I’ve recently come across a bug in one of my applications where
destroying a model can occur while it’s locked in another place for
processing. Just for the record: the destroy correctly aborts because of
this but there are files removed by dependent models being destroyed to
that can’t be restored when the rollback occurs (after_destroy is not
delayed until the actual commit…).
I know how I can solve this but I would like to create a test before
(TDD/BDD are my friends…) and I don’t know how I can handle this in
tests (reproducing a race condition isn’t especially easy).
Is there anyone on this list with experience to share on this particular