Page 3 of 7

Posted: 14.09.2006, 18:15
by Cham
Okay, after some time testing the most recent built of Celestia, I'm giving here my list of bugs and problems experienced with version 1.5.0 (sept 13 OS X Tiger built). Some people have already reported some of the problems described below. I'm repeating them just for completeness. I'm running this built on OS X Tiger 10.4.7, on a dual G5 with ATI x850 XT video card (256 MB vram).

1- The star glare is disappearing at some distance. Going to a star will make the glare texture to disappear suddenly.

2- Apparently, the galactic label transparency effect isn't strong enough on my system. I have discussed about this with Fridger, but here it is again, for completeness of my "problems" list. Galactic labels appears to be almost exaclty like it was with version 1.4.1. The code is working, but the effect is too subtle. The "3D" effect is not perceptible. I think the transparency should increase a bit faster with distance and there are too much galaxies which are popping in and out from nowhere. It doesn't seems to be very consistent with the transparency effect. In other words, there aren't enough variations (gradations) in the transparent labels.

3- There's a dark straight line in the middle of many galactic templates, as seen from the side. I've discussed about it with Fridger, but IMHO? that line is too straight and looks a bit unatural. It should be wavy in thickness. Maybe it's not possible to do it in the present code ?

4- The menu bar is behaving in some strange way, when in full screen mode (OS X). There's a delay (it's a bit too slow to me), and the menu bar moves slightly, verticaly, in some laggy way. It feels a bit uncomfortable and needs an adjustement.

5- Apparently, the French traduction is very incomplete on my compilation (there seems to be some conflict with what ElChristou is having on his built. I don't know why). The on-screen informations (verbose, speed, follow, FOV, etc ...) are all in English, while it was fully translated in version 1.4.2 (latest Dirkpitt built). Some files needs to be updated.

6- The moons and satellites are showing behind their parent planet (example : Saturn's moons are all visible behind). It? wasn't the case with version 1.4.1 (and 1.4.2).

7- FPS is generally MUCH slower than with the previous version (1.4.1 and 1.4.2 from Dirkpitt). Maybe I didn't made the compilation right (optimisation ?). It's the first time I use the XCode compiler on my Mac. Frame rate is about less than half what it was with 1.4.2.

8- There's a problem with skies, while cruising close to a planet's surface. For example, I get this sky while very close to Earth (about 1 km altitude) :

Image

9- Apparently, bump maps and normal maps aren't working at all on asteroids. I tried to use a bump map and also a normal map on Phobos, and on Eros (Chris version), but it didn't made any difference. I do have OpenGL 2.0 enabled and working (new atmosphere code working very well).

10- I also tried to add an atmosphere on a mesh (Eros), to reproduce what Chris have shown recently, and it gave the same result as with any previous version. Example below :

Image

11- The foreground clipping plane is cutting the orbital paths. I don't have this problem with the previous version (1.4.2).

Posted: 15.09.2006, 03:26
by chris
Cham wrote:Okay, after some time testing the most recent built of Celestia, I'm giving here my list of bugs and problems experienced with version 1.5.0 (sept 13 OS X Tiger built). Some people have already reported some of the problems described below. I'm repeating them just for completeness. I'm running this built on OS X Tiger 10.4.7, on a dual G5 with ATI x850 XT video card (256 MB vram).
Thanks for the bug reports . . . I can comment on a few of them.

1- The star glare is disappearing at some distance. Going to a star will make the glare texture to disappear suddenly.
Yes, there are known problems with the star code, and I'm working on them.

6- The moons and satellites are showing behind their parent planet (example : Saturn's moons are all visible behind). It wasn't the case with version 1.4.1 (and 1.4.2).
There are two version of the star code. The old star code doesn't assign a correct z coordinate to objects that are rendered as points. The new code does. You can toggle between the two by pressing the @ key . . . The old is still there just for debugging. It will eventually disappear when the new star renderer has been fully debugged.

7- FPS is generally MUCH slower than with the previous version (1.4.1 and 1.4.2 from Dirkpitt). Maybe I didn't made the compilation right (optimisation ?). It's the first time I use the XCode compiler on my Mac. Frame rate is about less than half what it was with 1.4.2.
Some of the shaders are more complex. Cloud shadows and atmospheric scattering can tax the geometry unit of your graphics card. In other cases 1.5.0 should be faster. The new star code is faster, as is the new depth sorting code. Anyhow, you need to give specific cases where you observe a slowdown so I have some place to start.

8- There's a problem with skies, while cruising close to a planet's surface. For example, I get this sky while very close to Earth (about 1 km altitude) :
That's an odd bug. I'm sure it's the same as what Fridger reported with his Titan experiments. I'm looking at that one now.

