[Ecls-list] Another bugfix release
toy.raymond at gmail.com
Thu Aug 13 01:03:07 UTC 2009
Juan Jose Garcia-Ripoll wrote:
> ECL 9.8.3:
> * Bugs fixed:
> - FLOAT-SIGN ignored the second argument.
> I know I should have done this before producing a new file, but better
> later than never. With this fix, the number of regressions in Maxima
> decreases enormously, leaving just about four or five failed tests.
> One of the failed tests is related to FLOAT, and the conversion of a
> rational to a double float. The point is that GMP's mpq_get_d
> truncates to zero, while Maxima expects FLOAT to round. I am not sure
> that the later assumption is reasonable, and without mpq_get_d it
> would get too complicated...
That's a good point. I couldn't find anything that says FLOAT should
round, but it is certainly my expectation to round instead of truncate.
You could steal the code from cmucl or sbcl which makes float round.
It's pure lisp, but not quite portable.
More information about the ecl-devel