[Ecls-list] trouble building Maxima w/ ECL

[Raymond Toy]

On 6/21/10 9:09 PM, Robert Dodier wrote:
> On Mon, Jun 21, 2010 at 4:14 PM, Juan Jose Garcia-Ripoll
> <juanjose.garciaripoll at googlemail.com> wrote:
>
>   
>> My apologies. Maxima was building just fine when I decided that it would be
>> nice to speed up the inline code for SIN / COS / etc even when the argument
>> does not have a well defined type. Problem is I uploaded only half of the
>> changes. I just finished a successful build of Maxima with ECL _as it is_
>> with no further changes, I promise :-)
>>     
> OK, thanks. I updated ECL and built Maxima successfully.
> The test suite has several failures in rtest16, several having to
> do with power series and a few with numerical evaluation of the
> zeta function. I built Maxima w/ Clisp and there are no errors
> in rtest16. I didn't look more carefully at the power series or
> zeta stuff to see what is the matter. Also, I don't remember if
> previous versions of ECL + Maxima had the same errors or not.
>   
I have not tried cvs version of ecl, but I suspect the zeta stuff is
caused by the fact that (expt 2 #c(1d0 0d0)) only gives single-float
precision instead of double-float.

I think it would be nice if ecl returned double-float precision for
this, but I don't think it's required by ANSI.  However, clisp, cmucl
and gcl do, and ccl will (changes applied but not updated),  I have
heard sbcl will too.

I'd rather not work around this in maxima, but if necessary, I guess
we'll have to.

Ray