[sldev] [ATTN-LL] Physics engine status update?

[Dale Mahalko]

It's now been about nine months since Andrew Linden or anyone else has
said anything official about the physics engine upgrade project.
Joints disappeared, and that was the end of it. I keep hearing rumors
that LL has Havok 4 implemented on a closed beta grid, but there is
nothing to back that up anywhere.

What's the latest info on the physics upgrade project?

Is there any continuing work to resolve the severe sim-crashing bugs
in the old Havok-I engine?

And finally, if anything is happening with a new engine on a beta
grid, is there some way for physics explorers like me to get involved
and help to stress-test it before it goes live? :-)

I submit my latest Havok-stressing project "Paddle-Whack" in Chilbo
(205,30,93) and my physics/art museum in Chilbo (137,47,98) as my
resume.


Last Thursday while intensely trying to do my Paddle-Whack physics
testing, I was crashing Chilbo about every ten minutes due to the
Havok engine collision bugs, until eventually the world map was
showing Chiblo to be red/Offline/invalid for half an hour.

I've documented one of these crashes on Google Video, showing a sim
crash within one MINUTE of starting up the physics experiment.

Google Video: Second Life Physics Engine - Tumbler Crash
http://video.google.com/videoplay?docid=1028823598159492193


I'm not intentionally trying to crash the sim, but I don't see how to
move forward without risking occasional crashes. Because the physics
engine is in the unlreleased server-side code, debugging is apparantly
out of our control. All I can do is try to learn how to step gingerly
around the major problems.

This Paddle-Whack project is so unstable that it apparantly cannot be
allowed to run unattended, because it may be causing the sim to crash
every 10 - 30 minutes all night long. Instead I've added a
user-triggerable control panel that loads a number of marbles and
knocks them around for just a few minutes to try and reduce the crash
frequency.


My partial hope here is that perhaps my continued physics experiments
are helping Andrew Linden and his team to narrow down the source of
the engine crashes.

This depends on whether the Havok-1 engine is still even being
debugged, or if it is considered "stable" and not being updated any
further, with the only fixes being like the anti-crash script-stopper
mechanism to mop up the mess after a buggy engine crash.

I don't know if its worth reporting these physics crashes to the JIRA
since they seem almost totally random, and the sim gives me no
debugging messages when it crashes. It just dies and the connection
immediately drops -- which you may happen to experience for yourself,
if you try the 5-ball setting with Paddle-Whack..

-Scalar Tardis, aka Dale Mahalko