Kinematic and dynamic body: interaction?

[Julio Jerez]

sorry, I am refactoring the VS 2008 and 2010 project so that it is easier to make the 2012.
try again please,

yes try to get VS 2012 express, let us see if this give more information,
I am making a new folder for VS 2012 solution and projects.

[Enclave]

Ok, i will download and try VS2012.

I don't know how soon i can do this, because a have some problems with internet connection..

EDIT:

Julio, do you know any debuggers/tracers, which works with visual studio and can help us?

[Enclave]

Julio, i am install VS2012 (full), but the project cannot compile.

C:\Program Files (x86)\MSBuild\Microsoft.Cpp\v4.0\V110\Microsoft.Cpp.Platform.targets(42,5): error MSB8020: The builds tools for Visual Studio 2010 (Platform Toolset = 'v100') cannot be found. To build using the v100 build tools, either click the Project menu or right-click the solution, and then select "Update VC++ Projects...". Install Visual Studio 2010 to build using the Visual Studio 2010 build tools.

[Julio Jerez]

Enclave do you mean VS 2012?

Ok guys I just check it legitimate VS2012 solution and projects that do not screw up the existing VS 2010 solutions.
You should delete your Newton dynamics folder and check out form SVN, the reason for that is that if you have and older PDB.90 for some reason in my systems VS 2012 linker fail.
The projects and solution are in folder pre-pended by .. /2012/..
I also check in a 32 bit exe of the sandbox build with 2012 in release mode.
In my system this run like a Cham, no warnings, no error, no crashes, nothing at all, the debug exe also run without problems.
Please try this exe to see if at least the compiled exes that I make works.

[JoeJ]

You mean VS 2012 - not 2008? Did you accidently open 2010 project instead 2012?

I open the 2012 express project, then some projects get updated from older version.
Then i build glew and get error: string literal too long. I change that to: VALUE "Comments", "xxx\0"

Then i build sandbox - no build errors - and run it, i get assert there:

void dgDynamicBody::SetMassMatrix(dgFloat32 mass, dgFloat32 Ixx, dgFloat32 Iyy, dgFloat32 Izz)
dgAssert (Ixx > dgFloat32 (0.0f));

I got the same assert previously when i skiped all stack corruption alerts.


@Julio: You don't get the glew error?

There's VC update pending. i'll do that now - maybe it makes a difference...

EDIT: I've overwritten demosSandbox_d.exe exe, but downloaded demosSandbox.exe crashes.
The 64 bit exe starts up now. It shows skymap and some totally black objects.
Turning off visible objects and turning on collision mesh / contacts / forces shows nothing.

[Enclave]

Yes, this was type mismath))

Now it compiles ok, but when i run it, i get the same assert. I have not received any additional information about debug from VS2012.

EDIT:

JoeJ

Then i build sandbox - no build errors - and run it, i get assert there:

I get the same error! Line 204.

The 64 bit exe starts up now. It shows skymap and some totally black objects.
Turning off visible objects and turning on collision mesh / contacts / forces shows nothing.

This happens because the mesh data is not loading/corrupted. See my screenshots, that shows corrupted data.

We must track/trace the mesh loading!!!

[JoeJ]

Someone knows how to do VS update?
Usually there is a tray icon telling me it's available - but i gnored it and i can't find option to do it manually


... if that does not help i'll see if i can debug with 2008 and 2012 at the same time. That should make it possible to find the difference.

[Enclave]

JoeJ

Usually i simple download a service pack and install it. That's all.

[Julio Jerez]

hwo codul it be that teh mesh is corrupetd I have tow very diffrent machine at home, plus my system at work, plus the Mac.
and in all those systems all the demos run without any error.

Joe now I believe we are runninmg the same build. what system do you have
what system do you have? I do no undernat what is wrong.

as for third party debugger for visual studio, I do no know of any, I do not think there are any because no one will make a debugger to compete with VS integrated debugger which is very good.

[JoeJ]

i have i7 920, 12 GB ram, gtx670
Win7 x64 ultimate service pack 1 german

vs update takes a while...
Then if it's possible to debug with both VS at once on one machine i may find something out...

[Julio Jerez]

that very much the same system than I have at work, at home I have a lower capacity win 7 proffetional and Gforce 450 so something that .
how could it be that consistently fail for you and no for me.
I did have the stack fail yestrday but th was becaus eteh prlject was all messup, after I fixed then do not have any more errors.

I nee to see if I cna test it on a system that can reproduce the crash.
you would think that if this was a problem with the code or the data, then it will also fail on the Mac. but so far I know it does not.

[JoeJ]

i would not worry too much about it - i think its some stupid little thing, like some ms library makes tinyxml behave different on german windows.
Or a buffer overrun.

I'll dare to try double debug while update is running - takes ages.

[Enclave]

if this was a problem with the code or the data,

I think a some code corrupt our data while loading the mesh. May be it is a buffer overrun, i don't know.
Can you explain a little, how the mesh is loading? Which places i must set a breakpoints.

Julio, did you seen a my last screenshots? It shows that normal data is corrupted, but i don't know when this happens.
I look inside to hangingBridge.ngd file, it is a XML format.

I want to trace and look how it perform loading the data (normals).

i think its some stupid little thing, like some ms library makes tinyxml behave different on german windows.
Or a buffer overrun.

This is possible and we need to find it

[JoeJ]

Oh * - i get the same assert with 2008 too. If i do a release build i get the black scene.
I always was thinking 2008 must be ok because ther never where those false stack alerts.

I'm a lot less optimistic now. If you can't find a machine to reproduce i'd say keep a copy of the last fox build

Edit: No news from the vs update.
I modified the box stacks demo to do nothing else than loading box.ngd.
But even with that simple scene it's hard to see anything in the large tree generated for the scene.
I'd like to compare vertex data with xml file, but so far i see only a pointer to the first float - too less knowledge to follow the flow of data from xml to tree.

It seems the only ngd file that loads properly is the flatPlane.ngd.

[Julio Jerez]

the simplest scene is the flatPlane when I get home I will hack a demo to just load that and chenk it in to see if it repordece teh bug.

then I will add some trace to the scene loader to see if whe findd the place were if goes wrong.