Hee. It gets better.
I have two unit tests now. One tests the DAO that wraps access to
unversioned nt:resources, the second tests the DAO that wraps access to
versioned nt:resources.
Either test, by itself, works.
Both run (via the eclipse junit test on a source folder) has the second
one fails. I don't know if it's due to the test itself or merely its
position as second. The error is on teardown - I can't delete the
repository files.
However if I comment out a final method that simply reads the properties
of the jcr:content associated with a specified node then both tests
succeed. I've single-stepped through the code and I'm not throwing any
unexpected exceptions that would prevent the session from being properly
closed. Plus I -know- that the unit test succeeds when I run it by itself.
My only idea was that there's a race condition where the second unit
test is trying to use a repository directory with the same name as the
first unit test, but that doesn't hold up either since I get the problem
even when I append a random number to the directory name.
Could there be other race conditions in transient repositories?
Bear