9- Apparently, bump maps and normal maps aren't working at all on asteroids. I tried to use a bump map and also a normal map on Phobos, and on Eros (Chris version), but it didn't made any difference. I do have OpenGL 2.0 enabled and working (new atmosphere code working very well).
Are you specifying the NormalMap in the SSC file or inside the cmod file? NormalMaps in an SSC file won't apply to meshes. I may change this, but for now it's expected behavior. Also, whatever mesh you do apply a normal map to *must* have tangents. The default Phobos model does not.

10- I also tried to add an atmosphere on a mesh (Eros), to reproduce what Chris have shown recently, and it gave the same result as with any previous version. Example below :
I know that this works . . . What atmosphere parameters did you use?

11- The foreground clipping plane is cutting the orbital paths. I don't have this problem with the previous version (1.4.2).


Can you give me a cel URL that shows the problem?

--Chris

Posted: 15.09.2006, 03:38
by dirkpitt
Cham wrote:4- The menu bar is behaving in some strange way, when in full screen mode (OS X). There's a delay (it's a bit too slow to me), and the menu bar moves slightly, verticaly, in some laggy way. It feels a bit uncomfortable and needs an adjustement.

The animation will be slow and jerky if Celestia is using a lot of CPU/GPU. The animation speed, and when/how it is triggered is completely up to the system, so nothing can be done (short of some very ugly hacking) to improve this.

5- Apparently, the French traduction is very incomplete on my compilation (there seems to be some conflict with what ElChristou is having on his built. I don't know why). The on-screen informations (verbose, speed, follow, FOV, etc ...) are all in English, while it was fully translated in version 1.4.2 (latest Dirkpitt built). Some files needs to be updated.

This is probably an easy fix -- once I get my Powerbook back.

7- FPS is generally MUCH slower than with the previous version (1.4.1 and 1.4.2 from Dirkpitt). Maybe I didn't made the compilation right (optimisation ?). It's the first time I use the XCode compiler on my Mac. Frame rate is about less than half what it was with 1.4.2.


Please post the size of your Celestia program. You should be able to obtain it by right/control clicking the program icon and choosing Get Info. The size of an unoptimized build will be larger than an optimized one.

Thanks for taking the time to test!

Posted: 15.09.2006, 04:14
by Cham
dirkpitt wrote:Please post the size of your Celestia program. You should be able to obtain it by right/control clicking the program icon and choosing Get Info. The size of an unoptimized build will be larger than an optimized one.


OOOHHAAAAHEEEEKKK ! I didn't noticed this ! Geez !!!

The normal Dirkpitt built (1.4.2) is about 5.8 MB, while my 1.5.0 compiled version is a whoppa 24.8 MB !!!! 8O 8O 8O 8O 8O 8O

I opened the package to see what's in there, and the normal resources (models, textures, ...) aren't in there. Only codes and app files.

What the hell !? Debugging codes ?

Posted: 15.09.2006, 04:18
by Cham
chris wrote:I know that this works . . . What atmosphere parameters did you use?
--Chris


Here's my Eros code :

Code: Select all

"Eros" "Sol"
{
   Class "asteroid"
   Mesh   "eros.cmod"
   Texture "eros.*"
   BumpMap "asteroid-bump.jpg"
   BumpHeight 3
   #NormalMap "erosnm.png"
   Color [ 1.000 0.878 0.853 ]
   BlendTexture true
   Radius 16.3 # maximum semi-axis
   MeshCenter [ -1.262 0.168 -0.164 ]

   EllipticalOrbit
   {
   Epoch       2451171.0      # 1998 Dec 23 12:00UT (NEAR-Shoemaker encounter)
   Period            1.761    # average
   SemiMajorAxis     1.458261 # at epoch
   Eccentricity      0.222885 # at epoch
   Inclination      10.830143 # at epoch
   AscendingNode   304.430882 # at epoch
   ArgOfPericenter 178.613184 # at epoch
   MeanAnomaly     208.403434 # at epoch
   }

   RotationPeriod         5.270
   Obliquity             78.70
   EquatorAscendingNode 107.23
   RotationOffset       338.165

   Atmosphere {
      Height 10
      Lower [ 0.43 0.52 0.65 ]
      Upper [ 0.16 0.37 0.74 ]
      Sky [ 0.40 0.6 1.0 ]
      Sunset [ 1.0 0.6 0.2 ]     
      Mie 0.0005
      MieAsymmetry -0.15
      Rayleigh [ 0.00025 0.0009 0.0015 ]
      Absorption [ 0.00018 0.00005 0.0 ]
      MieScaleHeight 18
      }

   Albedo 0.16
}

Posted: 15.09.2006, 04:48
by Cham
chris wrote:Are you specifying the NormalMap in the SSC file or inside the cmod file? NormalMaps in an SSC file won't apply to meshes. I may change this, but for now it's expected behavior. Also, whatever mesh you do apply a normal map to *must* have tangents. The default Phobos model does not.
--Chris


Ok, I was using an SSC definition. That explains the problem, thanks.

But then, how did you used the Eros model ? I have the addon you made with cmod model and the normalmap texture.

