Use of C99 extra long double math functions after r236148

[b. f.]

>This discussion confirms my impression, that it should be possible as an
>interim solution, to use a port for missing math functions (cephes alike
>or whatever). The port itself could warn the user about inaccuracies and
>edge-cases.

Parts of Cephes are already in ports: math/ldouble.  I had planned to
add the remainder.  It can be useful, but it has problems, some of
which have been described.  The same is true of other open-source
alternatives that I am aware of.


>I would be happy to take on any steep learning curve, and make
>contributions, if only I were part of a group that would steer me in the
>right direction.


A Functional Analyst offering to write code for a floating-point math
library!?!! ;)  You should send me a message off-list: maybe I can
find some time to help.


>Anyway, given that floating point is a big issue, and we are about a
>decade behind schedule, really suggests that a
>floating-point at freebsd.org mailing list is needed.  Or maybe there is an
>existing freebsd mailing list you guys already occupy.


I do not know that a separate FreeBSD mailing list would be of great
help, considering the likely volume of traffic, and the fact that we
already have the standards mailing list.  There is also:

http://mailman.oakapple.net/pipermail/numeric-interest/

which serves, among other things, as a forum for discussing changes
and improvements to the open-source math library from which parts of
our system math library were derived:

http://mailman.oakapple.net/pipermail/numeric-interest/2010-September/002054.html

A wiki page detailing problems and procedures, with a wish list for
the system libraries and toolchain(s) might help.  At present there
are only scattered messages in the mailing list archives, PRs, and
http://wiki.freebsd.org/MissingMathStuff .

b.