That was my observation as well. I think the txnmanager spec may shed some light? The original Jini specs are included in the doc folder.
2010/10/14 Patricia Shanahan <[email protected]> > On 10/14/2010 12:55 PM, Patricia Shanahan wrote: > >> Jonathan Costers wrote: >> >>> The complete current QA test suite passed with this change included. >>> Also, I ran the "txnmanager" category, all tests pass except >>> GetStateTest.td >>> (which hangs), but this one was already behaving that way without this >>> fix. >>> >>> I'd say commit it. >>> >>> PS: if anyone would be able to debug GetStateTest and figure out what is >>> wrong with it, we could add another 40-something tests (covering >>> mahalo impl >>> and the transactionmanager spec) to our QA run on Hudson (txnmanager >>> category). >>> >> >> I'm back from my vacation, and ready to resume River activities. I'll >> take a look at GetStateTest.td, unless there is something more urgent. >> > > I have a specification question. Is a TransactionManager *required* to > forget a transaction within some amount of time after it has been aborted? > > The test hangs because it is waiting for getState to throw an exception, > rather than going on reporting state 6, TransactionConstants.ABORTED. > > Patricia >
