"Joshua D. Drake" <jd@commandprompt.com> writes:
> How about *requiring* test cases that prove the patch?
There are lots and lots of things that can't really be tested with
pg_regress-based tests, and in any case a test does not prove the
absence of bugs; particularly not a test devised by the code author,
since almost by definition it won't test cases he didn't think of.
Certainly test cases have their place, but I have a lot more faith in
code-reading by knowledgeable developers as a way to spot problems
that got past the original author.
regards, tom lane
