If you run either the DoubleFlushTest or FloatFlushTest individually, then the test passes.
If you run them both together then the FloatFlushTest fails and doesn't ever persist the updates to the DB.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
The patch shows opening DB1, updating it, closing DB1, opening DB2, updating it, closing DB2, the opening DB2 and the changes made to DB2 will not have been persisted.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
The bug highlighted is very different from the original one even if it seems to be the same because the effect is the same.
The severity is high as well as that of the first one.
Still next week(saturday) for the release of the final fix.
Thank you William!
Last edit: Marco Amadei 2016-05-15
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Continuing on the discussion at http://chat.stackoverflow.com/rooms/111964/discussion-between-jamadei-and-william for http://stackoverflow.com/questions/36964783/ensuring-high-precision-float-value-doesnt-block-other-jpa-updates
It seems that to trigger it you need to exercise both the DoubleFlushTest and FloatFlushTest.
I have created a patch for UCanAccess that does exactly that.
See https://drive.google.com/folderview?id=0BzAjgkxDofpueGc5dUdaVmFTdEU&usp=sharing
It contains the patch plus the 2 test databases.
Run it via 'mvn test -Dtest=AllFlushTests'
If you run either the DoubleFlushTest or FloatFlushTest individually, then the test passes.
If you run them both together then the FloatFlushTest fails and doesn't ever persist the updates to the DB.
A better version of the patch is attached.
Database files for the patch are attached
And here is a patch that contains just a single test showing the error
The patch shows opening DB1, updating it, closing DB1, opening DB2, updating it, closing DB2, the opening DB2 and the changes made to DB2 will not have been persisted.
Note that the error only occurs when SingleConnection=true
The bug highlighted is very different from the original one even if it seems to be the same because the effect is the same.
The severity is high as well as that of the first one.
Still next week(saturday) for the release of the final fix.
Thank you William!
Last edit: Marco Amadei 2016-05-15
Tomorrow, if I can, I'll put the fix on svn and let you know when it's completed so that you'll be able to test it if you want.
FWIW, I just tried William's latest test patch against Revision 535 and both tests passed.