From: Richard R. <sf...@ol...> - 2003-10-05 02:49:02
|
On Sun, Oct 05, 2003 at 12:05:14PM +1000, Nigel Stewart and Fiona Smith wrote: > > I remember a message about fixing sphere and cone geometries, > >and since those are also bugs (in a loose sense--differences from GLUT) I > >think that if they are already in CVS we shouldn't take them out. > > Spheres and cones are broken on 2.0.0 when > back-face culling is enabled. There are > two main options: > > 1. Fix the current implementations. [...] > 2. Roll in a large scale revamp. [...] > If there is general support for (2) it would > be straight-forward to add a sphere replacement > following the cone and cylinder approach. As long as minimizing the ties to GLU is a priority, I'd say go ahead and commit your code. It fixes a bug and improves the code quality. And (less the sphere code, which you say is straightforward), it is already done. I would only add the reservation that I've seen mixed signals re. GLU dependance. (Personally, I don't particularly care about GLU independance, but if someone wants to try to sever that link, your code will help.) For the sake of posterity, you might add a comment at the top about the option to pass the buck to GLU---and why it's not done. [...] > precisely the same, rather than relying on > sin(0) == sin(2*M_PI*i/slices) == 0.000000000000 It may amuse you that in Octave: octave:1> sin(2*pi) ans = -2.4492e-16 (^& (Results truncated to 5 digits per default format.) -- "I probably don't know what I'm talking about." http://www.olib.org/~rkr/ |