Test regression in rtest_trig with cmucl



On Wed, Jul 18, 2012 at 9:17 AM, Barton Willis <willisb at unk.edu> wrote:

> This was my commit--I purloined the catch stuff from csign (defined in
> compar.lisp).  I believe  I tested the patch with CCL & SBCL.
>

BTW, do you mind if I change the tests in rtest_trig.  There are lots of
code like

(id : cot(2*x) - (cot(x)^2-1)/(2*cot(x)),0);
0$

(maxerror : 0,0)$
0$

(for p in pts do block([numer : true],
  maxerror : max(maxerror, cabs(expand(subst(x = p, id))))),0);
0$

is(maxerror < 1.0e-15);
true$

I think it would be beneficial if the initialization of id and maxerror be
grouped together.  The for loop part could be modified do something like if
is(maxerror < 1.0e-15) then true else maxerror.

That makes it clearer why the test failed.  And the number of tests is
"artificially" inflated because a lot of the tests are just setup for the
actual test, and then there's one additional test to see if it worked. :-)

Ray