Hi,

thanks for your reply. Yes we're calling the gl-function from within the GLEventListener-display method, and we're using JOGLAppletLauncher with -Dsun.java2d.noddraw=true set in the plugin parameter field. The strange thing is, on a different test computer here at the office with pretty much identical spec, same graphics card with the same drivers, the bug does not happen.

We do have a fallback to software rendering with gl 1.1, which kicks in when the gl-function call fails, but in this case, when the canvas has been initialized "offscreen", all that gets rendered is a black applet-window. Software rendering works great otherwise and is how we worked around this issue: http://www.java-gaming.org/forums/index.php?topic=16324.0

The second issue, the Opera crash only happens with ATI, but we get browser freezes on nvidia computers as well on occasion when using a GLJPanel.

Hi,

on some client computers, gl.isExtensionAvailable( "GL_VERSION_1_3" ); returns false when the applet and the GLCanvas is created outside the browsers viewport, i.e. when the user has scrolled down the page and the applet is created on the top of the page. If the applet is visible during the initialization, the method returns true. This happens on one test computer here at the office, tested with ie6 and firefox, java 1.5 and java1.6, GeForce 6600 LE latest drivers, but also on a few beta testers computers.

If using GLJPanel instead of a GLCanvas, we have seen random browser freeze when the user is navigating from the page containing the applet. In Opera, if the computer has an ATI-graphics card, a hard crash happens every time the user navigates from the page and a hs_err_pid log file is created. I'm attaching the log-file.

Any help is welcome

We have also seen this bug. We did a work around and used a GLJPanel instead, which does not have the repaint problem. Of course, a GLJPanel has slower performance, but in our project performance is not cruicial.

Johan
Agency 9
http://www.agency9.se

Its definitely the -Xmx flag, tested on two different computers with intel graphics. Removing the flag results in hardware accelerated context.

If we were to write our own plugin, it would not be as extensive as suns, it would only be able to run our special java class, and nothing else. If it could solve the problem, it would be worth the development time.
But if the problem is in the jvm.dll, I guess we would have the same issue anyway.

Hi Ken,

thank you for your reply.
Unfortunately a web start deployment strategy isn't suitable for our project, since we have to be embedded in a web page. We're trying to figure out possible work-arounds.

Is this bug only affecting jogl-applets or applets in general? Could we switch to D3D-rendering or software-rendering?
Would it be possible to write our own java plugin to IE, i.e. an activex plugin using jni to get our java project running in a browser, or would we face the same bug?

Regards,

Johan Göransson
Agency9

Hi,

we're using JOGL Applet Launcher to deploy an JOGL-project as an applet. Most clients have no problem running the applet, but some computers with Mobile Intel graphics cards especially have some problems in Internet Explorer when -Xmx is set. I tested the applet on an Mobile Intel 915/GMS, 910GML Express Chipset Family, with over 700 MB RAM, and had -Xmx128m specified. The result is that when creating a GLCanvas, we only get a software GDI canvas, and not an hardware accelerated canvas! This is only in Internet Explorer, the same settings works in Firefox and also stand alone on the desktop.

Is this an known issue?

Regards,
Johan Göransson
Agency9

Hi, Im just curious how you will use JOODE/ODE in you PhD studies? Whats the research topic/area?

yeah, dCollide would probably work, but then I'd have to collide the ray with all static geoms my self. And I never got dCollide to work with more contacts than one.

The point with ray - geom collision would be to implement ray cars for instance - shooting rays towards the static triangle mesh instead of using jointed spheres for wheels. Or, if a car has flipped over, you could shoot a ray from the car to detect where to drop the car again.

Hi,

has anyone got geomray to generate contacts with other (static) geoms? I got it working for geoms added to a body, but no contacts are generated for enviroment geoms.

EDIT:
Well, think I found the reason. In nearCallBack in odejava.cpp:
---
  // ignore if both o1 and o2 are plain Geoms (not bodies)
  if (!b1 && !b2) return;

---

Work around? Should I try the tedious path of recompiling odejava.dll?

Ok, so perhaps I should ask what you guys think how big a big packet is? I'm looking to send about 30 packets/s if they are big (250 bytes) or many more, perhaps 80 packets/s if they are small (100 bytes). The network topology is LAN or broadband internet.

Well, I will disregard out of order packets, but I've read somewhere that big packets are lost more often than small packets, but I dont know if there's any truth in that.

Hi,

I'm currently working on a network game using UDP, and I'm curious whats better, to send many small packets or to send fewer but larger packets in terms of packet loss and overall performance? Any thoughts?

