17 Commits over 0 Days - ∞cph!
Remove last remnants of tile cache
Correctly account for areaflags in path filer
Fix sample
Fix demo cmake get rid of tile cache
Ensure area flags are correctly combined when marking areas
Merge polyflags and areas into one concept (areaflags)
Deploying to gh-pages from @ Facepunch/recastnavigation@7efacd65d0620b39c4ece43fba7c4c0e72b9aba1 🚀
Enable 64bits for poly references
Again since we are only targeting x64 this does not have a performance impact.
Memory usage will be slightly increased, but only if many refs are held by the user.
Don't consider cells inside a triangle if the triangle just touches the cell
Deploying to gh-pages from @ Facepunch/recastnavigation@ac488df6fe9a3c2575e36c4eddc5b03815891b5f 🚀
Removed unused dividePoly function see a4e1e2b
Deploying to gh-pages from @ Facepunch/recastnavigation@217f9d61e5a45e71722fa58f21345a117a81c9bc 🚀
Bump RC_SPAN_HEIGHT_BITS from 13 to 16
Increasing the bits used, will yield more accurate height information in the generated navmesh polys.
This does not change the size of rcSpan on x64 plattforms. (See https://godbolt.org/z/4eKq45daK)
x64 is the only platform we are targeting for s&box and Rust. Therefore, this is free accuracy improvement without perf or memory impact.
Deploying to gh-pages from @ Facepunch/recastnavigation@a4e1e2bbfc518e12fd3b582e4d4aeae9b5d81cb2 🚀
Fix bug leading to stack smashing
Fixes recastnavigation/recastnavigation#687
To address this, I basically reverted recastnavigation/recastnavigation@e063ba6
& recastnavigation/recastnavigation@adcd4f4 which seem to be the most likely culprit for the issue. Contrary to the commit message in e063ba6, which claims that the changes made are an optimization, reverting the changes had no measurable impact on performance.
