Crash: "segmentation error" when demo comes to saturn orbit

[Cro Maat]

Hi,

My config is Mandrake 8.2, Ati Rage 128 and Xfree 4.2 with 3D enhancement.

Celestia crashes whenever the demo comes to saturn orbit. Everything else seems to be working OK.

Cheers

M.E.

[Cro Maat]

Reading other messages on the forum I tried a G to saturn or uranus and Celestia crashes the same way. Should I try a previous version ??



M.E.

[PeFu]

Version 1.2.0 didn't had the problem. Please try 1.2.2 and report back,
whether it works. I would like to track this down but until now I didn't
had enough time to spend on the subject.

[t00fri]

Version 1.2.0 didn't had the problem. Please try 1.2.2 and report back,
whether it works. I would like to track this down but until now I didn't
had enough time to spend on the subject.

It is most probably not a bug in the proper sense. E.g. as I reported elsewhere in this forum, I had this problem, too, with SuSE7.3 Linux and using the glx-headers from the xf86glx-devel package (Xfree 4.2.0). After deleting that package and replacing the headers by those in the mesa-3.4.2-devel package, Celestia-1.2.4 runs perfectly stable near planets with rings. Of course, the problem may also depend on the particular hardware drivers. I know exactly, where it happens in the code, of course. Chris has also spent some time to understand what is going on, without success, apparently.

Bye Fridger

[PeFu]

It is most probably not a bug in the proper sense. E.g. as I reported elsewhere in this forum, I had this problem, too, with SuSE7.3 Linux and using the glx-headers from the xf86glx-devel package (Xfree 4.2.0). After deleting that package and replacing the headers by those in the mesa-3.4.2-devel package, Celestia-1.2.4 runs perfectly stable near planets with rings. Of course, the problem may also depend on the particular hardware drivers. I know exactly, where it happens in the code, of course. Chris has also spent some time to understand what is going on, without success, apparently.

Bye Fridger

My Setup is very similar: I never had the xf86glx-devel package
installed on my box. But I still run XFree 4.1.0 and the NVidia
drivers version 1521. But celestia 1.2.0 through 1.2.2 run fine and
stable here, whereas 1.2.4 crashes reproducable (all compiled with
the same header files and libraries). Mysterious.

[t00fri]

My Setup is very similar: I never had the xf86glx-devel package
installed on my box. But I still run XFree 4.1.0 and the NVidia
drivers version 1521. But celestia 1.2.0 through 1.2.2 run fine and
stable here, whereas 1.2.4 crashes reproducable (all compiled with
the same header files and libraries). Mysterious.
[/quote]

Yes, it is. Currently, I am back from the latest (2880) to 1541 NVIDIA drivers, since there is apparently a subtle NVIDIA driver bug in XEphem that we are trying to track down. As to Celestia, there is, however no difference. Everything works perfectly in my case. I actually always use the latest CVS version, but this is probably irrelevant for this problem. I am also sure that you are experienced enough to make sure that there are no interfering files/headers from earlier distributions hanging around...

But it seems also various people using RedHat 7.2 are having the same problem.

Bye Fridger

[Guest]

PeFu:
------

In case you want to try the CVS version, you might check on another SEGFAULT under Linux, that becomes apparent since I have fixed all keys to work under Linux in the CVS version.

Try to go to Mercury and push CTRL P. You should see mercury now in a nice bumpmapped rendering. Next push CTRL V to activate vertex shading. That should give you a SEGFAULT whenever you do this for objects that have loaded an external bumpmap file via solarsys.ssc.

Would be very helpful if you could check this, since your installation seems to be quite similar to mine.

Of course, everybody else who can report on this issue is most welcome.

Bye Fridger

[Doug]

Try to go to Mercury and push CTRL P. You should see mercury now in a nice bumpmapped rendering. Next push CTRL V to activate vertex shading. That should give you a SEGFAULT whenever you do this for objects that have loaded an external bumpmap file via solarsys.ssc

I'm using a RH 7.2 system with a GF2 card and NVIDIA_GLX-1.0-2802

I see the above bug exactly as you describe.

Nice heads-up on those keys. The earth looks really cool with vertex shading! But, oooh, frame-rate drop! :)

Celestia is so cool! Love the new version.

