Hi Robert, sorry i am on vacation writing from cell phone. I should probably say more about all items for the submission but i assumed they were obvious. We were talking about LOD problems with shadows on users group. Let me explain why i modified light dir multiplication to use transform3x3. Light dir was stored on vec3 so if it was representing light direction ie point in infinity it should use W=0. vec3 * matrix assumes W=1 thats why i changed it. Third modification was really making shadow hud functional as it was simply not working. It also should not matter when hud is off. Last one ie texgen was well described i hope. I apologize for the inconvenience. I will be more verbose next time. I did not try to do anything fancy with these mods i wanted to fix the obvious bugs I found while developing ShadowMap derived class. I ll be back on 14th of march. I will be happy to discuss this in greater detail then. Regards, Wojtek > Hi Wojciech, > > I've merged all your changes, tested and submitted to SVN. I am > comfortable with your changes w.r.t texgen precision. > > I am however, a bit uncomfortable with the extent of the changes that > are happening to ShadowMap without me fully understanding them all. > The issue here is that once things stop working I'm in a weak position > w.r.t guessing what might be broken and how to fix it. It now feels a > bit like everybody is having a go at osgShadow implementations but > with this we're loosing coherency of the implementations, there are > also regressions happening. > > W.r.t regressions, running > > osgshadow --sm --base > > The result isn't right, the shadows from the base have a diagonal band > running across them that rotates with the eye point. I reverted back > before your changes and found the same issue so your changes aren't to > blame for this regression, but this regression as happened at some > point and I no longer know when. I fear that the implementation now > works better on some models but worse on others and its implementation > will flit between what works well depending on whom contributed the > latest "fix". >
_______________________________________________ osg-submissions mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