The smart thing to do might be a java bind (jni) to Novodex, Havok or Meqon. Coding a physics engine is a major undertaking. Developing a bind to an existing also brings the benefit of being able to use their exporters and tools. ODE is a great physics engine, but since its not commercial it is not being developed as much as the commercial ones.

Well, ragdolls are mostly used for death scenes in games (hence the term "ragdoll" ), are they not?
It just seems so hard to tweak angular motors to get the desired result.

I used ODE and AgentFX when creating ragdoll demos for Agency9, (http://www.agency9.se/demo/index.php?t=2 http://www.agency9.se/demo/index.php?t=1). I'm stepping the world using quickStep with a small time step, 0.01, and it works like a charm. The ragdolls are built using 12 bodies, 11 joints with no motor parameters. Controlling the joints with motors to get realistic movement sounds  complicated, it seems better to use animations for walking and such.

I'm setting the desired velocity parameter on the hinge2 (2-axis) as well as max force on that axis. It's my understanding that ODE will accel up to that velocity using the max force, so its not like I'm manually setting the velocity of the wheel. I've so far used the default ERP (0.8?). I'm playing around with FDir and slip parameters, but I haven't really done anything useful yet.

well, I've modified the car example, but came across the "12.9. I've made a car, but the wheels don't stay on properly!"-problem as noted in the ODE-docs. The wheels are unstable at high speeds. I can't seem to solve this, using the method described in the docs. Anyone else had this problem?

Space-bug perhaps? Have you tried collisiontesting not using spaces or used a simple space?

Is there a changelog, for both odejava and for the natives?

Regards

Nice demos,

actually, you used the same car model as I did when I modified the car example to try to get something moving. 
Small world when it comes to free 3d models that are not completely crap

thx for the reply,

are hinge2 as fast and as stable as sliders and "whatnot"?
Chassis could perhaps be a box with capped cylinders as bumpers.

Hi,

I'm interested to know if any of you guys have anything cool to show, done with odejava or physics in general? Has anybody actually done something useful, like a game or application showing of a good use of physics, or is it just box stacking demos? 

Hi,

I'm about to try and create a car in ode with a good arcade feel to it, and I'm interested to hear if you guys have any good ideas. The goal is to create something with a nfs-underground feel, with suspension, gears and a handbrake.

What type of joints would you use, hinge2 with suspension or a slider joint with cfm and erp set or perhaps using raycasting? How about gears, using a AMotor or setting torques/velocities manually to the wheel bodies?
Spheres for wheels or capped cylinders? A simple box for chassis, or multiple geoms?

I've looked at the car example, but it is quite limited I think.

I enjoyed ep3, as opposed to the previous 2 which I hated. The thing you've got to remember is, that Star Wars is intended for 12-year olds, which explains entities such as jar-jar and all the funny sounds coming from droids and creatures. Did the wookie actually do a Tarzan-sound when attacking the boat-thing?

And I really liked the explosions, since I'm tweaking a particle system right now, trying to create a good looking explosion. I kept thinking, "how would I go about to create that explosion?"

well the method I described works quite well since ode will automaticly enable any disabled box if it collides. Since it is so easy to implement I would give it a shot before dusting of anything

I'm not sure, but I dont think autodisable is implemented in odejava just yet. You could however implement a quick hack by checking if the linear and angular velocity of an object is below a certain limit and disable it yourself with Ode.dBodyDisable( bodyId ).

From the ODE-docs: "When assigning a mass to a rigid body, the center of mass must be (0,0,0) relative to the body's position. But in fact this limitation has been in ODE from the start, so we can now regard it as a ``feature'' "

So.. I'm guessing, if its regarded as a feature its not likely to get fixed in the near future.

But I'm also trying to build a car and it is difficult to get it stable since the CM is way to high. Placing the geomtry lower using a geomtransform is not a good solution, since the geoms are also used for collision testing and the actual 3D-model the geoms are representing, is placed higher. One solution would be to have a displaced geom representing the mass of the car, and another geom for collisiontesting which I would set the position and orientation according to the mass-geometry minus the displacement. But this feels like a bad solution Any other ideas?

Has anyone implemented a arcade-feel car using odejava? I'd like some examples other than the cars in the test-package.

I say submit!
I only had native crashes on one or two meshes, but most work great.

okey, well, have you tried the log4j-webpage? I have no experience in running it on MacOSX (and hardly any on win either), so I can't help you :/