Peter Geoghegan <pg@bowt.ie> writes:
> I'm beginning to think that the technique that I came up with to make
> "getting tired" deterministic ought to be supporting as a debugging
> option if we're to do away with internal use of the generic/seedable
> backend PRNG.
I have no objection to providing such a thing as a debug option; I just
don't want it to be reachable from SQL (at least not without pieces
that aren't there normally, e.g. a loadable C function).
Replacing random() might actually make that easier not harder, since
we'd have more control over what happens when.
regards, tom lane