Re: Fastcode CharPosRev B&V 0.5.0

Hi Dennis,


I do not think that the random function matters much in this situation and
think that the buildin is ok. Opinions wanted.

I thought the problem with the random is that, it's random and Delphi's
random has gotten better, but is not always equaly distributed..
We could use MT for 2 options:
1. Better distribution
2. We could use a fixed state as input so each time the same random string's
are generated? (this would mean more stable results I would guess)

When using option 2, there would be the need for big test cases or else some
states might be a little bit baised..

Do we really want so long strings?

I was just guessing, maybe looking at the Pos BV?

Best regards,
Davy Landman


.