Re: Segmentation fault using digest from pg_crypto

Bruce Momjian <bruce(at)momjian(dot)us> writes:
> Tom Lane wrote:
>> If you don't re-run a contrib module's sql script then you don't get any
>> new functions that may have been added to the module. So I think the
>> real issue here is that we ought to deprecate the idea of skipping that
>> step, period.

> The good news is that the scripts have CREATE OR REPLACE function so it
> would clean up everything to match the library you just installed.

Hmm ... now that you mention it: some of them do, some of them don't.
And some of them wrap the whole script in BEGIN/COMMIT, which means that
a conflict against existing definitions would prevent any new
definitions from being added.

Seems like maybe we need an explicit project policy that contrib scripts
should be able to overwrite an existing older installation properly;
maybe with some error reports, but not failing outright.

If so, someone will have to do the legwork of really making that happen
... any volunteers?

regards, tom lane