Depth sorting of meshes is buggy in 1.3.2pre 11

[fsgregs]

Hi. I have discovered some continued problems with the display of model meshes by 1.3.2. Apparently, the program's latest version confuses depth sorting of meshes.

The Shinehah nebula add-on by Rassilon has weird rotating inner nebula lines, but in addition, has a central star whose bottom half is being blocked (???) or changed by another mesh when viewed from above. Here is a screenshot.



Another problem occurs with Kokaubeam. When viewed from space, I sometimes get the surrounding nebula and sometimes not. For example, the first screenshot is of the Kokaubeam Nebula surrounded by the bigger nebula. The second screenshot is from the same spot moved outward ever so slightly. The other nebula is gone.






A third rendering bug occurs with Rassilon's rotating protostar add-on. It too has weird spinning central black lines spinning around at high speed, and a central star (and jet) that is now being blocked by something, so the bottom half is no longer visible from above. Here is a screenshot.



Lastly, I am having trouble visualizing another add-on that Jack Higgins did for me, that might be related to the problem seen here. It is the mirror and reflected beam that I use in my Terraforming of Mars Activity. When viewed from a certain angle, the beam, which used to be transparent from all angles, now conveys its transparency to the mirror itself, so I can see through both the beam and mirror, when viewed from key angles at all distances out. Here is a screenshot. Notice the mirror portion covered by the beam has gone transparent. This did not occur in 1.3.1.





I am using a GeForce4 TI 4200 card with 128 MB video RAM on a P4 machine. I just downloaded the latest driver last night but it has not fixed the problem.


According to Rassilon, this is a continuing problem with the depth sorting of model meshes in 1.3.2. I don't know how tough it is to fix, but I love these add-ons and I hope the problem can be resolved.

Regards,

I also reported this problem as a reply to another post, including screenshots. To see the comments, go here.

http://shatters.net/forum/viewtopic.php?t=5507

Frank

[Rassilon]

Thanks Frank!

I was meaning to post all this last night. I have the same issues exactly as shown above.

I believe I have the latest OpenGL 61.xx drivers.
GeForce FX 5900 Ultra XT 128 megs
Windows 2000 Pro

[selden]

Frank or Ras',

Can you provide Cel:// URLs of the failing viewpoints? And the URLs of the addons?

FWIW, the vertices of the meshes of Nebula objects never have been depth sorted. The facets always have been drawn in the order that Celestia encounters them in the model file. This can result in some rather strange effects that depend on the viewpoint.

[fsgregs]

Dear Selden:

Thanks for the input. I may be wrong about depth sorting (I am following Ras's lead here) but the effects seem to be occurring across multiple add-ons.

I can provide cel:url's, but it will take some time to provide url's for the add-ons. The ones I cited are part of big add-ons. For example, the Shinehah Nebula and Kokoabeam are part of Ras's big Gateway/Kokoabeam Fictional world add-on. Actually, I don't know where the link to it is anymore. I seached for it and can't locate it.

Likewise, Ras's protostar add-on was at one point available but I have searched Ras's website and again, can't find a link to it any longer.

The mirrors are part of a 120 MB add-on that I have available on my own website under Celestia Educational Activity 5.2 - the Terraforming of Mars. Since this problem involves only a small part of that add-on, it may or may not be easier for me to package it as a separate partial zip file. However, given what you said about loading meshes in the order they appear in the models folder, perhaps you need all the models in the folder from the entire add-on to duplicate the problem

Other folks are also suffering the same problem with 1.3.2 using similiar or identical add-ons. See the related post about Kokoabeam in the bug forum. So, maybe it is more systematic of a bigger problem.

If one has to rearrange the model meshes in the models folder so that certain ones come before (and are loaded before) other ones ... oh boy! I can foresee a real pain in the #^%^* in trying to do so and keep the names straight. Surely, this can be fixed (1.3.1 did not have this problem!!).

Anyway, I'm sure Ras can post the link to the nebula and protostar add-ons, if they are still available. If not, I'll do what I can to rename/rearrange some of the model components/ssc file designations in my copy of the add-ons and see if that does anything. Unfortunately, I have absolutely NO idea what I'm doing!

Let me know what I can do to help.


Frank

[ANDREA]

Likewise, Ras's protostar add-on was at one point available but I have searched Ras's website and again, can't find a link to it any longer. Anyway, I'm sure Ras can post the link to the nebula and protostar add-ons, if they are still available.

