fx navmesh scale detection
fill in mission file based .nav file name
account for walkableclimb having more filter weight than walkableslope when it comes to slanted quads generating navmesh data
move the geo collection into each tile, seems to work better for large levels
add true to getnav in aiconver so it overwrites all goals
cache the triareas so we can use a tool later to modify them
Added ground work for tester tool
tester tool works but needs to fill out list of acceptable datablocks and spawnclasses
navpaths now share 1 navmeshquery
AIControllerData now has a vector of area costs for different polyareas
General cleanup
added an override flag to stop detour from setting our depth mask state. This was causing navmesh to draw through other objects when it wasnt meant to
Reset our bounds box for each draw cache
Adds off mesh connection tool
upgrade functionality to allow setting the direction to be bi-directional
added immediate draw to duDebugDrawtorque so we can draw offmesh connections
DebugDraw for recast now caches the results
We now have a drawmode dropdown selector
drawmode changes come from the gui itself no longer from console values
all recast drawmodes are supported with the exception of drawmodes that add abilities like navqueries until the nav tester tool is imlpemented.
cache tiles data if keep intermediate is on
(we only need to cache the results of recast)
fix tile generation (again)
Add !m_geo check so that buildTile can regen the geometry needed to build the tile again.
Added chunkytrimesh - this class splits up the geometry the navmesh is interested in into kdtree for fast traversal, makes the actual navmesh generation work with smaller chunks.
Now only 1 RecastPolylist per navmesh this can be saved out in a future commit.
This is a history commit, all functionality works same as it did before but it matches recasts recommended setup more closely. Future additions may break backwards compatibility.
1) Reset the build when adding potential dirties to the list to ensure it isn't trying to kill off a dirty entry that it's passed by.
2) mSaveIntermediates = true; causes leakage even with all this. still tracking that end.
3) Need to remove a dead tile whether there's new data to replace it with or not. Empty tiles are an entirely possible case. Even a probable one if you have high verticality in a level.
4) Likewise you don't want to re-feed the same geometry list for a given tile in case the conditions have changed.
5) If no vertcount then clear all geometry entries. (that one might be paranoia)
Many instances where we would create a object via a new call, and then check that it was non-null.
This is redundant, as if we ever were in a situation where new failed, we'd be crashing left and right already, so the additional check is wasted processing.
A lot of instances where some function args are not actually modified in any way, meaning that it is better for performance to convert them into const references. This prevents an additional copy, which can help performance.