[t00fri]

I'm using a RH 7.2 system with a GF2 card and NVIDIA_GLX-1.0-2802

I see the above bug exactly as you describe.

Nice heads-up on those keys. The earth looks really cool with vertex shading! But, oooh, frame-rate drop! :)

Celestia is so cool! Love the new version.

Thanks Doug,

that's useful to know. The SEGFAULT was there before, only the
respective keys did not work under Linux to make it show up.

An easy workaround would be for the moment to bumpmap mercury.jpg
under GIMP (Photoshop) with marsbump.jpg and comment out the
respective entry in solarsys.ssc. Same with Pluto I think.
An advantage is that e.g. for Pluto, you may include /both/ bumpmapping
and specular ice reflections (see my example on Bruckner's site)
with GIMP into a pluto.png or pluto.dds (DXT3!) file.

Note that in the CVS version I have also updated the Help file, and all
described mouse functions and keys should work correctly now also under Linux

E.g..

Shift + left drag =>
quickly change field of view from 45 deg standard field to
telescope view (field >33 arc secs).

CTRL +left drag ( or Left + right drag) adjust distance to object (zoom)

For all OS's now:

Middle mouse|wheel button toggling between the standard field (45 deg) and the
previous field (e.g. telescope view)

Also the maneuvring keys (UP, DOWN and Keypad keys) now work smoothly also
at /small/ field of view (telescope mode).

Finally, have a look at the new constellation figures...

Bye Fridger

[Paul]

If you have any further crashes when nearing Saturn, I would also check out this bug:

http://www.celestiaproject.net/forum/viewtopic.php?t=284

Saturn's moon Phoebe uses roughsphere.cms which was causing crashes in recent versions.

[Doug]

that's useful to know. The SEGFAULT was there before, only the
respective keys did not work under Linux to make it show up.
<snip>
Note that in the CVS version I have also updated the Help file, and all
described mouse functions and keys should work correctly now also under Linux
<snip>

I checked-out the CVS tree of Celestia last night and now I have a fully updated, freshly compiled version to play with. Sometimes using Linux is just cool!

The keys seemed to be working great now. The only odd thing I noticed is that I need to keep the mouse pointer in the Celestia window, or the keys act strangely (they stick sometimes). Yes, the window stays highlighted even when the pointer is off the window.

Can't wait until the crash bug is fixed. The earth looks so realistic with that extra stuff turned-on.

I had a great Celestia moment awhile ago. I remembered a APOD photo taken from Mir of a lunar eclipse's shadow showing on the Earth's surface. I dug-up the info on the APOD site, plugged the time into Celestia, and parked in a orbit above Earth. Sure enough, after a few hours, a dark shadow went across the surface, almost identical to the picture taken from Mir. So cool!

[Guest]

If you have any further crashes when nearing Saturn, I would also check out this bug:

http://www.celestiaproject.net/forum/viewtopic.php?t=284

Saturn's moon Phoebe uses roughsphere.cms which was causing crashes in recent versions.

I recite your post from May 1st from this thread here again:

I found a fix for this crash. It seems Celestia's mesh noise generation code does not handle single octaves of noise.

If I go to any asteroid/comet that uses "asteroid.cms" it crashes. By comparing the file with roughsphere.cms (which works fine), I tried changing the line:
Octaves 1
to:
Octaves 2
which is the same as that in roughsphere.cms. Now these objects will appear without crashing Celestia.

Since celestia 1.2.4 crashes in method renderRings() and since
the crash also happens when going to Uranus, "Phoebe" is very unlikely to be the cause of the "segmentation faults" discussed in this thread here.

[t00fri]

The only odd thing I noticed is that I need to keep the mouse pointer in the Celestia window, or the keys act strangely (they stick sometimes). Yes, the window stays highlighted even when the pointer is off the window.

This behaviour I can reproduce. It depends on the settings of your window manager. I use Fvwm2, for example. If I take out the option 'SloppyFocus', then the Celestia window looses focus when the mouse leaves the window and the key shortcuts stop working. With the option left in, the behaviour you describe follows. I shall look further into this.

Bye Fridger