Thanks David for the response. The transactions tests failing are actually new because they are related to @Transaction support by CDI. I double checked and no there is no clean up performed. So again, if you are lucky and you never get a thread already used before, then you are good. Otherwise, you get random failures.
I opened https://github.com/eclipse-ee4j/jakartaee-tck/issues/573 Le ven. 13 nov. 2020 à 11:52, David Jencks <[email protected]> a écrit : > I don’t recall having this problem getting Geronimo to pass the TCK, but > maybe these tests were added since then. > > Are you sure there isn’t something in the adapter/test harness goo that > you have to write that cleans up transactional state between tests? I do > remember getting everything working in the test adapter was rather > difficult. > > David Jencks > > On Nov 13, 2020, at 12:59 AM, Jean-Louis Monteiro < > [email protected]> wrote: > > Hi all, > > Been trying to look at JTA support and see if TomEE passes TCKs. > Looks like we are failing on most of the tests in the package > > com.sun.ts.tests.jta.ee.transactional.* > > After digging into why they were failing I realized, transactions were > leaking into the thread context (ThreadLocal) and therefore if the test was > randomly picking a test that was badly used before, it would fail on > UserTransaction.begin because nested transactions aren't allowed. > > After digging into TomEE and Geronimo Transaction I realized i was > probably the TCK test which was not too good. > > Here is where it fails. > > https://github.com/eclipse-ee4j/jakartaee-tck/blob/master/src/com/sun/ts/tests/jta/ee/transactional/Client.java#L420 > > I see only begin() but no rollback/commit in some methods which means that > some transactions may leak. > > What do you think? > I will open a challenge against the TCK if we are all on the same page > with the analysis > > -- > Jean-Louis Monteiro > http://twitter.com/jlouismonteiro > http://www.tomitribe.com > > > -- Jean-Louis
