RC 2

[Entropy]

Btw, do you guys use the latest video drivers?

[Beaverhunter]

ttlanhil:

Hmm oki, but I didn't really mean to totally exclude the possibilty that there could be a bug in drivers or GPU. Sorry if that was how it was what it seemed like.

 

The_Piper:

Hehe oki=D But Nvidia and ATI differs on other things too but I would still like to hear if anyone using ATI can get similar crashes=) Even though I couldn't do much about it with what ever result a crash could give (maybe someone else could), it's just my suggestion on approach to investigate this further.

 

Entropy:

Hmm well almost, actually I use the version 97.94, the newest is 158.22. Note though that these does not differ _that_ much even considering the HUGE jump in number, actually the jump was made from 97.94 to 158.19 and then 158.22. I use the older one because it's possible to make a work-around for TV-out Fullscreen video=) But I guess I could try and update.

But the older cards (not in 8800 series) like the one The Piper has? (or did he say before that he had 6600?) Uses completly different drivers. (And 8800 series drivers was completly rewritten from scratch, so it would be ver unlikely for the versions to contain same bug and make that much of a difference) =/

[ttlanhil]

Btw, do you guys use the latest video drivers?

pretty recent. guess this is as good a time as any to update...

[Entropy]

Yes, it's always a good idea to have the latest drivers, just in case.

[Beaverhunter]

I can confirm that the bug in Irinveron is still present in RC2, but I must say that it took abit longer before it actually occured this time. But maybe I wasn't at the right spot right away. But can someone more try this out? ttlanhil? The_Piper again? Some ATI ppl?

 

But note that I compile from CVS, not using the RC2 client.

 

Some "progress"!!=)

I just tried with Shadow mapping off, I rotated like crazy and I ended up crashing with same crash. Maybe it isn't shadow mapping?

Edited May 19, 2007 by Beaverhunter

[ttlanhil]

didn't help. I updated, rebuilt, logged on (test server), boat to WS, turned shadow mapping off, clicked DP flag, and when the progress bar was at full, had it crash (with the same back-trace as previously, with the shadow mapping on in irinveron)

 

ed: it's highly consistent. ws->dp (and not the other way around) with shadows on, and shadow mapping off, will cause a crash. and funnily enough, the backtrace shows that the same numbers are present as in the previous crashes

#0  0x00002b2f3ebf97b3 in _nv000052gl () from /usr/lib/libGLcore.so.1
No symbol table info available.
#1  0x00002b2f3ec64e21 in _nv000052gl () from /usr/lib/libGLcore.so.1
No symbol table info available.
#2  0x00002b2f3ec5ea55 in _nv000052gl () from /usr/lib/libGLcore.so.1
No symbol table info available.
#3  0x00002b2f3ec60009 in _nv000052gl () from /usr/lib/libGLcore.so.1
No symbol table info available.
#4  0x00002b2f3e95ae98 in _nv000061gl () from /usr/lib/libGLcore.so.1
No symbol table info available.
#5  0x000000000042515d in render_submesh (meshId=0, submeshCount=1, pCalRenderer=0x8a70370, meshVertices=0x7d24e0,
meshNormals=0x77a6a0, meshTextureCoordinates=0x73fd20, meshFaces=0x6ad560) at cal.c:199
	submeshId = 0
	vertexCount = 171
	textureCoordinateCount = 171
	faceCount = 208
#6  0x0000000000425cdd in cal_render_actor (act=0x9428c50) at cal.c:376
	pCalRenderer = (class CalRenderer *) 0x8a70370
	meshCount = 4
	meshId = 0
	submeshCount = 1

etc

Edited May 19, 2007 by ttlanhil

[RallosZek]

I will retest too, tomorrow. 96xx branch is latest for my card (nvidia 4200 go), cause nvidia dropped support for that range in the newer drivers. This is the latest version in FreeBSD ports (which is currently in freeze) and my kernel/base system is 6.2-STABLE as of Mar 16.

 

@ttl: did you compile your entire system with debugging to get that detailed info or just libraries that el links with?

[bluap]

Have the latest RC download and a client built from CVS on Linux Debin etch, nvidia graphics card.

 

I also get The_Piper's crash "at irinveron 357,608 at day time". There appears to be some kind of stack corruption going on and the actual position in the code of the crash moves around a bit but always in the shadow drawing code. I can't get a gdb back trace due to the stack corruption. Valgrind is clean apart from the bug I'll detail below *.

