Controller issue AGAIN

[Raul]

It seems the problem is not solved. I worked in the last days to my map so I did notice until yesterday.

 

I have created a very simple map to test it. Only a terrain + the SDK Oildrum. I attached the full source code + exe in the rar archive. Simply running the exe you will see sometimes there is a wrong behavior with he barrels when they are hit by the player. This happens only sometimes. It's a random behavior. Like 1 from 3 tries.

 

Also I noticed I have a little freeze after starting the application. Immediately after I run the exe (1-2 seconds) I have a small pause, like 1 sec, then the app continue. I think this is wrong too.

[Marleys Ghost]

I have been trying for the last 20 mins but I can get the behaviour using your exe

[Raul]

I tested on my laptop (Intel Core 2 Duo, T5800 @ 2GHz, Nvidia 9200 M). I run it like 10 times and there was no problem. Now I am waiting a response from a friend of mine.

 

I assume there is a problem on my pc then. But what can be? I just installed the windows before buying the 2.3 B)

 

Maybe is a hardware issue?

[Raul]

I tested the application on my friend pc and the problem is reproducing: Sometimes when hitting the barrels, they are thrown away.

[wh1sp3r]

going to test it, have a free time B) hehe

 

edit: I can't reproduce your problem.

[Raul]

something I observed: After a while I used the engine (for example using the editor or even trying my code) I observed the problem does not reproduce anymore. If I do a computer restart then the problems appear again B)

[wh1sp3r]

i remember .. i had strange problems too with character long ago .. collision was ... strange, slow .. hard to explain, i rewrote my demo application .. and since this, i have not experienced any similar problem anymore, so, It's strange B)

[Raul]

I rewrote the whole application again and again. In the archive I uploaded there is also the source code. For almost 24 hours I try to understand how is this possible and no one can help me.. B)(

 

I do not understand why i have that freeze, why the collision are not working properly... why why why.. it's frustrating. :|

[macklebee]

I rewrote the whole application again and again. In the archive I uploaded there is also the source code. For almost 24 hours I try to understand how is this possible and no one can help me.. B)(

 

I do not understand why i have that freeze, why the collision are not working properly... why why why.. it's frustrating. :|

 

do you have raycasting in your program? the freeze might be caused by the first raycast pick... and apparently its a worse delay if its in debug mode...

 

as for the controller acting weird (i havent looked at your code sorry), have you tried increases the iterations in the updatecontroller function? or have you tried increasing the mass of your controller to something like 100 (which means you will need to increase the maxaccelearation of the controller update to like 300-500)? I had this problem back in 2.2X, but i havent had this issue with recent versions... at least not with controller

[Pixel Perfect]

I have experienced similar problems with collision in the past, usually where Newton has become unstable for some reason. If I have large amounts of physics bodies moving at speed (100+ during explosions) I tend to see this type of problem where even after they have settled down the physics becomes unstable and I see random small movements of dynamic objects from that point onwards and collision problems like those you describe; where colliding with objects that were previously ok now send either themselves or my controller flying off at great speed.

 

Quite why its happening in your case I'm not sure as you appear to have little happening but from looking at your code you might try one or both of the following to see if it improves it:

 

1.

Increase the acceleration value of the UpdateController command from 40 to something like 500. I found values of around 40 - 50 too low for some geometry and that caused me lots of problems in certain levels in the early days. I have known people use values up to 15000 for acceleration.

 

2.

Use the SweptCollision function on the controller and other dynamic objects, this slows the system down and produces generally more stable physics but can be too slow for large amounts of physics movement.

[Raul]

No, I do not use any raycasting..

 

I tried with the bodymass = 80 and 100 and the maxaccleration = 500 and even 15000

 

No difference. Same behavior.

 

I also tried the SweptCollision function: No difference.

[Pixel Perfect]

ok ... well you are not going mad lol. I downloaded your demo project and I can reproduce the exact same problem on my system as per spec below. The barrels fly off all over the place with exagerated force and sometimes spin about at speed.

