Candle holder question

[Ishtvan]

I'm trying to fix the Grabber bug where candle holders with candles bound to them don't rotate right. Found something odd though, holder_tall_* doesn't work right, but holder_round_* rotates okay, regardless of which candle is bound to them. Any idea what the difference would be between the round candle holder and tall one? Specifically, any difference in the collision model? Could it be that the candle CM overlaps with the holder CM in one case and not the other?

[Ishtvan]

Btw, probably the root of this problem is that idPhysics::ClipRotation seems to be returning collisions with bind-slaves even when they are set solidForTeam 0. To fix that we'd probably have to go pretty deep into the trace algorithm. We can't set the candles nonsolid either, becaus then you could turn the holder upside-down and drop the candle part through the floor up to the holder.

 

But I still don't know why it's different on these two candles. One has a CM with slightly more polies than the other one. My guess though is that perhaps on one of them, the CM for the attached candle extends a bit into the CM of the candle holder. That could make it never collide with it in a trace, because it already starts out colliding with it. I can't really tell if this is true or not though, because ingame I'd have to zoom in by a huge factor to see if the CMs are overlapping or not. It's really close to overlapping but I can't see the difference on the two holders.

[Springheel]

I can't check right now, but I'll try to remember to look tomorrow. Just give this a bump if you don't hear anything.

[Springheel]

The only obvious difference I can see is that the square holder uses common/textures/collision and the round one uses our tdm_collision_metal. Other than that, they both seem to have the same amount of clipping at the top. I'll modify the cm material of the first one and see if that does anything. My only other thought is whether the slightly concave sides cause some kind of problem, though we have lots of models with that kind of cm.

[Springheel]

I've uploaded a simpler cm that doesn't have any concave sides. See if that helps.

[Ishtvan]

Well that's better, but, wtf, this is weird, if you hold the candle so the flame end is pointing away from you, then try turning with it, the behavior changes depending on the "roll" rotation (that is, keeping the flame pointed away from you and changing which face of the CM is pointing up/down/left/right). Along one axis, it turns fine, rotate 90 degrees, it turns with you in one direction (say you turning clockwise), but if you turn the other way (counter-clockwise), it breaks and rotates away from you.

 

Is there any asymmetry in the CM if you're looking straight down at it from flame to base of holder? If there's no asymmetry in the candle part of it, then this asymmetry must be coming from the interface between candle CM and holder CM? But then why did the behavior change if you changed only the holder CM without changing how much it came up to hit the candle??

 

This is weird. There's probably a more fundamental problem that idPhysics::ClipRotation() shouldn't be detecting collisions between the bindmaster and the bound entity at all in this case, but it's still weird how this bug is manifesting itself.