Android build notes from latest git

[Slick]

I might be wrong with some of these so they are suggestions of problems:

(1)
dContainers\dBezierSpline.cpp
__debugbreak on line 516 is not supported on Android.

(2)

Code: Select all

[arm64-v8a] Compile++      : dNewton <= dNewton.cpp
S:\Newton\packages\dNewton\dNewton.cpp: In constructor 'dNewton::dNewton()':
S:\Newton\packages\dNewton\dNewton.cpp:100:127: error: cannot convert 'NewtonWorld*' to 'NewtonOnAABBOverlap {aka int (*)(const NewtonMaterial*, const NewtonBody*, const NewtonBody*, int)}' for argument '4' to 'void NewtonMaterialSetCollisionCallback(const NewtonWorld*, int, int, NewtonOnAABBOverlap, NewtonContactsProcess)'
  NewtonMaterialSetCollisionCallback (m_world, defaultMaterial, defaultMaterial, m_world, OnBodiesAABBOverlap, OnContactProcess);

At first glance it looks like a parameter is missing:

Code: Select all

// Name: NewtonMaterialSetCollisionCallback
// Set userData and the functions event handlers for the material interaction between two physics materials .
//
// Parameters:
// *const NewtonWorld* *newtonWorld - is the pointer to the Newton world.
// *int* id0 - group id0.
// *int* id1 - group id1.
// *void* *userData - user data value.
// *NewtonOnAABBOverlap* aabbOverlap - address of the event function called when the AABB of tow bodyes overlap. This parameter can be NULL.
// *NewtonContactsProcess* processCallback - address of the event function called for every contact resulting from contact calculation. This parameter can be NULL.

I see a commit from 2 days ago so this is probably in process. Maybe I should build against a zip download if the source is in flux. For now I removed the 2nd mWorld on line 100 of dNewton.cpp.

[Julio Jerez]

on the first one, yes that was a mistake I will remove it. I was tracking for a divide by zero bug that only happen in release mode. but since I added that line stop happing.
What happen is that by adding the line the compiler save the value to memory and the error goes away.
I need to figure out how to reproduce it. because I still do no see why happens.

the second yes I change the interface, the call back was setting the user data, but if I want to set a different call back the there was no way to set the user data. Since I added the custom collision call back the needs a userdata, so I decided to make set used data independent call.

I probably did not compiled the dNewton, so that bug fell trough.
I can make that quick fix and commit tonight.

[Slick]

I couldn't post again earlier but after the changes I made I did get it all to build.

I'll try tomorrow to see if it has any impact on my non callback problem in the other thread.

Thanks for the replies.