[Titan]

sure.......... why not try it.......... as soon as i get home................ GAAAAAAAAAAAAAAAAAAAAAAAHHHHHHH!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!  

[Getter Robo G]

Frankly, 90% of the models I've poffed over the last 4-5 years are only in game because of Modelview...

I just wonder why for ONE example out of hundreds, model hangs up PCS for 15 mins before I turn it off and in MV it's done in real time 5 seconds... It's a fighter and teh conversion works fine in game (completely tested). Does this mean MV can be more powerful in some instances? Meaning if it hits a minor error it can ignore it while PCS goes mental?

(And no, until PCS2 comes out I will not stop using modelview...)

[Bobboau]

probly cause modelview doesn't build an actual BSP tree resulting in a horrendously inefecent collision detection and geometry sorting. if you don't use PCS this 9 year veteran and pioneer of FS modding says it's not good, and just cause you don't notice it when you have half a dozen of them in a mission doesn't mean everything is fine.

[Titan]

i already hate PCS. it doesn't work. everytime i try to cob-to-pof a model it gives an error that theres like to many things in one place and that maybe i used something other then truespace bla bla bla............

[Flipside]

Do you mean it's telling you they aren't grouped or that there are too many polygons in a small area? The second problem is a common one, maybe try doubling the size of the model in Truespace and halving the conversion factor? Iirc the problem isn't with PCS, it's with the number of decimal places Truespace is accurate to or something.

[Vasudan Admiral]

Use the Wiki Luke!

The problem and normal fix is there, just as Flip described:

If PCS gives a "Too many polygons in same average location!" error, it can mean one (or both) of two things. Either there is a duplicate of one of the objects in the SCN in the EXACT same position as another of the EXACT same shape and size, or your object is too small in TS for PCS to be able to clearly tell the difference between two polygon locations during conversion. (This is a common problem for fighters and bombers because of how small they will appear in TS in order to fit the standard conversion ratio of 20 FS meters to each TS meter.)
To fix the first problem, find the duplicate object and delete or move it. To fix the second, scale the whole ship up in TrueSpace to the size you want it to appear as in Freespace - meter for meter. Then, before converting, set PCS's scaling to 1 rather than 20 and convert.
This should fix it by effectively giving PCS more accuracy when it gets to really small numbers. The up-scaling of your model is counteracted by telling PCS that one TrueSpace meter should equal one Freespace meter, so you don't end up with cruiser sized fighters and the like.

=======

And Getter Robo G: If you won't listen to me about modview's converter over on Game Warden, then please listen to Bob here? The message is exactly the same.

[Getter Robo G]

Oh I listened, I just don't actually have a CHOICE... (other than convert or not convert)... 

[Scooby_Doo]

Frankly, 90% of the models I've poffed over the last 4-5 years are only in game because of Modelview...

I just wonder why for ONE example out of hundreds, model hangs up PCS for 15 mins before I turn it off and in MV it's done in real time 5 seconds... It's a fighter and teh conversion works fine in game (completely tested). Does this mean MV can be more powerful in some instances? Meaning if it hits a minor error it can ignore it while PCS goes mental?

(And no, until PCS2 comes out I will not stop using modelview...)

You know, I might have some a possible workaround for that.  It sometimes works, but not always. 

• Find the object thats causing the lockup, mostly trial and error, remove a piece from the cob at a time
• Once you've found out which one(s) it is, go back to your original designer (max in my case)
• Throw a box around the entire rogue subobject
• Do a boolean with intersection from the box to the subobject

• This, if done correctly, will create a new object that will look exactly like the original
• Reconvert over and hopefully pcs won't complain this time

I think there might be something goofy with the original subobject mesh that doing a boolean fixes.


Also the newer pcs (the one with useable import glow points) is a lot more finicky.[/list]

[Getter Robo G]

I appreciate the well thought out procedure. 

I'll keep it in mind if I ever get any 3d people on my team. I've gotten to the point where I have messed with existing subobjects and the random sections TS makes when you use the break into subobjects button for meshes all in one piece, but that's as far as my skill level is so far. (and a hell of a improvement then 5 years ago.)

