i just looked at the output of randi in ..
the output is from 0-2*xamp 
did new improvements forget to subtract xamp on output?
7)matt

---------- Forwarded message ----------
From: vjs <vjs@ndirect.co.uk>
Reply-To: csound-dev@eartha.mills.edu
To: ingalls@mills.edu
Subject: re: randi bug (was sr=1 orcs and scroll bug)

you wrote
>have you tried unirand?

unless i misunderstand, unirand has neither band limitation nor
interpolation, also no negative numbers

my point -
in 3.492b, there is a bug in randi and randh - they output values greater
than the input argument xamp


a test

;orc
sr = 44100
kr = 441
ksmps = 100
nchnls = 1

instr   1
    a1  oscili  20000, 5, 1
    a2  randi   a1, 1000, 0.5
    out a2
endin

;sco
f1  0   4097    10  1

i1  0   5
e


similarly for randh

i do not have any other platforms to hand so unable to tell whether this is
just PPC or to do with canonical code


vjs