit
> without real testing, so now I will pay the price :-)).
>
> After that I will make another switch attempt, slowly, one step at a
> time, just like we planned before. I think I know why the unit tests are
> failing, and I can probably do a quick patch, but instead I want to
> cleanup access layer architecture that became real messy recently.
>
> Hope I can do this without switching to a branch...
>
> Andrus
>
>
> At 05:47 PM 2/12/2003 +0100, Holger Hoffstätte wrote:
>>Andrus Adamchik wrote:
>> > I will try to fix everything tonight.
>>
>>I am currently running all the test suites against the different
>> databases that I have here (Oracle8, MySQL, hsqldb, Postgres 7.3) and
>> can send you the zipped HTML pages so you don't have to run them all
>> for yourself. The two remaining failures that I had with hsqldb
>> (related to constraints) are gone with the new commitChanges(), but
>> instead I now get lots of
>>'unsupported operation' from the JDBC driver (it doesn't like
>> addBatch). It seems batching throws off most of the other adaptors,
>> too. Many failures (e.g. NPE in DataContextDeleteRulesTst) are common
>> to all adaptors.
>>
>>Holger
This archive was generated by hypermail 2.0.0 : Thu Feb 13 2003 - 10:31:43 EST