Mikko Mononen

Got it :) I think that kind of filter is doable. It is possible that it will ran into false positives. Digging out neighbour information of heightfield is not fun,...

If the filter is set up so that it will try to "fill in" only if the height diff between left and right is less than climb, then that above...

I'm pretty sure the code tries to find a delaunay edge (no other point can be inside the circum circle of 3 points), but due to floating point inaccuracy there...

The dungeon level was kindly provided by John Ratcliff. I got bunch of nice levels from him which were big help in developing Recast. I think he distributed some of...

Zero area polygons should not happen. A screenshot of the case would be a good starting point to see where things might go wrong. --mikko On Wed, Dec 4, 2019...

I'm not sure what I'm looking at. Tha does not look like the navmesh in those pictures. On Fri, Dec 6, 2019 at 7:00 PM Erik Wijmans wrote: > Screen...

I have to say, I cannot remember anymore which way the polygons wind :) One thing I always found tricky was that a lot of algorithms assume that X goes...

@jakobbotsch when I implemented the span pool, it made the allocation significantly faster (compared to growing array) especially on larger levels because realloc would not need to move previously allocated...

@jakobbotsch It was big list of spans with indices for linked list.

The problem for this is 2 fold: 1) The tile cache code should remove the extra vertices along the tile seams, but for some reason it cannot https://github.com/recastnavigation/recastnavigation/blob/master/DetourTileCache/Source/DetourTileCacheBuilder.cpp#L1926 2) Each...