Compound collision problems after upgrading to newer build

[Julio Jerez]

I am let it run using vs 2013 in multi thread just liek you are doing.
I can let the log going for so long the file is huge. but so far I can't make this happend

[Julio Jerez]

let us take a step back, this seem to start when I added
this #define DG_FIX_CONVEX_BUG

can you sync and go the begin of the file and comment out that define. let us see if that remove the bug. then we know is the last new code.

[Julio Jerez]

I let it running all day and when I came back the crash does happens. Now I will add soem test code so that I can reproduce it because this is was just a test run to see if It can be reproduced.

[Julio Jerez]

ok I found it, I am fixing it.

[Julio Jerez]

ok I think I found and fix it, I will let is run over night, to make sure is fixed.

[JernejL]

So far very good!! no crash after a few hours of testing, what was causing it?

[JernejL]

I said it too soon. Crashes in beamclipping function again:

Code: Select all

Unhandled exception at 0x546D19C8 (newton.dll) in TDC.exe: 0xC00002B5:  Multiple floating point traps (parameters: 0x00000000, 0x00009D33).

on line 286 - "edge = edge.Normalize();"

Code: Select all

>   newton.dll!dgCollisionConvexPolygon::BeamClipping(origin={...}, dist=0.000000000, parentMesh=0x17436200) Line 286   C++
    newton.dll!dgCollisionConvexPolygon::CalculateContactToConvexHullDescrete(world=0x17436060, parentMesh=0x17436200, proxy={...}) Line 743   C++
    newton.dll!dgWorld::CalculatePolySoupToHullContactsDescrete(proxy={...}) Line 2518   C++
    newton.dll!dgWorld::CalculateConvexToNonConvexContacts(proxy={...}) Line 2404   C++
    newton.dll!dgWorld::ConvexContacts(pair=0x1743fd9c, proxy={...}) Line 1770   C++
    newton.dll!dgWorld::CalculateContacts(pair=0x1743fd9c, threadIndex=0, ccdMode=false, intersectionTestOnly=false) Line 1900   C++
    newton.dll!dgBroadPhase::CalculatePairContacts(pair=0x1743fd9c, threadID=0) Line 1086   C++
    newton.dll!dgBroadPhase::AddPair(contact=0x146dc720, timestep, threadIndex=0) Line 1135   C++
    newton.dll!dgBroadPhase::UpdateRigidBodyContacts(descriptor=0x13f5fe18, nodePtr=0x1a5b8ea0, timeStep=6.33289843e-025, threadID=0) Line 1580   C++
    newton.dll!dgBroadPhase::UpdateRigidBodyContactKernel(context=0x13f5fe18, node=0x1539b500, threadID=0) Line 1507   C++
    newton.dll!dgThreadHive::dgWorkerThread::ConcurrentWork(threadId=1415870941) Line 242   C++
    newton.dll!dgThreadHive::dgWorkerThread::Execute(threadId=0) Line 260   C++
    newton.dll!dgThread::dgThreadSystemCallback(threadData=0x14ef2ba4) Line 166   C++
    newton.dll!std::_LaunchPad<std::_Bind<1,void *,void * (__cdecl*const)(void *),dgThread *> >::_Go() Line 187   C++
    newton.dll!_callthreadstartex() Line 376   C
    newton.dll!_threadstartex(ptd) Line 354   C
    kernel32.dll!@BaseThreadInitThunk@12()   Unknown
    ntdll.dll!__RtlUserThreadStart()   Unknown
    ntdll.dll!__RtlUserThreadStart@8()   Unknown



I will enable log and recompile.

[Julio Jerez]

I wasn't fixed yet. now it is.
please sync and try again.
The bug happened because is possible that the result of a clipped face was a edge of a point, and the funtion was not dealing with those cases. because they will be reject later on, however the newer optimization for edge handling only assumed well formed conve polygon, so and edge or a point will make fail in the normalization. even produce logic bug on very small edges.
it is fixed now. I let it run over nigh and is still going thsi morning.

[JernejL]

Crash doesn't happen but now a assert is triggered in BeamClipping:

Code: Select all

Debug Assertion Failed!

