On Fri, Mar 20, 2009 at 3:22 PM, Dieter Kaiser <drdieterkaiser at web.de> wrote:
> Very often I have the problem, that expressions which look equivalent do
> not pass the test_suite. Today, I have observed an issue with $ratsimp.
The Maxima evaluator (meval) simplifies the result of every
calculation, so $fff functions are allowed to return unsimplified
results, as in this case. On the other hand, almost all Maxima
internal functions (including alike1) expect simplified arguments, and
are not guaranteed to work correctly on unsimplified arguments.
So there seems to be a bug in laplace where it is calling $ratsimp
without calling simplifya. This is taken care of if you call sratsimp
instead.
-s
> (%i16) integral:laplace(sinh(a*t)/a,t,s);
> (%o16) 1/(s^2-a^2)
> (%i17) result:1/(s^2-a^2);
> (%o17) 1/(s^2-a^2)