Subject: Integrator loops endlessly for lambert_w(1/x)
From: Stavros Macrakis
Date: Fri, 26 Dec 2008 13:14:40 -0500
On Fri, Dec 26, 2008 at 7:15 AM, Barton Willis <willisb at unk.edu> wrote:
> Ouch! Here is the bug displayed on the command line
> ...
This reminds me somewhat of the Methuselah bug (see SF bug 776441,
> orderlessp not transitive).
>
orderlessp is indeed not transitive in this case:
a: 1/(f(1/x)+1)$
b: 1/x$
c: %e^-f(1/x)$
orderlessp(a,b) => T
orderlessp(b,c) => T
orderlessp(a,c) => F
A slightly simpler case is the same as above but b:x$
-s