On 1/9/07, Stavros Macrakis <macrakis at alum.mit.edu> wrote:
> > (%i1) prederror : false$
> > OK more or less:
> > (%i2) if f(k) then 0 else %pi;
> > (%o2) if unknown then 0 else %pi
>
> No, this is worse than useless --
I agree this behavior is suboptimal.
> it seems that the unevaluated conditional experimentation has
> leaked into the production version.
That is not the case. This behavior has been present at least since 5.9.1.
> 2) if there are going to be unevaluated conditionals, then they should
> *preserve* things that cannot be evaluated completely,
Agreed.
load (boolsimp);
if f(k) then 0 else 1;
=> if f(k) then 0 else 1
Robert