| Commit message (Collapse) | Author | Files | Lines |
|
Copied BulletSNPlugin.BulletSimAPI to a new
BulletSPlugin.BSAPIXNA.cs and then modifyed the latter to
comply with the BSAPITemplate definition. Not totally debugged
but the code is all there for an INI variable to select either
unmanaged C++ Bullet or the C# version of Bullet.
|
|
|
|
to have pointers and managed to have objects.
Initial paste of XNA code. Commented out.
|
|
1. The error checking for the case where there's no "My Inventory" folder was
incorrect: it checked the wrong variable.
2. If GetSystemFolderForType() is called to get AssetType.RootFolder then
it should return the root folder immediately; not look for another root
folder below it.
|
|
A different approach to the patch in http://opensimulator.org/mantis/view.php?id=6462
that doesn't involve further forking of SmartThreadPool
|
|
pool on first use rather than constructing it ourselves.
No functional change.
|
|
BSAPITemplate.
|
|
relied on prebuilt construction info structures.
|
|
Remove all related calls from the unmanaged and BSAPITemplate interfaces.
Update DLLs and SOs to include the version without HeightMapInfo structures.
|
|
Update BulletSim DLLs and SOs with simplier step function interface.
|
|
Only initialization and debug fuctions left.
|
|
(.ToString(X)) and replace it with a method on BulletBody, BulletShape, ...
|
|
functions.
|
|
unmanaged (C++) and managed (C#).
|
|
BulletSim core to using the BulletSimAPITemplate. Physical object creation and destruction first.
|
|
instantiations for PhysBody and PhysShape when BSPhysObject is created
to account for them being classes and not structures.
Update TODO list.
|
|
by reversing the sign on the recoil computation and adding a
[XEngine]RecoilScaleFactor parameter which defaults to zero.
Testing in SL seems to show that there is not a recoil action there.
Or, at least, it is very small. If someone knows how small, the default
for the scale factor should be changed.
|
|
|
|
did the right thing for stopping (speed reducing to zero), it prevented
movement from starting (speed increasing from zero). Will revisit
when the generalize PID motor is debugged.
|
|
Looks like the viewer bombards the server with avatar orientation
information (we're talking several hundred a second) when the avatar
is being turned or when walking. This change just reduces the number
of 'set' calls into unmanaged code.
|
|
asymmetrical avatar capsule work now that rotation is being passed
from the simulator. Turns out the Bullet capsule is just not very
functional: it doesn't scale properly, the implementation only half
does asymmetry and, in general, is hard to work with.
Avatar shape is about what it was before these changes.
Added initial data structures for avatar shape mesh.
|
|
will be a no-op for ODE but enables asymmetrical avatars for BulletSim.
|
|
to implement the 'always run' feature.
|
|
by moving the movement motor to a pre-step action and out of its
questionable previous home in UpdateProperties.
|
|
Remove unused code. Add comments and TODOs.
|
|
Another instance of the underlying Bullet doing, ah, helpful things
when items are added to the world.
|
|
dynamially turn on/off detailed, unmanaged data dumping of prims and vehicles.
|
|
individual sub-classes and up to parent BSPhysObject class.
|
|
velocities of zero.
|
|
object is selected.
Update TODO list.
|
|
rebuilt linkset. I was burnt by making get/set methods with side
effects. I should know better.
|
|
like raw numbers scattered around the code.
|
|
I did last time).
|
|
|
|
step time so it will be applied completely the next step. The internal
AddForce routine does not scale the force.
|
|
object's individual gravity to the world gravity when the object
is added to the physical world.
|
|
only zeroing the movement motor in the UpdateProperties routine.
|
|
to get ready for...
Start creation of BulletAPITemplate. This defines the abstract interface
functions. Following commits will move over to the new interface.
This will enable switching between the managed and unmanaged version of
Bullet.
|
|
to the SL value.
Fixed frictin values for physical materials which were just wrong
which caused things that should have slipped to not.
|
|
Properly return root mass as mass of just the root prim rather
than the mass of the linkset. SOG has the logic to add the masses
together to get the linkset mass.
Update TODO list.
|
|
less CPU intensive.
|
|
of BulletSim. This is designed to be /as close as possible/ to the BulletSim plugin while still being entirely in the managed space to make keeping it up to date easy as possible (no thinking work). This implementation is /slower/ then the c++ version just because it's fully managed, so it's not appropriate for huge sims, but it will run small ones OK. At the moment, it supports all known features of BulletSim. Think of it like.. POS but everything works. To use this plugin, set the physics plugin to BulletSimN.
|
|
up some usages. Disable motor when done.
|
|
pre and post step event invocation routines to Trigger* to be consistant. Remove old, unused code.
|
|
|
|
|
|
updates for vehicles with per-frame action registration. One fewer special case.
|
|
over time.
|
|
problem was Bullet deactivating the object between the pushes (when, as far as the physics engine is concerned, it isn't moving).
|
|
Use same to implement setForce and setTorque so the values are restored at the beginning of each step (since Bullet zeros forces applied last step). Simplify implementation of AddForce and AddTorque by relying on the addition of forces in Bullet.
|