MySQL + PostgreSQL broken

[thomas.diesler]

A recent change seems to have reintroduced the clean tables problem.

http://jbpm.dyndns.org:8280/hudson/job/jBPM3-Matrix/container=jboss422,database=mysql,jdk=jdk1.5/lastBuild/testReport/org.jbpm.bytes/ByteArrayDbTest/testManyBlocks/

org.jbpm.JbpmException: couldn't execute sql 'alter table JBPM_ACTION drop foreign key FK_ACTION_EVENT'
 at org.jbpm.db.JbpmSchema.execute(JbpmSchema.java:376)
 at org.jbpm.db.JbpmSchema.cleanSchema(JbpmSchema.java:244)
 at org.jbpm.db.AbstractDbTestCase.ensureCleanDatabase(AbstractDbTestCase.java:103)
 at org.jbpm.db.AbstractDbTestCase.tearDown(AbstractDbTestCase.java:74)

Again, I believe that a data base test should not have to execute DLL statements. Instead the test should properly clean up after itself.

This is related to http://www.jboss.com/index.html?module=bb&op=viewtopic&t=142374

[tom.baeyens]

in the meantime, till all tests clean up their data, i think it is best to fail that test, delete the data and then continue. no ?

the extra checks i added reveal that a lot more tests don't yet clean up their data then originally was indicated in terms of jira issues.

[thomas.diesler]

Well, as you can see here

http://jbpm.dyndns.org:8280/hudson/job/jBPM3-JDK1.5/14/

a large portion of tests are failing because of this. Generally, if one introduces regression he should fix it the next day, which should always possible with rollback.

We did have successful test runs where these additional left over records did not matter. Yes the test does not clean up properly, but if the following tests are not effected this is not a critical issue and can be dealt with in the next release cycle.