Graphical glitch in Linux/Windows, different glitch with Mac

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

Graphical glitch in Linux/Windows, different glitch with Mac

wspear
Greetings. I'm converting an application from a very early version of jogl to the latest build (12 December) and it's mostly gone smoothly but I've run into a couple weird problems. I would like to see if  these are known issues / potential user error before I try to issue a bug report (it's a reasonably complicated application and extracting a succinct example will take some time).

Basically, we're rotating charts. From one side the charts look normal:
Good Angle

From the other side some objects don't seem to be properly opaque to the other objects. We can see all of an object that should be hidden by the object in front o fit. I am a rank newbie at jogl programming and don't have any intuition about what might cause this problem:
Bad Angle

This has been observed with java 6 and 7 on 64 bit Linux and 64 bit windows. This doesn't happen on 64 bit mac, but on that platform resizing the window doesn't resize 3d frame. If the window is shrunk below the size of the original 3d frame it exhibits weird stretching artifacts. That is presumably an unrelated problem.

Any ideas or prior knowledge on either of these issues would be much appreciated. Otherwise, confirmation that I should be pulling out some code to stick in a bug report (as opposed to some kind of 'lite' bug report) would be helpful.

Thanks,
Wyatt
Reply | Threaded
Open this post in threaded view
|

Re: Graphical glitch in Linux/Windows, different glitch with Mac

jmaasing
wspear wrote
Basically, we're rotating charts. From one side the charts look normal:
...
From the other side some objects don't seem to be properly opaque to the other objects. We can see all of an object that should be hidden by the object in front o fit. I am a rank newbie at jogl programming and don't have any intuition about what
...
That almost looks like the boxes are drawn in the same order in both pictures but without depth testing. Strange though that it doesn't appear on Macs, maybe different driver defaults or different graphics configurations. Hard to say without knowing more.
Reply | Threaded
Open this post in threaded view
|

Re: Graphical glitch in Linux/Windows, different glitch with Mac

Wade Walker
Administrator
jmaasing wrote
wspear wrote
Basically, we're rotating charts. From one side the charts look normal:
...
From the other side some objects don't seem to be properly opaque to the other objects. We can see all of an object that should be hidden by the object in front o fit. I am a rank newbie at jogl programming and don't have any intuition about what
...
That almost looks like the boxes are drawn in the same order in both pictures but without depth testing. Strange though that it doesn't appear on Macs, maybe different driver defaults or different graphics configurations. Hard to say without knowing more.
Yeah, it looks like they turned off z-buffering (or just didn't turn it on), and manually rendered in "back to front" order (i.e. the painter's algorithm). So then when you rotate the graph, that draw order is wrong. Maybe the new JOGL doesn't turn on z-buffering by default and the old one did?

Reply | Threaded
Open this post in threaded view
|

Re: Graphical glitch in Linux/Windows, different glitch with Mac

wspear
In reply to this post by wspear
I've posted a bug with a tarball of source to demonstrate the issue here: https://jogamp.org/bugzilla/show_bug.cgi?id=936
Reply | Threaded
Open this post in threaded view
|

Re: Graphical glitch in Linux/Windows, different glitch with Mac

jmaasing
Just a friendly tip, if you want more people to try and help I suggest you make a test case demonstrating the particular glitch. Ideally that should be a single "main" class with only the depth sorting issue demonstrated.
The source you attached is like your entire application with no explanation of what the classes do so fewer people will want to take the time to dig in.
Reply | Threaded
Open this post in threaded view
|

Re: Graphical glitch in Linux/Windows, different glitch with Mac

wspear
Thanks. I've updated the bug with reduced demonstration code. There are still three source files but I believe most of the extraneous stuff is gone now. I had hoped the issue might be with some initialization boilerplate that would be easy to spot for someone better versed in JOGL. I took a look at the depth buffering/buffer-clearing stuff in particular but so far it seems to line up pretty well with what demo code I've seen.