Posted: 15.09.2006, 05:02
by Cham
More info :

About the frame rate, there's a huge difference between ElChristou's built and mine. On the same scene (simple view of the solar system, orbits on), I get more than 45 FPS with ElChristou's version, while I get about 14 FPS with my built. ElCh app takes about 4.2 MB on disk, while mine is 24.8 MB !! So the frame rate problem is only related to the way I made the built, unless there was an important difference in source codes (ElChristou made his built about a week ago, while mine was made two days ago).

About the orbital path and the front clipping plane, I get exactly the same result with ElChristou version and mine. See the picture below :

Image

Posted: 15.09.2006, 09:36
by phoenix
your executable is 24 MB? 8O

well normally celestia is 1.6 MB on windows (2.4 MB with SPICE).
it seems you have compiled a debug-version, this would also explain your low framerates.

Posted: 15.09.2006, 11:59
by dirkpitt
phoenix wrote:well normally celestia is 1.6 MB on windows (2.4 MB with SPICE).
it seems you have compiled a debug-version, this would also explain your low framerates.


That was the point I was trying to make when I asked Cham what the size of his executable was. :) Anyway, I'm glad that was cleared up. Maybe I should add something to the splash screen saying "DEBUG VERSION" when compiling as debug.

Posted: 15.09.2006, 15:15
by Cham
Good ! My compilation problem is solved. I now have a proper version of Celestia 1.5.0 (application weight : 4.2 MB on disk). Speed issues seems to be mostly solved. And the French translation problems on the screen are solved too. 8)

Posted: 16.09.2006, 04:05
by Cham
Chris,

the stars orbits may have a problem at some scales. They tend to disappear suddenly and reappear while moving around. See for example those binaries :

HIP 23166 A
bet CrB B

Posted: 16.09.2006, 04:16
by Cham
OOooooOOo !

My Messenger model now looks exactly like in my 3D modeler ! The material aspect is now MUCH better ! 8O :D 8)

Image

Posted: 16.09.2006, 19:29
by Cham
There's a lot to say about the new model rendering. There are some aesthetical problems with specular reflections, which becomes too exagerated on some "old" models. See for example the Mariner 9 probe (sorry, no picture about that one here). Many other probes are having the same problem, sometimes worst, actually. However, on other models, the reflection is much better, giving a metallic feel to the surface. Most of my own models are affected in a good way : they are looking a bit different, but they are ok. Sadly, some of my custom asteroids (in my Zet Reticuli SF addon) are now looking completely plastic ! :-(

I also noticed a problem with orbit display. Look at the blue line below. In Celestia 1.5, it's "blinking" in some weird way :

Image

Posted: 16.09.2006, 20:58
by ElChristou
Cham wrote:There's a lot to say about the new model rendering. There are some aesthetical problems with specular reflections, which becomes too exagerated on some "old" models. See for example the Mariner 9 probe (sorry, no picture about that one here). Many other probes are having the same problem, sometimes worst, actually. However, on other models, the reflection is much better, giving a metallic feel to the surface. Most of my own models are affected in a good way : they are looking a bit different, but they are ok. Sadly, some of my custom asteroids (in my Zet Reticuli SF addon) are now looking completely plastic ! :-(


True, many old models (Cassini for example) have now a very high spec, but I suppose a new def of the spec in the 3DS file will solve this problem...

Posted: 16.09.2006, 21:05
by Cham
It will be painfull (long) to edit all those models.

Maybe an option in the SSC, about the old rendering style, or just about specular ?

Posted: 18.09.2006, 04:37
by Cham
I started to edit my afflicted models (specular reflection). The new code is doing a fantastic job on metallic materials :

Image

This is almost a revolution, on modeling with Celestia ! :D

Posted: 18.09.2006, 12:09
by ElChristou
Cham wrote:I started to edit my afflicted models (specular reflection). The new code is doing a fantastic job on metallic materials:


I don't recall this one... What model is this?

Posted: 18.09.2006, 18:41
by Cham
ElChristou,

this model is actually one of the several Babylon5 models available out there. This particular one is very beautifull. However, I've edited it heavily, since I didn't liked the typical "babylon5 style" solar panels on it. Also, the original "command center" on its front was horrible compared to the rest. The original model was extremelly heavy on the frame rate (too much details on the solar panels). So I simplified the model and changed the textures. This version is very satisfying as a "luxury space tourist cruiser" on its tour around some exoplanets. :wink:

Re: 1.5.0 changes

Posted: 19.09.2006, 05:27
by Cham
chris wrote:* Mac: cursor now changes on draggable view boundaries


Hmmm, I don't have this on my Mac. Is it really on CVS ?

Re: 1.5.0 changes

Posted: 19.09.2006, 11:48
by ElChristou
Cham wrote:
chris wrote:* Mac: cursor now changes on draggable view boundaries

Hmmm, I don't have this on my Mac. Is it really on CVS ?


Of course, when you split the view, the cursor change when you pass over the white line (boundarie)...