0.8.8 Bugs

[Graion Dilach]

I can think of corrupted/missing AAS too, tho I never had an OA server.

If we'd know which map triggered it, would be easier to figure out the cause.

[RMF]

I'm not sure if I get it right, but you're saying maps without bot support will crash the server when you try to add bots? Preventable by testing or not, this wouldn't be nice.

[Gig]

It seems to me that a missing aas file prevents you from adding bots, but does not crash the game. At least when playing locally... I (we?) should check in dedicated mode. And I (we?) should check if one loads a map with bot support, adds them, then switches to a map that does not support them.

Some other tries may include using a wrong aas file (compiled with a previous version of the map), or a damaged aas file (e.g. with some parts of it deleted with a text editor).

[grey matter]

Bots still cause Server Shutdown (Server crashed: netchan queue is not properly initialized in SV_Netchan_TransmitNextFragment) - AAS shutdown on dedicated servers?

Afaik this happened in old ioquake3 builds when e.g. "/kick allbots" was kicking bots "too fast".

[Gig]

Uhm... maybe that guy simply asked a question, to know if that old bug was fixed?
The sentence ended with a question mark, and it was posted in the section of the page called "other - these are old bugs that may or may not be fixed yet".

[GrosBedo]

From my own experience, this specific bug (with netchan and packets sending) happen when the server tries to send/simulate (for bots) packets to players/bots that are not there anymore.

@RMF: So indeed, a missing bot support would not crash the server, but it will produce warnings.

What could crash the server is a weird and wrong management of players/bots status, or maybe a corrupted AAS file. But mainly I think it's either AAS or a tampered server's binary/mod.

[Coder]

One thing i noticed...
I felt having a ffa with bots on wrackdm? (longestyard remake) and i fall off the map but i dont die?

[Gig]

@ Coder
Maybe this is a long term bug in wrackdm17, where occasionally the trigger that should kill you fails, but I'm not able to reproduce it exactly (maybe there it some place of the "floor" (in the void) where the trigger is missing?). Someone should take a look to the .map...

[Gig]

Can we talk a little about the bugs of anaglyph stereo (3D) rendering?
The first one is that the crosshair is missing, and the second one is that the infos in the upper right corner are messed up if you enable both cg_drawfps and cg_drawspeed.

Screenshot here.
They are there since before 0.8.8.

I asked tcpp (author of "cascaded blur" bloom variant) if he could take a look into it, and he ended up with some lines of code that should fix the crosshair. Considering that I am not able to do anything with such lines of code, I copy his reply here:

Problem solved.

First, the cause of the invisible crosshair is cg_draw.c:CG_DrawCrosshair3D(void). Drawing a visible sprite by calling trap_R_AddRefEntityToScene(ent) requires ent.shaderRGBA to be set, but CG_DrawCrosshair3D doesn't set ent.shaderRGBA.

Therefore, CG_DrawCrosshair3D should say:

static void CG_DrawCrosshair3D(void)
{
   …
   // scale the crosshair so it appears the same size for all distances
   ent.radius = w / 640 * xmax * trace.fraction * maxdist / zProj;
   ent.customShader = hShader;

   // set color based on health
   if ( cg_crosshairHealth.integer ) {
      vec4_t      hcolor;

      CG_ColorForHealth( hcolor );
      ent.shaderRGBA[0]=(byte)(hcolor[0]*255.f);
      ent.shaderRGBA[1]=(byte)(hcolor[1]*255.f);
      ent.shaderRGBA[2]=(byte)(hcolor[2]*255.f);
      ent.shaderRGBA[3]=(byte)(hcolor[3]*255.f);
      
   } else {
      ent.shaderRGBA[0]=255;
      ent.shaderRGBA[1]=255;
      ent.shaderRGBA[2]=255;
      ent.shaderRGBA[3]=255;
      
   }

   trap_R_AddRefEntityToScene(&ent);
}


Second, poor performance in 3D anaglyph mode is caused by an combination of r_bloom and r_greyscale. When r_greyscale>=1, bloom screen textures are created with GL_LUMINANCE internal format, but GL back buffer's internal format remains GL_RGB. As a result, qglCopyTexSubImage2D called by R_Bloom_BackupScreen involves conversion of internal format from GL_RGB to GL_LUMINANCE, which is extremely slow, since it cannot be done in some GPU.

