It is a known issue. Moving the mouse or pressing keys on the keyboard will make that app gain much more CPU time. I guess is a fltk thingy.
About the timestep of the physic simulation, it is read from the file, default is 0.05 if not found  and it is not dependent of anything else. 
So the physics are faster because everything is faster.

The twisting, etc of the models ... You found a bug.

On Fri, Apr 18, 2008 at 5:47 AM, Benjamin Kloster <> wrote:
Hello again,
normally I work on a rather slow machine (one that really screams when I
load up the bandit model with its many polygons). When I installed
Gazebo on my desktop at home, that is not quite high end, but almost, I
noticed something peculiar. When I leave the simulator alone, it will
run at 30 FPS and the speed of the simulation (measured with a stop
watch and the "simulation time" in the status bar) seems about right.
But when I move the mouse over the Ogre window (the 3D window to the
simulated world), the framerate jumps as high as 300 FPS, accompanied by
a five to tenfold increase in the simulation speed. I'd also like to
remark that it seemed like the time step wasn't constant anymore,
because my robot started to twitch and jump around. I didn't test this
thoroughly, but I recall that it (the robot) only behaved weird when I
stopped moving the mouse and the framerate started dropping, not when I
built up the framerate or kept it at a high level.

I suspect the window has to be redrawn quite often when some maniacs
frantically moves his mouse cursor over the image, so I can understand
the jump in the FPS. But why the increase in simulation speed? I was
pleasantly surprised when I discovered that the timestepper just "takes
his time" when updating the world, regardless how slow the machine or
small the timestep is. I wouldn't have thought that it would misbehave
like this on fast machines.


This email is sponsored by the 2008 JavaOne(SM) Conference
Don't miss this year's exciting event. There's still time to save $100.
Use priority code J8TL2D2.;198757673;13503038;p?
Playerstage-gazebo mailing list

Jordi Polo Carres
NLP laboratory - NAIST