Perspective FOV/aspect wrong, ... and more
The aspect = xfov / yfov calculation described in the COLLADA documentation of the perspective node is wrong, since the fovs are angles. Anyone using it will not produce proper displays. I've shown the correct formula in the bug report on sourceforge.
It looks like the OSX viewer posted on sourceforge has it wrong, I'm not getting the right results from it. Having it barf because there is no profileCG when there is profileCommon doesn't exactly inspire respect.
I'd really appreciate it if someone could start interacting with me (PM?) on these posts. I'm a commercial developer about to announce a product that can write substantial COLLADA files containing geometry, node hierarchies, animation, and texturing. They pass verification testing ok.
I need a known-good full-featured "reference" COLLADA viewer to use as a gold standard to test the exporters, and someone with some experience with existing commercial importers so I can tweak exports if needed to compensate for reader problems.
I want COLLADA to work, it has a lot of potential. I'd love some pointers to the importer developers at the various companies, for example.
Hello? COLLADA? COLLADA? The ref is counting... 10... 9.... 8...
Sorry to whine but I don't want my development on this to be a loss. Thanks.
Re: Perspective FOV/aspect wrong, ... and more
sourceforge is not the place to report spec bugs. It's http://www.khronos.org/bugzilla .
Originally Posted by Rasson
In that instance the spec is describing a relationship, not a literal calculation, of horizontal over vertical. I agree it could use some clarification as you point out and even in regards to angle vs field of view, as different domains used the terms differently.