On Sat, Aug 16, 2008, Adriaan de Groot wrote: > On Saturday 16 August 2008, David Schultz wrote: > > On Sat, Aug 16, 2008, Daniel Eischen wrote: > > > Additions of some (and not all) long functions have been > > > breaking ports that just perform cursory checks for a couple > > > of long functions and then think they have access to them > > > all. > > > > The last time this issue was brought to my attention, the problem > > port was also a part of KDE, by the way. It might be a good idea > > to talk to the KDE developers about it so they don't keep > > replicating the same mistake. > > Agreed. The checks have traditionally assumed the presence of all the > functions in a family of *l-suffixed ones. Whenever I've noticed, I've added > checks upstream. Please do submit a bug report at bugs.kde.org with the OS > set to FreeBSD - I check those periodically and clean up / fix up what's > needed.
I looked at the kcalc source, and the whole thing is so broken that you are better off just turning off HAVE_L_FUNCS regardless of anything. KCalc casts the input and output of each libm function it calls to double precision even if it's calling the long double version. This means that it actually achieves *worse* accuracy by using the long double functions, due to rounding twice. _______________________________________________ kde-freebsd mailing list [email protected] https://mail.kde.org/mailman/listinfo/kde-freebsd See also http://freebsd.kde.org/ for latest information