Program: J:\AppBuilder\Delfi\Top Down City\Game\Libraries\Newton.dll
File: ..\..\dgPhysics\dgCollisionConvexPolygon.cpp
Line: 285

Expression: edge.DotProduct(edge).GetScalar() > dgFloat32(0.0f)


Assert is here:


Call stack:

Code: Select all

    newton.dll!dgCollisionConvexPolygon::BeamClipping(origin={...}, dist=0.343024731, parentMesh=0x1850cfa0) Line 285   C++
    newton.dll!dgCollisionConvexPolygon::CalculateContactToConvexHullDescrete(world=0x14bbc120, parentMesh=0x1850cfa0, proxy={...}) Line 743   C++
>   newton.dll!dgWorld::CalculatePolySoupToHullContactsDescrete(proxy={...}) Line 2517   C++
    newton.dll!dgWorld::CalculateConvexToNonConvexContacts(proxy={...}) Line 2404   C++
    newton.dll!dgWorld::ConvexContacts(pair=0x1850fbdc, proxy={...}) Line 1770   C++
    newton.dll!dgWorld::CalculateContacts(pair=0x1850fbdc, threadIndex=1, ccdMode=false, intersectionTestOnly=false) Line 1892   C++
    newton.dll!dgBroadPhase::CalculatePairContacts(pair=0x1850fbdc, threadID=1) Line 1086   C++
    newton.dll!dgBroadPhase::AddPair(contact=0x14cf74c0, timestep=0.0166666675, threadIndex=1) Line 1135   C++
    newton.dll!dgBroadPhase::UpdateRigidBodyContacts(descriptor=0x1402fd90, nodePtr=0x15140580, timeStep=0.0166666675, threadID=1) Line 1580   C++
    newton.dll!dgBroadPhase::UpdateRigidBodyContactKernel(context=0x1402fd90, node=0x15140580, threadID=1) Line 1507   C++
    newton.dll!dgThreadHive::dgWorkerThread::RunNextJobInQueue(threadId=1) Line 224   C++
    newton.dll!dgThreadHive::dgWorkerThread::ConcurrentWork(threadId=1) Line 243   C++
    newton.dll!dgThreadHive::dgWorkerThread::Execute(threadId=1) Line 260   C++
    newton.dll!dgThread::dgThreadSystemCallback(threadData=0x14d0321c) Line 165   C++
    newton.dll!std::_Bind<1,void *,void * (__cdecl*const)(void *),dgThread *>::_Do_call<,0>(_Myfargs={...}, __formal={...}) Line 1149   C++
    newton.dll!std::_Bind<1,void *,void * (__cdecl*const)(void *),dgThread *>::operator()<>() Line 1138   C++
    newton.dll!std::_LaunchPad<std::_Bind<1,void *,void * (__cdecl*const)(void *),dgThread *> >::_Run(_Ln=0x01f4f3ec) Line 196   C++
    newton.dll!std::_LaunchPad<std::_Bind<1,void *,void * (__cdecl*const)(void *),dgThread *> >::_Go() Line 187   C++
    newton.dll!_Call_func(_Data=0x01f4f3ec) Line 28   C++
    newton.dll!_callthreadstartex() Line 376   C
    newton.dll!_threadstartex(ptd=0x1498a8d8) Line 359   C
    kernel32.dll!@BaseThreadInitThunk@12()   Unknown
    ntdll.dll!__RtlUserThreadStart()   Unknown
    ntdll.dll!__RtlUserThreadStart@8()   Unknown



[Julio Jerez]

that should not be possible, contact out is teh output buffer that has no changed in years.
I take this happen in debug, because is an assert.
I will let it run to see if I get it.

do I have to do anythong other than let is play?

[JernejL]

I uploaded a new build of the game, release build seems ok (so far), only in debug mode i got that assert.

https://we.tl/t-bh6CUJh9Io

I have moved all dlls to Libraries folder (they are all dynamicly linked), so to test, put your dll in there now.

[Julio Jerez]

ok downloaded the last demo, no sound on this one,
I am running it and so far no assert.
I will let is go while I go to work. Is some special I need to do to make it happens?

[JernejL]

it happens even if you just leave it running, but not in background as it pauses - if you click start server it will not go to pause so there is that

[Julio Jerez]

yes I noticed the action stop after a while when I let it run, I guess I was lucking I have it braking on the assert before.

I will click the server button tonight if it no assert happens, I am still so confuse as to while that assert happens when the pointer is the destination points for the joint to place the contacts calculate by the collision routine. I know I will not happens in release but a null pointe should generate a memory violation exception.
can you try comment opu the assert in debug and see if is crashes in the next loop? until I run with the serve option?

[JernejL]

Looks like i made a some kind of bug related to audio system, it is also the reason why game starts to be slow after a while - i will try to fix this.