I had a problem with my model going through an invisible wall when it shouldn't have. The same wall with INVISIBLE not checked prevented my model from going through the wall. I will describe the process I went through to figure out why as it may help others in finding other problems that they may come across.

Model used: Raven's spaceship
http://www.xecu.net/raven/Dload/Viper.zip

1. When you change anything in a computer application there is always the possibility of something getting corrupted for one reason or another. I had been working with this level for some time so there was a good chance that something in the level was causing a problem. So with this in mind I removed everything from the level except my model and one prefab.

Result = Wall penetration.

2. I went to the OFFICE level and made a small building invisible. There was always the possibility that I was doing something incorrect.

Result = No wall penetration.

3. I brought the building into my level (Copy & paste).

Result = Wall penetration.

4. Since scripting can change many things in your level I loaded up the FLYLEVEL as it used a similar script as I do. I pasted the building into this level.

Result = No wall penetration.

5. I took the model from this level and substituted it for the model in my level.

Result = No wall penetration.

That told me that maybe the problem is the fault of the model.

6. I scaled the model up 5 times.

Result = No wall penetration.

7. I changed the scale many times to narrow down the point to where the model would and would not penetrate the invisible wall.

At a scale of:

1.091 = Wall penetration.
1.092 = No wall penetration.

The model was 30 quants high at this point and since I did not know what it was about the model I decided to use a primitive cube from MED.
I re-scaled the cube until I found the transition point where the model could and could not penetrate the wall.

Scale factor:

1.667 = Wall penetration.
1.668 = No wall penetration.

Result = The cube was 53 quants on each side.

8. I went back and measured the width of the model in my level and it was also 53 quants wide. So it was the width and not the height of the model that was determining wether or not there was penetration of the invisible wall. I was looking for something that had a relationship to the sizes of the Bounding Box, the Hull size (Fat & narrow) or the Point size listed in the manual (32, 16, 8). I would not call this a bug but an unknown limitation.

There seems to be a fractional point in the 53.xxx quant range where the actual transitional point is located.

This is what I found and it would be a good idea if a few other people could do some testing to see if they come up with the same conclusion. I would suggest first to use the same model that I did (Raven's spaceship) and then use another model of their own as it could actually be something else and not the width that is the determining factor.

Maybe Doug and JCL could give some input regarding what I seem to have found. Then again I could be all wet.

Ron

If a model is too small, it doesn't get a hull by default - it is assumed to be something like a bullet. Models without hull penetrate invisible blocks like thin air.

The easy solution is to set the narrow hull for that model.

Yes JCL you are so right and those facts are stated in the manual on page 82. But with a model that is 50 quants wide the narrow hull is automaticly set for the model. The model only gets no hull asigned if it is smaller than 8 quants. So if the model is 53 quants wide and 30 quants high it should have the narrow hull asigned and should not penetrate an inviisible wall.

If you take the large primitive cube in MED and bring it into WED and do the scaling that I mentioned below you will have the same problem. Both models are larger than 8 quants and therefore have the narrow hull asigned.

Please correct me if I am in error. If I am not in error then please explain in detail why the penetration of an invisible wall with a model that is 53 quants square. The model size measurement is taken directly from the view windows in WED.

Ron

[This message has been edited by Ron Holmes (edited 05 March 2001).]

Just a dumb idea but I figured I'd throw it out there. Have you tried making a very small model in MED and bringing it into WED? Conversely, have you tried making a big model in MED and bringing it into WED and using its scale property to make it small? I'm wondering if one of the two methods above have different effects on the engine or if it really amounts to the same thing as far as A5 is concerned. Hmmm.

No I have not done that Dave but it is something that I could try but I expect the results to be the same. The building that I took from the OFFICE level was scaled up and scaled down to see if there was a correlation of model size to invisible wall size. The results were the same. As long as the wall was not selected invisible the model did not go through the wall but only when the invisible flag was selected.

Tomorrow I will try and scale up the cube in MED so it is necessary to scale it down inWED and see what happens. Or I could try and find a model that is larger than the 53 quant transition point.

I would rather do that because it isolates the possibility that something I could be doing is what is giving me these results. And the model by a third party is also availble to everyone that would like to do their own testing to independantly verify what I say is happening.

Since others on the forum have had the same experience with models going through invisible walls I thought it was important in what I had come across. Or maybe we are all just doing something identical that is causing the problem. I do hope that JCL comes back with some specifics as I can not tie what is happening to anything other than the model size at 53 quants.

It does not matter if I am doing something incorrectly or if there is a bug or limitation in the engine, script, etc. The important thing is to find out why I am getting the results that I am and then working around it so as not to have a model go through a wall when everything else says that it should not.

Ron

I just scaled a cube up in size in MED and broght it into WED. I zeroed in on the transition point and then took a measurement. The size of the resulting cube was again at 53 quants. So the direction of scaling is not important.

Ron

Here's another idea. What about not checking invisible but instead turn each face to "none" in the faces section?

I will try that when I get home from work tonight.

ron

I thought that I had tried that before but went through the motions anyway. With no texture assigned (None) the default stone texture seems to be applied anyway. With the invisible not checked the ship went right through the wall.

Now if you stop and think about this invisible wall problem you could turn it into something useful. I would still like a little more input from JCL and Doug concerning what I have found as I am sure they could give us some more information that we could use to actually exploit this.

But to confirm that there is nothing in my system that is causing a fluke we need a few people to run some tests of their own. I am using version 4.23 with all of the updates to WED, MED and the Templates. Do we have a couple of volunteers?

Ron