On 1/8/08, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote: > The https://issues.apache.org/jira/browse/MATH-164 issue presents a > multiplication case which is neither consistent with double multiplication nor > compliant to C99. Considering the division case which already handles lots of > special cases, I think we should really be C99-compliant for this method. > > Do we change the behavior of the method and become C99-compliant (thus > resolving > the issue) or do we change the javadoc only and explain what is done in this > case (thus choosing wontfix for the issue) ? >
If you can find a way to achieve C99-compliance without killing performance and maintaining clear documentation, that would be optimal. IIRC, however, there are two problems with this. First, unless things have changed, unfortunately the spec is not public, so we can't provide a public reference to it, so we have to basically embed full documentation of all of the corner cases in the javadoc. Second, when I looked at this before, I could not see a performant way to fully implement the spec in Java. Of course, that could just be stupidity on my part ;) I would be fine with anything in between what we have now and full C99x compliance for 1.2, as long as we maintain complete and clear documentation. Presenting the computational formulas used and referring to JDK documentation is preferable, in my opinion, to referring to a non-public C spec. Phil > Luc > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
