🚀 go-pugleaf

In article <886424243snz@inf-eden.demon.co.ux>, Chris Kennington -
newshound <Newz@inf-eden.demon.co.ux> writes
>
>    There is a very easy way to get a divide overflow, and this is
> by using some sort of processing-speed test designed for a slower
> m/c.  Typically this does some piece of computation and times how
> long it took.  If it forgets to check for zero time (ie no
> clock-tick taking place) before dividing to get the speed, then
> it will eventually fail as speeds get faster.  MS-K 3.13 gives no
> trouble on my Cyrix P133+; but that's some way behind what you
> have now got.

One of the problems is (ISTR) that the standard Borland libraries suffer
from a bug of this kind. (It may well be that similar MS libraries do
to.) The result of which is that these bugs are quite common.

Still, it's better than the old way of using a fixed timing loop
(expecting to see a reply in 200 loops, for example); software such as
that (and there's a lot of it) fell over a long time ago.

-- 
Mark Rogers (Software Engineer)       -o-      mark@controlware.demon.co.uk
2M_ControlWare, Yaxley, Peterborough, ENGLAND.