[Octave-bug-tracker] [bug #58800] BIST for rng sometimes fails

[Michael Leitner]

Follow-up Comment #13, bug #58800 (project octave):

@Hg200: Yes, all you are saying about init_mersenne_twister() is correct. The
point I want to make is however that rand("state","reset") should not call
init_mersenne_twister() if we cannot guarantee that it will get a new seed
(different from the last invocation, however recent that was -- /dev/urandom
would guarantee exactly that), but that in this case something like
shuffle_mersenne_twister() should be called, which is guaranteed (in the sense
of pseudo-random numbers) to change the state so that the resulting stream of
random numbers has no recognizable relation. And I personally do not know, but
I am quite sure that it is known in which way seeds have to differ for this to
be fulfilled (or rather, it is definitely known which differences do not
suffice). But of course, if we can guarantee that we have a source (on all
supported platforms) for different seeds (even if we call it at the same point
in time) then it would be safer to use that.

    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?58800>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/