 

I can prevent the crash if I comment out the calls to draw_3d_object_shadows() in shadows.c display_shadows() . I've tried various things in draw_3d_object_shadows() but I'm a bit out of my depth. That function has many gl enable() disable() like calls, many are conditional, compile and runtime wise. Could one of these be the culprit? Sorry I can't help any more, I try again tomorrow...

 

* The valgrid error is in new_actors.c with a read from uninitialised. This is caused by guild_id not being set properly if you have no guild tag. The switch at line 679 should probably have a zero length case.

Edited May 19, 2007 by bluap

[ttlanhil]

I can prevent the crash if I comment out the calls to draw_3d_object_shadows() in shadows.c display_shadows() . I've tried various things in draw_3d_object_shadows() but I'm a bit out of my depth. That function has many gl enable() disable() like calls, many are conditional, compile and runtime wise. Could one of these be the culprit? Sorry I can't help any more, I try again tomorrow...

I'll have a look through various functions... maybe I can find when it crashes... blech

* The valgrid error is in new_actors.c with a read from uninitialised. This is caused by guild_id not being set properly if you have no guild tag. The switch at line 679 should probably have a zero length case.

probably. I'd be inclined to use either 0 or -1, but someone involved in custom_look should probably do that (eh, labrat? )

 

to rallos: I'm using debian, system libs are pre-compiled... I have -dbg packages for much of the system, though

[The_Piper]

Updated my drivers, there was a newer version from march 2007, still the same.

 

Shadow mapping on, no crash. Shadow mapping off, crash.

 

Piper

[Beaverhunter]

I crash with shadow mapping on or off as long as Shadows are on. So I would say it's rather shadows than shadow mapping?

[ttlanhil]

okay, to those who have the crash... do you have -DSIMPLE_LOD enabled in your make.conf (or compile options in MSVC)? is there anyone with a crash like this not using that flag?

 

ed: oh, and if removing it fixes it for you

Edited May 19, 2007 by ttlanhil

[The_Piper]

okay, to those who have the crash... do you have -DSIMPLE_LOD enabled in your make.conf (or compile options in MSVC)? is there anyone with a crash like this not using that flag?

 

ed: oh, and if removing it fixes it for you

 

I have -DSIMPLE_LOD in my make.conf, will test tomorrow if removing it helps.

 

Piper

 

Edit:

After a short chat with ttlanhil about day and night, i tested to leave the cave in glacomor with/without shadow mapping enabled. No crash, so the crashes only happen at daytime, so they *ARE* related to shadows.

Edited May 19, 2007 by The_Piper

[0ctane]

Have to report the same object face bug as in the last RC:

That table to my right is supposed to have four solid sides.

I will have to wait a few hours until "daytime" to see if the shadowing problem still exists.

Can anyone else go to these coordinates and check out that table?

Port Anitora 262,201

 

Yes, this is on a Mac (intel variety) with latest data files and CVS code.

 

BTW, here are my build flags/defines (in case I am missing something):

OSX86 ELC BUG_FIX_3D_OBJECTS_MIN_MAX NEW_ACTOR_ANIMATION NEW_FRUSTUM NEW_TEX CUSTOM_LOOK COUNTERS ALUT_WAV USE_ACTOR_DEFAULTS ZLIB ZLIBW SFX EYE_CANDY NEW_E3D_FORMAT FONTS_FIX

[Entropy]

You can remove BUG_FIX_3D_OBJECTS_MIN_MAX, it was for the old format only.

[ttlanhil]

Can anyone else go to these coordinates and check out that table?

well... yes I can... but I went there, and, what table? piccie

 

using:

ALPHA_ACTORS, ANTI_ALIAS, BUG_FIX_3D_OBJECTS_MIN_MAX (yes, I just saw ent's message), COUNTERS, CUSTOM_LOOK, DEBUG, EYE_CANDY, FONTS_FIX, FUZZY_PATHS, MASKING, NEW_ACTOR_ANIMATION, NEW_ALPHA, NEW_FRUSTUM, NEW_TEX, NO_MUSIC, NOTEPAD, OPTIONS_I18N, PNG_SCREENSHOT, SFX, USE_INLINE, USE_LOW_MEM, USE_SHADER, USE_SSE, USE_SSE2, USE_VERTEX_ARRAYS, X86_64, ZLIB, FEATURES += NEW_E3D_FORMAT

[Tanyia]

I still have to log in to see the effects of my LOD settings. I have tested it so far in the C1 portals room and around a WSC fountain. Same thing in both places: I have to relog while in the map before the LOD settings take effect.