 

The odd thing is that when run on my old P4 PC which has an ATI 3850 it behaves normally! This is looking hardware or driver related.

 

[EDIT] Sometimes this does not happen and it behaves normally. It would appear to always manifest itself when first executed after a reboot, but then may work fine for subsequent executions. However, this is not always the case. I restarted the computer, ran it once and saw the issue, stopped and restarted the demo program five times. Four of the five worked fine but the fifth time exhibited the same issues again! I really have no explanation for what is going on here.

 

It looks to me like the barrels get sucked into the controller as its gets close and are ejected generally behind the controller. I saw a barrel rolling away from me actually reverse its roll direction as I approched it and roll into the controller. Weird!

[Raul]

Yes Pixel Perfect. You describe the behavior exactly.

 

My friends pc has a dual core, 8800 gt and Win7 x86. And he can reproduce too (except the little freeze )

[Pixel Perfect]

Hopefully Josh, with his knowledge of how the controller is implemented, might be able to offer an explanation.

[Josh]

I can't produce any problems in this demo. The only error I see is the oildrums should have their step mode set to 0.

[Raul]

does not have any impact....

[Pixel Perfect]

I've rebuilt your project RVL using my 2.31 files and I can't get the executable to fail any more, unlike your original which is still exhibiting the same problem on my system.

 

I've attached the rebuilt project if you want to give it a try and see if it's fixed it on your system.

[Raul]

I do not understand whats happening.

 

First I tried to compile your code (which is same as mine of course ) and I receive the following error message:

 

Run-Time Check Failure #0 - The value of ESP was not properly saved across a function call.  This is usually a result of calling a function declared with one calling convention with a function pointer declared with a different calling convention.

in

inline TController CreateController(flt height, flt radius, flt stepheight, flt maxslope)
{
return leCreateController(height, radius, stepheight, maxslope);
}

 

Second:

I saw in your exe that the player height is lower than in my original exe... :|

 

EDIT:

Also the engine.dll file is NOT the same as mine :|

Your dll is a little bigger than mine.

[cassius]

I just ran the exe on my machine and did have the same problem. The oildrums are not seen falling over unless you pull back the controller quickly, but they should be falling forward away from the controller.

Some people may not see this as a problem but I do. As I said before a clean re update solved the problem for me after weeks of altering values to no avail.

 

EDIT: I have run your example several times and sometimes it works fine and sometimes not. Strange.

[Josh]

The first thing you should do is update to version 2.31.

 

I actually did reproduce your error after a few runs. I would like to see the project compiled with version 2.31 to see if it makes any difference, as I did recently make some fixes to the controller. It appears the force that should be applied is randomly being reversed, which causes a positive feedback resulting in the objects being "sucked into" the controller.

[Pixel Perfect]

The one I supplied is compiled against 2.31 but was compiled in Debug mode which would account for the increased exe size and the difference in the dll. I'm not sure why the height of the camera has changed as I've not changed any of your code (I did notice that).

 

What version was your's compiled against? As far as my testing has shown the problem is fixed in 2.31 but confirmation that my exe runs ok on your system would be the final proof.

[Raul]

Is there another downloader for the 2.31 version? I only have this file: "LESDK2.3.exe" which was mailed me by Josh when I bought the update. I cannot find any other NEW downloader...

 

and in the old "developer" zones there are only old SDKs

[Marleys Ghost]

http://leadwerks.com/werkspace/index.php?/files/file/107-sdk-installer-and-updater/

 

 

make sure to select the version of the SDK you want to DL when you run the above updater. (2.3/2.31/2.32beta)

[Raul]

Thanks. Downloading 2.31

[Raul]

Now with the right version of the engine (2.31) is working perfectly. I tried like 10-15 times and there was no issue. I HOPE is solved now.

 

Thanks guys for helping me. I was at the edge..