Hello Frank, I've both protostar.zip (8729 Kb) and h-alpha.zip (3249 Kb).
If needed, I can contact Motherlode in order to upload there the files or, if your browser allows for those dimensions, I can send you them by e-mail.
Inform me, please.
By

Andrea

[fsgregs]

OK, I've tried changing things in one of the problem add-ons (the mirrors). I reordered the model meshes, changed their names, edited the ssc file, all to no avail. The semitransparent beam still blocks the main part of the mirror (as if the mirror wasn't there any longer). This may also be what is happening in the other add-ons, since each of the problems involves a semi-transparent mesh in front of another mesh.

I also tried separating the two models (the beam and mirror are touching in the original add-on). That did not do anything constructive.

Andrea, Ras may already have a link to his protostar add-on. I suggest we wait for him to respond before you upload his add-on to Motherlode.

Frank

[ANDREA]

Andrea, Ras may already have a link to his protostar add-on. I suggest we wait for him to respond before you upload his add-on to Motherlode. Frank

OK Frank, waiting for Rassilon's info.
By

Andrea

[selden]

Frank,

I'm sorry: it seems you misunderstood my "FWIW" comment.

The problem I was trying to describe only affects Nebula models defined within DSC files, not object models defined within SSC files. The facets within an individual Nebula model are not depth sorted, so sometimes parts of a model that should be drawn first, and thus should be partially hidden by nearby facets, are drawn last, in front of facets close to the observer.

Most of the problems you are describing are with SSC models.

[Rassilon]

Andrea, Ras may already have a link to his protostar add-on. I suggest we wait for him to respond before you upload his add-on to Motherlode. Frank

OK Frank, waiting for Rassilon's info.
By

Andrea

thats fine by me...


Selden,

From what I was aware of I believe Chris had started messing with depth sorting in the recent versions of Celestia for all models...I believe this entails depth sorting by way of ray casting vs loading procedure...but Im not positive...Since the way models are rendered have changed I believe its safe to assume its the way Chris patched up the rendering of models in the most recent versions, but what makes me second guess is the area of effect isnt alone with 3ds models but also with stars which are software rendered. The clipping of the central star raises a few questions weither its depth or rendering...

[selden]

Ras',

As best I can tell, although there seem to be some minor changes in how DSC meshes are handled, the depth sorting done for SSC models simply is not done for DSC models, even when building Celestia from the latest CVS source code. The same model acts quite differently depending on whether it is used as an SSC model or as a DSC model. That's why the bug descriptions have to be clear as to which type of catalog a model is being used with.

Obviously there are still bugs in the new model code.

[ANDREA]

Andrea, Ras may already have a link to his protostar add-on. I suggest we wait for him to respond before you upload his add-on to Motherlode. Frank

OK Frank, waiting for Rassilon's info.
By
Andrea

thats fine by me

OK, I've just uploaded to Motherlode both h-alpha and protostar .zip files.
Don't know how long will take to have them available, anyhow, it's holiday time!
By
Andrea

(added later): thanks to Don, both files are now available in Motherlode creators Rassilon's folder.
By

Andrea

[chris]

I think that I have a fix for this bug. It's very simple and I can easily slip it into 1.3.2. But, I can't seem to find the Kokaubeam add-on in order to test my change. Can someone provide a link?

--Chris

[ANDREA]

... But, I can't seem to find the Kokaubeam add-on in order to test my change. Can someone provide a link? --Chris

Chris, you can find it here:
http://sourceforge.net/project/showfiles.php?group_id=96264
in Rassilon's page.
By
Andrea

[Rassilon]

For future reference in case no one was aware all my addons will from now on be available at sourceforge unless its a very small file I will post on the addons section of my site. If you go to sourceforge and do a search on celestia I almost always come up...

Plus the links to my sourceforge page is in my sig...

[chris]

Thanks for the report and links to the add-ons. I verified that my change fixes the artifacts in the protostar and Kokaubeam add-ons. It's checked into CVS and it'll be in 1.3.2 final. I did not test the mirror add-on, but it's probably working now as well.

--Chris

[fsgregs]

Chris, that's great. Thank you.

I don't have the url to the updated exe files in CVS. Is the fix available for download at 1.3.2pre12, or should we wait until 1.3.2final? If it is available, what is the url address to the CVS Celestia exe page?

Frank

[Rassilon]

Thanks again Chris, your efforts are appreciated...