 

Reference from RC1 thread: http://www.eternal-lands.com/forum/index.p...st&p=357490

[0ctane]

Can anyone else go to these coordinates and check out that table?

well... yes I can... but I went there, and, what table? piccie

Okay, you can slap me. My data files were getting all messed up. That problem has been solved. I still have to wait for daylight to check the previous shadows problem.

[KarenRei]

Wow, lots of posting between when I last checked in and now I had to work late last night, so didn't get any EL work done.

 

Tanyia: Yes, the problem will still exist, because I haven't been on to investigate it yet I'll look into it this evening. Glad to know that the config options let you customize eye candy to your liking. Just got to make sure that the changes activate without making you log in/out.

 

Hmm, how to test? I can't teleport to the portals room myself. I can walk there, but that might not be the same. Well, I can at least start with that. Taniya, what settings did you end up choosing?

 

Shadow mapping:

 

Wow, we have stacktraces on the crash now! Suddenly, we not only have *a* stacktrace, but we have, what, five? Something to work with! I'll take a peek at that after I work on Taniya's issue.

[Tanyia]

Yeah, I just didn't want it to get lost in the other thread

 

Here are the eye candy settings I have now:

Max framerate: 45

Min framerate: 20

Light column threshhold: 15

[KarenRei]

Looking at the code, a piece of code that I said a while back would probably cause crashes when NEW_E3D is enabled is still in there in draw_3d_object_shadows...

 

#ifdef NEW_E3D_FORMAT   // Always supplies a texture pointer
	glEnableClientState(GL_COLOR_ARRAY);
	glEnable(GL_TEXTURE_2D);

 

Someone show me where the corresponding glColorPointer statement to that glEnableClientState is. Grep shows only one glColorPointer statement in all of EL's code, and it's inside an #if 0'ed block in particles.c.

 

Why do we have so much trouble keeping track of our draw array states? Thankfully, I already knew about -- and mentioned -- this potential one. Two simple rules will prevent this: 1) don't enable a state until right before you set the pointer. Worst case, you enable the state too often. Not a big tragedy there. 2) Before you start, clear all of your states. Once again, you might occasionally disable states too often. Again, not a problem.

 

The CPU time to change a state is almost irrelevant. Too many optimization guides tell you "minimize your state changes", when in reality, the effect is usually trivial(**), but the risk of introducing bugs is significant. Esp. with these sort of draw arrays.

 

Aanyways...

 

Someone who has shadow crashes want to take this line out and see if they still crash? Even if this isn't *the* problem, I don't see how this isn't *a* problem.

 

 

** -- I'll write a demo benchmark program if people want to see how little impact it actually makes.

Edited May 20, 2007 by KarenRei

[0ctane]

Fountain in C2 portals room...looks like water is flowing through structure rather than over each pedestal.

 

While that table/boxes issue in PA was solved, the shadowing issue remains:

Port Anitora 260,200

 

My actor shadow is correct (unless I am walking on a bridge then it is on the water rather than the bridge). The torch and sign shadows extend directly into the ground, and when I rotate they continue to point straight down. The boxes have no shadowing and appear to be floating due to it. I have no idea what is throwing that huge shadow on the lower left.

 

Oh, and those gold things on the rug are upside down baskets. In the shadow at the edge of the rug you can see the handle of the basket. This "I can see into the ground" effect is best illustrated here:

The tree stump has a "shadow" of the entire tree object below it. The fallen tree also has this effect. Idaloran 122,312.

 

Shadow mapping (256) still causes a system freeze. Not even the mouse could not move this time.

[KarenRei]

Tanyia: Can you use the CVS client? I just committed a potential fix for you.

 

 

Fountain in C2 portals room...looks like water is flowing through structure rather than over each pedestal.

 

Can't be helped for this release. There are different fountain geometries, but currently only one hard-coded geometry for the fountain effect. Next release will allow effects to be customized. The ability is already in the effects, but not yet integrated with EL and the map editor (in progress).

[MistickIce]

I saw CelticLady posted about the black screens for Idaloran and Iscalrith storage. I have the same problem all through Iscalrith storage. I had problems with EVTR and Willowvine before the new client but not anymore. I still have a problem with the IotF igloo to Iscalrith as well. I have screenies but they are exactly the same as CelticLady's. I did notice that we both use Intel Onboard Video. Tirun has an actual video card and does not have this problem. I am posting a screenie for the Onboard Video though by way of comparison. No seems to have responded to CelticLady's post on the other thread and this is happening with the new client so I thought it needed a bump.

 

 

 

 

 

 

__Icy__

[Entropy]

Try playing with your settings, first try the poor man, see if ti works, then try other things as well. If you find something that works, let us know.