Very curious crash in Celstia 1.3.2pre4/5

jim · Post #1by **jim** » 03.03.2004, 18:36

Hi Chris,

I've detected a small bud very curious crash bug in Celestia 1.3.2 pre4 and pre5 only. After the start I choose Saturn's moon Pan and goto this one. At the distance where the model 'asteroid.cms' should be loaded Celestia is crashing.

First I thought it's a problem with the 'cms' model but other bodies that use the same model and texture work fine. Than I found a trick: if I go first to Saturn's moon Atlas and than to Pan no crash happens. But only this and no other combination works - really very strange!

Now I've tested other models with Pan. The crash happens only if Pan use a 'cms'-model ('asteroid.cms', 'roughsphere.cms') while with '3ds'- and 'cmod'-models all works fine.

Now I guess this bug is not really a problem because 'cms' model will be no longer relevant for Celestia.
Chris, can you build 'cmod'-models to replace all 'cms'-model?

Bye Jens

--------
My sytem: Win98SE, Duron900, 512mb ddr ram, Geforce3 TI200 64mb, 1024x768x32 (no AA), Nvidea driver 52.16

Post #2by **granthutchison** » 03.03.2004, 19:51

No such problem with XP Pro 2002 SP1, GeForce4 Ti 4600.

Grant

Post #3by **selden** » 03.03.2004, 20:26

Jens,

You need to update your drivers to 53.03.

Post #4by **granthutchison** » 03.03.2004, 20:54

OOPS.
But I can precisely reproduce it with a 4k ring texture!
Jens, what happens if you use 1k?

Grant

jim · Post #5by **jim** » 03.03.2004, 23:35

selden wrote:Jens,

You need to update your drivers to 53.03.

There are any know bugs with my driver version? It works on my system so far without any problems.

granthutchison wrote:OOPS.
But I can precisely reproduce it with a 4k ring texture!
Jens, what happens if you use 1k?

I foregot to mention that I use only the clean installation without any addon. Further the pre3-exe in the pre5-installation works whithout such a crash.

At the moment I fixed this problem by useing a 3ds model for Pan.

Bye Jens

OT: Grant, I know you work on new Saturn ring texture. If you want I can mail my 8k PSD raw version to you?

Post #6by **granthutchison** » 03.03.2004, 23:59

jim wrote:OT: Grant, I know you work on new Saturn ring texture. If you want I can mail my 8k PSD raw version to you?

That'd be nice, thanks. I have what I feel is a very good alpha channel built directly from occultation data (which I've forwarded to Chris so that he can adjust shadows against the denser B ring this produces), but the albedo and colour variation need tweaked.
I'll send you my alpha in return, if you want it.

Grant

Post #7by **selden** » 04.03.2004, 00:48

Jens,

I'll admit that I don't recall which bugs were fixed between the version you have and v53.03. (There is a document on Nvidia's Web site that describes them.)

However, I was running 43.45 with my Ti4200 at home and found that it had bugs in the routine GL_ARB_vertex_buffer_object that pre5 uses to render planets. Use of that routine is new in pre5. It was not used in pre3. Those bugs were fixed when I upgraded to 53.03.

I was hoping that 53.03 might also fix the problems you and Grant are having, since I don't have them.

jim · Post #8by **jim** » 07.03.2004, 13:27

Hi Selden,

I've now the latest driver installed (5306) and as I've expected nothing has changed. Why should a driver make a difference between cms model and 3ds/cmos model?

Now I found out more about the circumstances of that crash. It happens also whith Atlas. It seems that only moons within the rings with cms model led to this crash. I've tried also the basic render path and dissabled ring shadows with the same result. But sometimes no crash happens and I don't know why. I only can say if one of this moons (Atlas or Pan) has correct loaded then Celestia does no more crash.

Still very curious to me :roll:

Jens