I would like to know why bullet simulations, even when declared as independent classes, somehow manage to cross-talk to each other, particularly in regards to the usage of ::clientresetscene() and ::initPhysics(). No amounts of books or tutorials on Bullet Physics could ever answer this question. This is a backend problem requiring the advice of a developer of the SDK. Perhaps one of you visits this portion of the forum, I don't really know.
(1) If simulations are declared independently, and then the first one is initialized, run through simulation steps, stopped, exited, and cleaned out of the memory -- then any created simulations secondary to this process will crash with an "AABB out of bounds" error.
(2) If the simulations are declared independently, and then initialized, kept-alive in memory, then only the very last simulation will operate correctly. That is, only the last simulation that received an ::initPhysics() call will operate correctly.
These weird backend problems will have enormous impact on anyone who is trying to multithread separate concurrent simulations in the same program. Not multithread the same simulation, but multithread independent ones. There does not appear to be a clear method for making them genuinely independent.
Your thoughts..?Statistics: Posted by Berezovsky — Sun Dec 28, 2014 1:13 am
]]>