Jonne:
Thank you very much for the speedy and informative answer, it's much
appreciated. J
Not that it matters in terms of the bug, but for my own information: are you
quite sure that ConvexHull shape-types are always static? Because in an
un-edited Avatar scene, looking at the Entity-Component Editor, Entity #3,
"Fish", currently has a Shape type of ConvexHull, yet it's definitely moving
when I run into it.
I have added a bug-report (#503) to the issue tracker as you requested. I
didn't do that initially because, as a very inexperienced Tundra-user, I
don't want to risk clogging up the tracker with bugs that turn out to be
user-errors. J
Cheers,
Cale.
From: Jonne Nauha [mailto:[email protected]]
Sent: Monday, 2 July 2012 5:36 AM
To: [email protected]
Cc: [email protected]
Subject: Re: [realXtend] Physics/Collision bug in the "Avatar" scene?
Yes, setting 0 mass is correct if you don't want it to move. Also note that
using TriMesh and ConvexHull shape types cannot be pushed/have movement even
if they have mass. They are static shape types. If you want movement you
have to use some of the basic shape types like the av or the fish has there.
The bug where you go inside the object is in fact a bug and it should not
happen. I started noticing it at the time when we got the new optimized
network code in that is tied to the physics (as it optimized networking for
objects with EC_RigidBody). My hunch is that its tied somehow to the client
side interpolation/dead reckoning etc. Am not a real expert on that
particular field so, I'll let others answer better. But just wanted to say
its a bug an afaik that should not happen in a proper simulation, that would
be weird, you could help by submitting a bug report to issue tracker :) You
already have all the info there that you need to describe the bug very well.
https://github.com/realXtend/naali/issues
Let us know if you have anything else. Cool that you are giving Tundra a
spin!
Best regards,
Jonne Nauha
Adminotech developer
On Sun, Jul 1, 2012 at 4:44 PM, Cale Vinson <[email protected]>
wrote:
Hi.
I was trying to get to grips with the Tundra framework by playing around
with the very simple "Avatar" scene, and in the course of that
experimentation I observed something odd. Or at least something that seemed
odd to me. :-)
In the un-modified Avatar scene, the fish has a non-zero mass, and so if the
avatar (AV) walks into it, the fish is pushed away. Am I correct in
believing that if you wanted the fish to be a stationary obstacle (maybe a
statue or something), the way to achieve this is to set its mass to zero?
Because when I tried that, aside from the one issue below, it seemed to
have the desired effect.
Once the mass was set to zero, if I walked my AV into the fish at an angle
(say 45 degrees), the collision/physics was fine - my AV did not penetrate
the fish, and just "slid" along parallel to it. However, if I walked my AV
into the fish and carefully chose my direction so that I was hitting the
fish at right-angles (hence not sliding to the left or right), and kept the
"walk" key down, then, intermittently, my AV would start *walking backwards
away from the fish* (and to a considerable distance too) until I released
the walk key, whereupon the AV "snapped" back into the correct position next
to the fish.
Is this a recognised bug with either the Tundra core physics/collision, or
the sample avatar control/motion scripts in the "Avatar" scene? Or am I
doing something silly at my end?
Cale.
My PC specs:
Processor Intel(R) Core(TM)2 Duo CPU E8400 @ 3.00GHz Memory (RAM) 4.00 GB
Graphics NVIDIA GeForce 8800 GT Windows Vista (TM) Home Premium
Experimentation using Tundra 2.3.3.1, running both the server and viewer on
the same PC.
[email protected]
--
http://groups.google.com/group/realxtend
http://www.realxtend.org
--
http://groups.google.com/group/realxtend
http://www.realxtend.org
