Query.setLockMode() doesn't seem to actually set lockMode

dmlloyd · **Joined:** Tue Jul 27, 2004 11:28 am **Posts:** 3

Under Hibernate 2.1.5 and 2.1.6 against Oracle 9.2.0 (via Oracle9Dialect), it seems that executing an HQL query with a lock mode set on a column alias via setLockMode("alias", LockMode.UPGRADE) doesn't actually mark the object as UPGRADEd in the session, at least in conjunction with joins. It does remember to put "for update" on the SQL though.

So something like this:

Code:

String MY_HQL
    = "          select foo "
    + "            from com.dmlloyd.Foo foo "
    + "inner join fetch foo.bar bar "
    + "           where foo.id = ? ";

Query q = session.createQuery(MY_HQL);
q.setLockMode("bar", LockMode.UPGRADE);
q.setLong(0, 1234);
Foo foo = q.uniqueResult();
LockMode currentMode = session.getCurrentLockMode(foo.getBar());
if (currentMode.lessThan(LockMode.UPGRADE)) {
    throw new RuntimeExcpetion("Barf: " + currentMode);
}

would end up showing: "Barf: READ" even though the object is locked in the database (and then some... I couldn't get Hibernate to do FOR UPDATE OF; even modifying Oracle9Dialect to return true for supportsForUpdateOf() didn't do it, so it just used a blanket FOR UPDATE across the join).

Any ideas?

dmlloyd · **Joined:** Tue Jul 27, 2004 11:28 am **Posts:** 3

Ah, corrections.

1) Hibernate will indeed do FOR UPDATE OF ..., for some reason my modified Hibernate didn't deploy.

2) For Oracle 9i, the syntax of FOR UPDATE OF is "FOR UPDATE OF TABLE.*", not "FOR UPDATE OF TABLE", which I assume is why Oracle9Dialect currently returns false for that capability; though this seems like a trivial fix to me. Thus I have reverted to stock 2.1.6 again.

The main point of the message remains, however.

gavin · **Posted:** Thu Aug 12, 2004 3:35 am

um

i don't quite understand what you are saying in the first post ... but this area could be buggy in 2.1 (its better in 3.0), so submit a bug report to JIRA (please be a bit more specific though).

TIA