Angled env_physics_blocker entities block hittables

[Derpduck]

env_physics_blocker entities rotated by an angle will always block physics objects (hittables) regardless of the block type used, where the intended behaviour is for only block type 4 to block physics.

This means that any angled blocker that is supposed to only prevent exploits will interfere with hittables, and without fixing this bug the only solution would be to compromise on exploit blocks, or replace them with the obsolete env_player_blocker entity.

[SirPlease]

Do you know of a certain spot where this would be easy to test?

[Derpduck]

This blocker by the forklit should be fine (on 2.0) https://i.imgur.com/18ig6Ji.jpg

[Derpduck]

Here's a video to demonstrate this issue
https://youtu.be/q3ya0TUZEP0

[SirPlease]

Thank you for saving me the time, I'll test a fix tomorrow, promise.

[SirPlease]

I'm fairly this has to do with the fact that they're added by TLS, from what I've been able to test and try so far I haven't been able to modify its settings through a plugin.

[Derpduck]

From what I know this issue has always existed, it just hasn't been noticed much before because angled clips were rarely used, but it affects all clips including those added by Stripper.
I don't think theres anything wrong with the settings of the clips themselves, they still correctly block the survivor/infected types they are set to block.

[Derpduck]

Did some testing and found a page that might point towards the reason for this bug https://developer.valvesoftware.com/wiki/CollisionProperty
It could be that physics props or the blockers themselves are set to have collision with env_physics_blocker entities by m_Collision or m_CollisionGroup net props in situations where it shouldn't collide.

Edit: Tested setting various collision related properties on physics props and blockers but hittables were still always blocked

m_CollisionGroup
m_Collision.m_nSolidType
m_Collision.m_usSolidFlags

[A1mDev]

Physical object (prop_physics) when approaching an angled env_physics_blocker entity just collides. It seems this should call the function 'CEnvPhysicsBlocker::ShouldCollide(CEnvPhysicsBlocker *this, int, int)', in it it is checked whether it can collide with this type of objects or not (depending on the set property 'BlockType'), but it doesn't happen. I think this is because the property m_nSolidType or m_usSolidFlags is different for the object angled env_physics_blocker entity, and this collides due to the fact that some conditions are the same in the code. I don't think it's good to change these properties. There is another option, the collisionhook (ShouldCollide) will handle this very well :D.

[jensewe]

void CEnvPhysicsBlocker::Spawn()
{
	SetBlocksLOS(false);
	GetAbsOrigin();
	
	if ( m_nBlockType != 4 // All players and physics objects
		&& m_angAbsRotation != vec3_angle) // Not a null vector
	) {
		SetMoveType(MOVETYPE_NONE, MOVECOLLIDE_DEFAULT);
		SetCollisionGroup(COLLISION_GROUP_PLAYER_MOVEMENT);
		m_Collision.SetSolid(SOLID_BBOX); // an AABB
	}
	else
	{
		SetMoveType(MOVETYPE_VPHYSICS, MOVECOLLIDE_DEFAULT); // uses VPHYSICS for simulation
		m_Collision.SetCollisionBounds(m_vMins, m_vMaxs);
		m_Collision.SetSolid(SOLID_OBB); // an OBB (not implemented yet)
		VPhysicsInitStatic();
		m_Collision.SetSolid(SOLID_VPHYSICS); // solid vphysics object, get vcollide from the model and collide with that
	}
	
	SetBlocksLOS(false);
	SetBlockState();
}

Make sense for VPhysics_ShouldCollide helping in this scene given the above.

It's been over a year and there's always an easy fix via CollisionHooks according to @A1mDev though it's not working for Windows currently. However, a few questions in order for a possible way without noticeable consume:

• Is there any alternative to avoid certain collision judging from above code?
• Is it acceptable to use multiple blockers to simulate rotations?
• Could we apply angles inputs to blocker afterwards and make it work as intended?

[A1mDev]

Yes, collisionhook is the worst option, but a working one, for those who do not want to delve into the problem.