Character Virtual missing contact normal

[LeXXik]

Trying to debug why the character virtual sometimes doesn't report a ground contact normal. It appears in my case that when the character moves on a slope uphill, the contact normal is received, but when it is going downhill, then there is no contact reported.

Initial debugging shows that when character moves downhill, OnContactValidate is called correctly, but the subsequent OnContactSolve is not. Any ideas why that could be or what is causing a validated contact not to process in solve? Could it be supporting volume being too high/low? Although, I'd suspect, it would behave same going up or down the slope, so must be something else.

[jrouwe]

Typically if you go downhill gravity is not strong enough for the character to stay connected to the ground and you'll float a little bit above the terrain before falling back onto it in subsequent frames. This is because a character has 0 vertical velocity if it is on the floor and usually a game character can accellerate very quickly, quicker than gravity accelerates in the vertical direction.

Are you using CharacterVirtual::ExtendedUpdate with a non zero ExtendedUpdateSettings::mStickToFloorStepDown? This mechanism tries to correct for that effect by forcefully moving the character back onto the floor if it is floating a little bit. Looking through the code, there may be a missing callback there too. If that's the case I'll fix it.

[jrouwe]

I've pushed a fix for CharacterVirtual StickToFloor so that it will trigger OnContactAdded callbacks. OnContactSolve will not be called, since that is only used when trying to find a sliding direction, which is not something that triggers during StickToFloor. It's also integrated in JoltPhysics.js now.

[LeXXik]

Thank you! Another question on the same topic. It looks like I can set the step height via the update settings and also via character virtual:

JoltPhysics.js/JoltJS.idl

Line 2750 in e2b9468

[Value] attribute Vec3 mWalkStairsStepUp;

JoltPhysics.js/JoltJS.idl

Line 2913 in e2b9468

boolean WalkStairs(float inDeltaTime, [Const, Ref] Vec3 inStepUp, [Const, Ref] Vec3 inStepForward, [Const, Ref] Vec3 inStepForwardTest, [Const, Ref] Vec3 inStepDownExtra, [Const, Ref] BroadPhaseLayerFilter inBroadPhaseLayerFilter, [Const, Ref] ObjectLayerFilter inObjectLayerFilter, [Const, Ref] BodyFilter inBodyFilter, [Const, Ref] ShapeFilter inShapeFilter, [Ref] TempAllocator inAllocator);

I guess the one on the character is more flexible, as it looks in 3 directions and can do filtering. I think I am safe to assume that I can use it instead of update settings in all cases. Is that correct?

Edit:
Actually, it looks like it might be used by the Character Virtual internally? As I don't think I need to track the forward direction of the character, etc. I guess defining the step heights and whether to stick to floor when going down should be controlled via the update settings then?

[jrouwe]

It is indeed the same thing. Calling ExtendedUpdate with the ExtendedUpdateSettings parameter passes that information on to WalkStairs internally.

ExtendedUpdate is nothing more than a call to Update, WalkStairs and StickToFloor in a particular order and with some glue code.

The forward direction is taken from the linear velocity.