I have no illusions that I know what I'm doing...(I don't)

But as Coach Ota says: "Noriko, nothing is impossible if you have Will and Determination!"
Which helps when you're outnumbered like 100 million to 1.

[Bobboau]

let me tell you a bit of advice on looking for things that you learn when you are a compsci guy, take half out at a time, test that half if it's in there test half of that, if it isn't in there test half of the other half, keep doing this untill you only have one object left, you will find what you are looking for in log2(n) steps which means if you have a million subobjects it'll only take you about twenty tries.

[Scooby_Doo]

let me tell you a bit of advice on looking for things that you learn when you are a compsci guy, take half out at a time, test that half if it's in there test half of that, if it isn't in there test half of the other half, keep doing this untill you only have one object left, you will find what you are looking for in log2(n) steps which means if you have a million subobjects it'll only take you about twenty tries.

binary search 

[Bobboau]

[Titan]

  

getting back on subject, could someone help me make a texty for my first ubership? The name, GTF Cube, says it all.  The tech room description is " The GTF Cube, the newest in a line of ships made by LEGOtm, is the best ship the GTVA has. It will soon become the mainstay of the GTVA fleet.

[Flipside]

let me tell you a bit of advice on looking for things that you learn when you are a compsci guy, take half out at a time, test that half if it's in there test half of that, if it isn't in there test half of the other half, keep doing this untill you only have one object left, you will find what you are looking for in log2(n) steps which means if you have a million subobjects it'll only take you about twenty tries.

binary search 

Fastest way of getting rid of off-screen polies iirc

[takashi]

PCS fail to be accurate when it comes to gun positioning and such. as far as i know, PCS is just a converter.

[FlareBaffled]

PCS is a POF EDITOR, not just a converter. All the tools have benefits and drawbacks, but the general consensus is that PCS is a lot more fussy, but produces a much more complete and accurate end result. I use modelview to line stuff up, but I do all my actual number changing in PCS. Modelview is more visual, and it's easier to see where you are positioning a gun point or an engine glow.... but I don't trust it to produce the raw POF.

Most of the issues with PCS can be avoided once you know them.

But...when it comes to texturing... UV mapping in Max does it for me, with the unwrap plugin to export a blank texture to paint once I have the mapping done. Max's UV mapping tools got much more bearable with the more recent releases, and there is no need to go to the trouble of mapping in  a3rd party tool unless you really don't like the native tools. With the Max POF exporter, it's a one stop shop for me

[Scooby_Doo]

I've also noticed some other oddities with pcs such as Hull converts ok, lod 1 converts ok, hull with lod 1 locks up.   

[aldo_14]

let me tell you a bit of advice on looking for things that you learn when you are a compsci guy, take half out at a time, test that half if it's in there test half of that, if it isn't in there test half of the other half, keep doing this untill you only have one object left, you will find what you are looking for in log2(n) steps which means if you have a million subobjects it'll only take you about twenty tries.

Only in a sorted list, though.....

[Scooby_Doo]

let me tell you a bit of advice on looking for things that you learn when you are a compsci guy, take half out at a time, test that half if it's in there test half of that, if it isn't in there test half of the other half, keep doing this untill you only have one object left, you will find what you are looking for in log2(n) steps which means if you have a million subobjects it'll only take you about twenty tries.

Only in a sorted list, though.....

Technically your correct, however in this case, binary search still works. Your just not looking for one exact particular entry.
There is a problem however, multiple objects can be causing the problem,  which complicates the searching.


Oh another thing about PCS, if you get a stack overflow warning, shut down pcs and start it again.  Otherwise you'll always get stack overflow warnings even if the mesh is good.

[Bobboau]

Only in a sorted list, though.....

well actualy it works on any list you can test large groups of data cheaply, sence you are doing this search by hand useing PCS to test half the model is relitively free. sorted lists are a requierment in programs generaly cause it's the only way to make a computer tell if the item you are looking for would be in a subsection, but if there is another way to do that (perhapse haveing a machine capable of massive paralelism), then you can use a binary search without sorting.