The solution for this problem is forcing tr_image.c:Upload32 to use GL_RGB internal format during the creation of bloom screen textures.

Although performance can be improved by the solution I've mention above, there is still another problem: strange behavior of bloom effect in 3D anaglyph mode, for example, bloom effect is applied only for red channel. There are some work-arounds for this problem, but resolving this problem without any overhead might require somewhat larger modification of ioquake3 code.

So, he also noticed there are some some problems with bloom and anaglyph mode together. I don't know what we should do with that.. maybe trying to fix them or disable bloom when in anaglyph mode? Well, for the problem that r_greyscale is slow with bloom on, I suppose we may simply suggest users do to not enable both things at once (if one has got good hardware, he may want it anyway). UPDATE: I've written in the wiki about enabling both bloom and greyscale brings a drastic framerate drop, advising do to not use both of them at the same time.

For the sake of completion, there have been some changes in anaglyph colors in the last year. I write the next lines in a "quote" just to separe them from the rest, and it's not so important.

About the anaglyph color combinations:
In the OA wiki
http://openarena.wikia.com/wiki/Manual/Graphic_options#Stereoscopic_view
we have the the color codes I found in ioquake3 wiki
http://wiki.ioquake3.org/Stereo_Rendering

* 1: red-cyan
* 2: red-blue
* 3: red-green
* 4: cyan-red
* 5: blue-red
* 6: green-red

But maybe recent ioquake3 changes.. changed it:
at the moment, ioquake3 README in their SVN
http://svn.icculus.org/*checkout*/quake3/trunk/README
says

r_anaglyphmode - Enable rendering of anaglyph images
                                      red-cyan glasses:    1
                                      red-blue:            2
                                      red-green:           3
                                      green-magenta:       4
                                      To swap the colors for left and right eye
                                      just add 4 to the value for the wanted
                                      color combination. For red-blue and
                                      red-green you probably want to enable
                                      r_greyscale

... this due to a change they did in their README file in november 2011:
http://svn.icculus.org/quake3/trunk/README?r1=2191&r2=2198
I don't know when they did that in the engine. Anyway probably the old version is still the official, considering that the binary download from ioquake3 site tells its from 1.36 branch (that, AFAIK, is some years old... does someone know when ioquake3 1.37 will be officially released?).
OpenArena 0.8.8 binaries do not have the "green-magenta" support, while "new" OA 0.8.x version 30 test binaries do have.

Doing that research, I noticed this other ioquake3 improvement, that is more recent than OA version 30 binaries and so we don't have it yet (however, it is not related with anaglyph mode or 0.8.8 bugs at all, sorry for the OT)

[grey matter]

Can we talk a little about the bugs of anaglyph stereo (3D) rendering?
The first one is that the crosshair is missing, and the second one is that the infos in the upper right corner are messed up if you enable both cg_drawfps and cg_drawspeed.

I asked tcpp (author of "cascaded blur" bloom variant) if he could take a look into it, and he ended up with some lines of code that should fix the crosshair. Considering that I am not able to do anything with such lines of code, I copy his reply here:

Problem solved.

First, the cause of the invisible crosshair is cg_draw.c:CG_DrawCrosshair3D(void). Drawing a visible sprite by calling trap_R_AddRefEntityToScene(ent) requires ent.shaderRGBA to be set, but CG_DrawCrosshair3D doesn't set ent.shaderRGBA.

This sounds like a bug. Please also report this upstream.

[Gig]

This sounds like a bug. Please also report this upstream.

