What steps will reproduce the problem?
Invoke a transaction callback that raises an exception, which will cause a
rollback before the exception is re-raised in ndb.Context.transaction. If
tconn.async_rollback fails (which is difficult to do on demand but happens
every so often) the original exception is swallowed.
What is the expected output? What do you see instead?
I'd like to either see both exceptions re-raised as as JointException (as done
in taskqueue, for example), or at least see the lost exception logged. Losing
all trace of the original exception makes debugging difficult.
What version of the product are you using? On what operating system?
1.7.4 on OS X
Original issue reported on code.google.com by pi...@knowlabs.com on 9 Jan 2013 at 10:45
Original issue reported on code.google.com by
pi...@knowlabs.com
on 9 Jan 2013 at 10:45