div / 0 in BuildJacobianMatrix

[Julio Jerez]

Thanks I'll take a look into this, but I will reach the same problems when I'll try to update the joints.

on what? the newer OgreNewt?

[arkeon]

I have updated my joint code to use an inverse timestep like you've done in other joints and avoid the / 0.
this ones works well now.

I still have a crash on dgVector truncated (_mm_cvtepi32_ps (_mm_cvttps_epi32 (m_type))); in dgVector Floor.
It is on very specific scene only, it happen when my controlled body (first person camera made with 2 collision scaled spheres and a slider joint) hit a body which use continuous collisions. When it happen the newton matrix can return nand numbers and my camera goes far away ^^

I'll try to figure it out (i know i'm obstinate but I would prefer doing mine own wrapper)
I will make a video tomorrow.

[Julio Jerez]

do you have a test demo that I can test?

[arkeon]

I can make one.

[Julio Jerez]

please make one that use the dlls, so that I can see why is wrong

[arkeon]

Doh !
I build all in static :/ and modified the cmake to just build the needed libs with correct install and preprocessors.

you need the network debugger right ? I didn't add this one because of it's dependencies.

I can build a non static version if you want.

I testing my scenes for an hour and could find one that make it crash.
The scene with crash or strange behavior is at work, I'll send it tomorrow if I don't get on failing here.

[Julio Jerez]

Not the network debug does not work. you nee to build using the dlls so that I can attach the visual studio debugger

[arkeon]

ok so the dll in debug mode
I'm adding an option to cmake files for static. You could get them if you want.

do you need the DGASSERT enable ? I've remove them since there is a lot ASSERT(0) in the code.

[Julio Jerez]

the assert are important yes

[arkeon]

I've updated the cmake files with a build static and disable assert option
you can find the 3 cmake files here : https://svn.scolring.org/trunk/dependen ... -dynamics/

the cmake files on you git repository are not working on windows you should consider updating them with this ones.

I will send you the sample tomorrow.

[Julio Jerez]

I do not know what I am looking at here, this is a subset of the newton files, I did not see any demo.

[d.l.i.w]

I'll try to create a diff to the existing CMake files and issue a pull request on github...

[arkeon]

yes I've only put libraries on my svn

The demo part is just commented in the cmake file.

[Julio Jerez]

when I uncomment that last line:
add_subdirectory("${NewtonSDK_SOURCE_DIR}/applications/demosSandbox")

CMake Error at CMakeLists.txt:92 (add_subdirectory):
add_subdirectory given source
"D:/newton-dynamics/applications/demosSandbox" which is not an existing
directory.

I do not understand why you make this so complicated.
what demo would I be looking at, because what I see is a copy of Newton SDK,
and the path is a path top the sandbox demos.

why will this be different than why the Newton SDK that is in GitHub?
are you making changes to the SDK internals?

I compared you copy to the current copy in GitHub and I see you have and older version that misses a significant number of updates, but all those change are relate to collision improvement and bug fixes that were already working, so I do not see how this will be different.

[arkeon]

Ho yes I didn't update the demossandbox cmake file

I will send you a compiled demo.
there is no demo in here.

It is pretty complicated because this is a language development with scripts.

We have the Scol virtual machine, then the 3D engine as a virtual machine plugin which contains Ogre and newton. then the application is written in scol language.