Bruce Momjian <pgman@candle.pha.pa.us> writes:
> We do have:
> #geqo_random_seed = -1 # -1 = use variable seed
> that lets you force a specific random seed for testing purposes. I
> wonder if that could be extended to control VACUUM radomization too.
> Right now, it just controls GEQO and in fact gets reset on every
> optimizer run.
Actually, just the other day I was thinking we should take that out.
Since there is only one random number generator in the C library,
GEQO is messing with everyone else's state every time it decides to do
an srandom(). And there is certainly no need to do an explicit srandom
with a "random" seed every time through the optimizer, which is the
code's default behavior at the moment. That just decreases the
randomness AFAICS, compared to letting the established sequence run.
regards, tom lane