Andrew Sullivan <andrew@libertyrms.info> writes:
> Yes, 3 weeks of testing with the patch that Jan provided for 7.2. We
> had no failures that I know of. It provided no measurable
> performance gain, but it also never deadlocked, and we tested under
> conditions where it sometimes did in the past.
> Note, however, that we were testing it indirectly; that is, while we
> were testing to see if it would break, our application (which does a
> number of the referential checks itself, multiple-connection-safety
> notwithstanding :( ) tends not to try to violate the foreign keys
> anyway. I wouldn't want to claim that we've tested it real heavily.
Nonetheless, this does seem to speak to my real concern, which is
whether the patch introduces any unexpected side-effects. The code
was getting executed, whether or not it detected any FK violations,
so we can have some hope that any bizarre problems would have been
noticed.
I'll go ahead and apply it for 7.3.3.
regards, tom lane