Hi Andy,
this is great news! I must have missed that you changed back JPOX to the
behavior of 1.1.0 with respect to object deletion.
I can confirm that the TCK passes uses the latest JPOX SNAPSHOT. There
is only one query failure (see JIRA JDO-390 for details), but this is
not related to the issue of object deletion (JDO-392).
Regards Michael
Current TCK against the current JPOX snapshot fails in tearDown due
to foreign key constraint violations. The TCK might need to change
based on the resolution of JDO-392. AI Erik change JPOX to allow the
TCK to use the old behavior; the TCK might need a PMF flag for interim.
JPOX (nightly builds, available on our Maven repo) has for the last several
weeks used the same behaviour that was used at 1.1.0 ... namely
* if delete-dependent then delete dependents
* else if nullable null the FK
* else leave to the DB
It has no such issue with any TCK snapshot that i've checked out over the last
2 months. Not sure what people are using there.
I totally agree with Jorg that the use of <foreign-key> should be taken into
account in this process. What it should be IMHO is
1. If delete-dependent, then delete dependents
2. If not dependent, but an FK is specified leave to the DB to sort it out
3. If not dependent, and no FK, null the FK (if nullable)
4. Else leave it to see what happens.
The code for this is in JPOX CVS commented out (for 1-1, 1-N, M-N, N-1
relations), but am waiting on JDO-392 to consider the issue before I enable
it. If JDO-392 (and the content of the spec on this issue) doesn't get any
action in the near future then it will become a JPOX extension.
--
Michael Bouschen [EMAIL PROTECTED] Engineering GmbH
mailto:[EMAIL PROTECTED] http://www.tech.spree.de/
Tel.:++49/30/235 520-33 Buelowstr. 66
Fax.:++49/30/2175 2012 D-10783 Berlin
