persistence@glassfish.java.net

Transaction bug?

From: Scott Oaks <Scott.Oaks_at_Sun.COM>
Date: Fri, 19 Jan 2007 15:59:02 -0500

[I'm resending this -- apologies if some of you receive it twice, but
many apparently didn't get it the first time.]

Our QA group has a test that executes code similar to the attached. This
worked fine in 9.0 and 9.0 UR1, but is now failing on 9.1.

In a nutshell, the attached code has a 2.1 session bean which calls a
custom query on an entity manager; the custom query issues a "for
update" to lock the row. This all happens within a user-defined
transaction. In 9.1, the transaction doesn't seem to commit correctly;
the row remains locked in the database forever. Hence, successive
attempts to access the row hang.

Is there something wrong witht the test case (it is quite an odd test,
I'll admit), or it this a new bug?

_Scott