In TexMgr_LoadImage8 you have "if (glt->flags & TEXPREF_ALPHA && !(glt->flags & TEXPREF_CONCHARS))" (also the same in a few other places) - "glt->flags & TEXPREF_ALPHA" needs parentheses. 

The initial value of gl_max_anisotropy should be 1. This fixes the infinite loop in the cvar callback and is conformant with the spec (i.e. a value of 1 specifies normal isotropic filtering). 

fitzquake disabled gl_fog and just does a fixed-opacity blend over the sky, unless gl_skyfog is 1 (fully fogged) and then it just draws solid-colored, textureless sky polys.

As for the other bugs, thanks... i'll have to check those out. 

Just learned that there's a 256 frame limit which seems to be an engine side cap.

Going to make changes to have the thing fixed on our end, but thought I'd ask - is there any reason not to fix it? Apart from the obvious 'hardly anyone wants 256+ frames of animation'. 

That limit was raised in protocol 666, and the frames seem to always be passed as 32-bit ints internally, so i'm not sure what the problem would be. Is it crashing? 

Without a warning and only when I add more than 256 frames. I thought I remembered something about it being mentioned in documentation but couldn't find it again.

In any case the model is going on a diet - there are some unnecessary frames in there that can be pruned. Just thought I'd mention it. 

One trick you can use to get rid of frames is to create framegroups from some animations. The safest candidate is the stand animation, since there's no action the monster performs which needs to be synced with the frames. Anything more complicated than that would almost always cause difficulties, but that one should be safe enough. If you're only 8 or 10 over the limit, that might just get you back.

Once you know that trick, you can also make longer idle animations, knowing that they will only cost you one frame. I've not done it yet, but it's one of those ideas that's been knocking about... 

That's a cool idea - although there are actions that need to be run like idle sounds, and we've got some idle tic stands as well.

Having said that there's some stuff that can use it such as the simpler trap style entities like wasps, spikemines and maggots. 

okay, i'll look into it. Thanks for the report. 

http://forums.inside3d.com/viewtopic.php?p=19144

Heh ;) 

cool! thanks! :D 

Hey, just d/l'd FQ 0.85 the other day after wanting to run through a little Quake SP and remembering I didn't really like DarkPlaces last time I messed with it.

Anyway, FitzQuake has just replaced DP for me. :D

I'm having one problem, though - can't seem to get my fog settings to stick (tried it in both autoexec and in config after figuring out that scr_crosshairscale had to be changed there :D. Anyone have an idea why fog density wouldn't stay set? R,G,and B seem to stay set, fwiw.

System specs: XP Pro, A64 X2, 2GB DDR & 8800GTS 512.

Other than that, FQ works a lot like the UGL Quake I loved back in the day, but MUCH nicer (love the working fullbrights and gamma)- Thanks for the updated version! 

That's a known issue, the current behavior is for each map load to set the fog based on it's setting, and if it has no setting, it gets reset. 

I should have guessed it was something like that, since everything else works quite well. I'll just make a keybind for it if I find myself really missing it. :D Also wow, your update feels like Quake the way it was back in '95, thanks again!

(Now, time to drop some of CZG's, Vondur's and Elek's maps in to try 'em all out again...) 

Travail, Quoth and Tronyn's latest efforts :) 

I receive an error I didn't know, as it appears only in hard skill

8062 byte signonbuffer exceeded the standaard limit of 7998

Aguier's glquake doesn't give a message.
Is it the same as too many eddicts? 

no, it just means the overall level has enough stuff that the initial message to new clients is too big.

BUT: if you don't get this warning in aguirre quake, it might be okay. Try in fitzquake using sv_protocol 15 for a more accurate test, or just try it in regular quake (glquake or winquake) and see if it crashes. If it doesn't crash, i think you're okay. 

I was surprised it was only in hard skill.
easy/159 normal/169 hard/175

So it seems hard skill has 6 entities over the limit. 

and yes, the map is rather full.
925 entities, 263 lights, 26288 faces. 

to give more explanation, the "Signon Buffer" is basically a snapshot of the initial map state, which includes the initial state of all edicts, plus any static entities in the level. So you could reduce it by reducing edicts or by reducing static entities. 

The signon buffer has to detail every entity which spawns with a model set. One of the ways you can reduce the size of the buffer content is to have some entities set their models a few frames after the server starts.

It is possible to do this with a hack that works in virtually any quake mod that contains a func_wall and an info_notnull, but it relies on you having some func_walls to apply this to. Take one of these existing func_wall entities, and change it's classname to info_notnull(without making it a point entity). Then add the following keys:

"nextthink" "0.5"
"think" "func_wall"


This will make the func_wall spawn 0.5 seconds after the server starts, which removes it from the signon buffer. Instead it gets sent as one of the first update packets. You need to be careful picking a func_wall to use for this trick. If some other entity will spawn on the wall or may move into it before it spawns, then this trick will probably trap that entity inside the wall. So be a little careful.

This trick cannot be used on any entity which requires a precache, which limits it a great deal. Brush entities have the advantage of getting their model precached automatically. A silent func_plat is possible, although those will give you warnings about sounds, so not a good idea. A func_illusionary will not work, as they are made into static entities which must be in the signon buffer for players to see them.

Incidentally, this trick can be easily adapted to make func_walls spawn in levels when triggered. Take the info_notnull wall, and instead of think/nextthink, add "use" "func_wall" and give it a targetname. Just be careful about entities getting trapped inside... 

Is Fitzquake-SDL likely to run on an ARM-based device (Nokia N900) running Debian Linux, assuming the required libs are present? 

It would have to be recompiled for ARM, that's for sure. AFAIK, SDL can be compiled for ARM, so there's no problem there. What about GL support on that platform? I'm not sure, but I doubt that OpenGL ES is enough to run Fitz. 

Considering that even much "lesser" phones like N95 and even N80 could run accelerated GLQuake and Quake2, I would be surprised if this wasn't the case for N900. 

It seems that Nokia N95 has OpenGL ES 1.1 while the N900 ships with ES 2.0.