But in ioquake3, in anaglyph mode, a crosshair is drawn. I checked with ioquake3 1.36, and it is shown there.
I don't know what is exactly the problem they mention here (and to which ioquake3 versions it refers), but anyway it seems it's not the same problem we have in OpenArena anaglyph mode (may be somehow related).

[Gig]

I don't remember if this has been said before...
Anyway, I was checking a thing in 0.8.8 unified package, and noticed that there are some "readme-like" files (README, CREDITS, COPYING, CHANGES) without extension. This is not a problem for Linux users, probably, but is not friendly for Windows users. I think future releases should add .txt extension to all these files (like readme_085.txt and readme_088.txt already are).
More, che "CHANGES" file there arrives up to 0.8.1... I suppose it should contain the main changes also of later patches.

[jackoverfull]

I'm afraid that's some sort of stardan for open source unix projects.

yes, the CHANGES file should definitively be up-to-date.

[Gig]

About 0.8.8 bugs...
- If cg_leiEnhancements is set to 1, then your cg_oldplasma value is ignored (you always see like with cg_oldplasma 1).
- Are we sure cg_leiSuperGoreyAwesome works as expected? It seems to me it's not the blood bath one would expect.

These seem to have been pointed out before...
http://openarena.ws/board/index.php?topic=3386.msg28429#msg28429

[fromhell]

supergoreyawesome jsut adds a few more splats and I think IIRC splat noises. I used to have excessive blood with it but it hit the temporary entity limits too much. To have a true supergoreyawesome would mean a particle system based in the renderer (which has been done before in Quake3MME which had its source released under my pressure )

[capodecima]

I have this errors on every map, its always sounds errros or music errors, i would ignore it but it always drop fps which is not nice. I heard this is some bug which will repair in next version. Or can i solve this now ? Download some missing files somewhere?

[Gig]

I have this errors on every map, its always sounds errros or music errors, i would ignore it but it always drop fps which is not nice. I heard this is some bug which will repair in next version. Or can i solve this now ? Download some missing files somewhere?

About the "music error that drops fps", do you mean that yellow warning that says the music "format is not 22khz stereo" (not seen in your screenshot)?
That message is shown on-screen everytime the music sound track ends and starts again, and I can guess it's displayed because OA music has got an higher sound quality than the one designed for the original Q3 soundtrack.

I hope that message could be fixed by changing it to "format is not 22khz or 44khz stereo" (accordingly adjusting the check that triggers it).
Or, if that fix is not possible, a (worse) fix could be to show it only in "verbose" console output (with "developer 1").
I don't know if such changes would also fix the fps drop when music re-starts or not.

Those error messages that are in your screenshot, instead, only appear in console just after map loading, right?

[capodecima]

i got this errors during game also yes i got error with 22khz, so is possible it fix ?

[Gig]

@Capodecima:
Not sure, but I fear just sliding music volume to the minimum may not prevent the game from "playing" the music. Did you already try?

Maybe you could fool the game by creating a .pk3 containing an empty file with the same path and name of the music file that runs. You may call the pk3 something like "z_nomusic.pk3". I suppose that may stop the music from playing, when playing locally. But that would not work when playing on internet, on any "pure" server. UPDATE: easier workaround found in the next posts.

[Neon_Knight]

There is a console command for disabling music, isn't it?

[Gig]

The music volume "slider" in the GUI controls s_musicvolume variable (1.0=max volume, 0=no music, default=0.25). I don't know if setting it to 0 prevents the game from internally executing the file or not. It would be interesting to know. (UPDATE: considering capodecima's post below this one, it does. Good. )

However, I just found another workaround for capodecima's music problem: it exists the "music" (client-side) command to play a music of your choice, instead of the one pre-set by the map. If you use it with a non-existing file name (e.g. "/music qqq"), you get a file not found (or unknown extension) message, and the music stops.
I suppose this should solve his problem (music format warning with fps drop each time music re-starts), at least until the match ends.   Of course, renouncing to music.

[capodecima]

 s_musicvolume 0 solved my problems , i always thought i have turned off music but i had set it to 0.175.