[aldo_14]

I had a few (minor) problems with this (minor because it's not a very frequent event).

Basically, I had a very large bomber, with the eyepoint still set as null (centre).  The wings would thus clip straight trhough other ships, but the centre wouldn't.... so I was wondering - is the collission bounding box defined by the eyepoint in player-flow ships?

[IceFire]

I think a similar problem may have caused crashing with the TBP WhiteStar.  What we need is a new class called "GunShip".  Assumes that guns are forward like a fighter/bomber but has other attributes that are consistant with a Cruiser.

[aldo_14]

How long is the WS, though?  The ship I'm talking about is only around 80m long, IIRC.

[Galemp]

Originally posted by aldo_14
The wings would thus clip straight trhough other ships, but the centre wouldn't.... so I was wondering - is the collission bounding box defined by the eyepoint in player-flow ships?

Exactly the same problem as Karma, Black Wolf, and I are having with the X-Wing's nose and wings. I DO think it has to do with the eyepoint, since when you're in the cockpit you don't see your nose sticking out in front of you, so you collide where your eyepoint is.

[IceFire]

WhiteStar is some 200-300 meters long.  I forget what the "best" figure is.

[Solatar]

Try flying an Orion. The eyepoint is directly in the center, so you can fly halfway inside a Demon before you start colliding. Also, you have to turn off the damage indicator, all thos turrets will crasht the game if they get hurt. its kind of fun.

[Bobboau]

there was a problem with my WS,
why did you not tell me this?

[IceFire]

Originally posted by Bobboau
there was a problem with my WS,
why did you not tell me this?

I did

I think it got lost in the usual noise of doing this that and so on and so forth.

[Nuke]

i find of colision detection problems with modded ships. i think its mainly because of the cob format, which is prone to bad geometry problems. most models at one point or another must go through truespace to be texture mapped. im not saying turn away from truespace, i couldn't mod without it, but we need to find a way to fix bugs like this without changing the established ship design procedure.

[Bobboau]

poor modeling leads to bad geometry, truespace is a lot more prone to keeping geometry good than ... 3DS for instance, the only problem with TS is that if the geometry does get srewed up there isn't much you can do with it any more

the problem lies not in the geometry ediotor but the BSP data compiler (COB converter), but then PCS is vastly superior than any of the other programs, I have not had any major problems with PCS,
the white star model I made predated the current incarnation of PCS

you are useing PCS right?

[KARMA]

it happened to me also with another ship, and i noticed that sometimes changing converter solve the problem

[Nico]

Originally posted by Bobboau
poor modeling leads to bad geometry, truespace is a lot more prone to keeping geometry good than ... 3DS for instance,  

how surprising to hear that from someone who uses only TS
This statement is most likely false, I never had any pb with 3ds geometry. I don't think there's any pb with cobs either. what's screwed is the TS 3ds importer. Use deep or 3D exploration, you'll never have any pb.

[KARMA]

as far as i noticed after some tests, the problem is not related in any way to the "quality" of the geometry, it doesn't matter if you use or not booleans, intersecting polys, faces with more than 3 vertices or other odd strategies
it seems instead something related to the shape, sometimes you get better results using cob2fs2 or pcs, but sometimes you have problems anyway, like if fs2 has some problems with "non solid" objects. with "non solid" i intend a shape with a lot of parts that came out from the general body
for example an xwing has the long nose, the wings, the laser cannons, the engines thrusters, the laser cannons, i also have a similar but very very smaller problem with the tf and ta solar panels
the mass setted in pof edit also seemed to have a certain effect in reducing the problem (but may be wrong)
the speed at wich collision occurs also seems to play some roles: sometimes you detect the problem only rotating the model at zero speed, and if you engage top speed sometimes the problem disappear
another thing noticed is that sometimes when this problem happen you can even shoot through the model
the size of the model also seems to don't have effect, its true that i tried only fighters but i also tried models with size 50x obtaining exactly the same result
i already have to try instead to split the mesh in subobjects or to work with eyepoint

[Nuke]

most of my newer models were compiled under pcs. but i have models (like the first chaos and the seker) that were compiled using anchient tools (the original cob2fs2 for example) and old versions of pcs. ive also had cases where pcs refused to compile a model, so i had to use cob2fs2 instead. many of those models are well overdue for a recompile. im gonna start with the apis because it has the most noticable problems (bad turrets and a huge missing polygon under one of its wings). the original chaos is a total mess (i used the many simple objects ideology used in descent 2 models), ive given of thought to just retire it and stick with the newer model. suprisingly, the original chaos is one of my more stable models.

another bug i found was when i crash a moded ship into my ragnarok corvette, it usuly goes right through, but do the same with a herc and the colision detection is fine. likewise you take the modded fighter and run into an orion and clipping still apears ok.

wherever the problem lies, it needs to be fixed. until we understand what the cause of the problem is, we should avoid passing blame to the modelers or the mod tool programmers. we could also use an official list of modeling guidlines that are intended to prevent errors, but we must understand what causes those errors first, else create guidlines that cause bugs or impede the modeling process.

[Bobboau]

just realised something
wern't the V converter tools suposed to be put out with the sorce??

[Killfrenzy]

I dunno - where they?

If there are some converters, WE MUSH HAVE THEM!!:devil:

[KARMA]

just to make clear what i'm talking about, the problem is this:


you can't do that while flying, or, if you can, it is very difficult to
to have this weird result you have to "park" the fighter near the other one and then rotate it until the offset part of the model from the center (wings, nose...) pass through the other ship
i saw a similar effect in other cases but never so weird (and xw is a ship with a lot of offset parts), i thought that maybe it was something related to the way the model was build so i made a new model by scratch in ts5 with a sweep/point editing strategy (so no edges generated by booleans and no interceptions) but i obtained exactly the same result
for these reasons (and since i'm not the only one having this kind of trouble) it is very possible in my opinion that it is something related to how the collisions works in fs2 with "convess" (don't know the exact word) models
i tried almost evrything as said before, different converters, i changed mass, changed the size, with no (or minor)effect
i tried just today moving the eyepoint and splitting it in subobjects, again with no results

[Anaz]

Originally posted by Killfrenzy
I dunno - where they?

If there are some converters, WE MUSH HAVE THEM!!:devil:

The converters drew upon several hundred megabytes of extra information in libraries that existed since Decent 1, and had just been modified a bit for FS2. In otherwords, If you want to d/l a few gigs, then go ahead, but you'd have to talk to DaveB about it first....

[Killfrenzy]

Okay then - maybe not!:D