|From:||Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>|
|To:||Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com>|
|Subject:||Re: [HACKERS] user-defined numeric data types triggering ERROR: unsupported type|
|Views:||Raw Message | Whole Thread | Download mbox | Resend email|
Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com> writes:
> FWIW I originally constructed the tests merely to verify that the fix
> actually fixes the issue, but I think we should add some tests to make
> sure it does not get broken in the future. It took quite a bit of time
> until someone even hit the issue, so a breakage may easily get unnoticed
> for a long time.
Oh, well, that was another problem I had with it: those tests do basically
nothing to ensure that we won't add another such problem in the future.
If somebody added support for a new type FOO, and forgot to ensure that
FOO-vs-not-FOO cases behaved sanely, these tests wouldn't catch it.
(At least, not unless the somebody added a FOO-vs-not-FOO case to these
tests, but it's hardly likely they'd do that if they hadn't considered
I think actually having put in the coding patterns for what to do with
unsupported cases is our best defense against similar oversights in
future: probably people will copy those coding patterns. Maybe the bytea
precedent proves that some people will fail to think about it no matter
what, but I don't know what more we can do.
regards, tom lane
|Next Message||Tom Lane||2018-03-04 19:46:22||Re: IndexJoin memory problem using spgist and boxes|
|Previous Message||Tom Lane||2018-03-04 19:06:48||Re: PATCH: psql tab completion for SELECT|