Can I make a recursive function? What is the maximum stack depth? 

You can create a recursive function, the maximum stack depth is 32. This is possible to hit, see http://tomeofpreach.wordpress.com/2013/08/09/going-overboard-with-field-pointers/ for an example and a bit more detail.

Spike posted a warning somewhere on here that local variables to a recursive function are not reset between calls. So you should initialise all your variables like you would in C when writing a recursive function, rather than assume they get initialised to zero. 

Thanks. 32 is very small for my task. I'll have to think of something else 

How much time can be in progess the function? Can the engine to break it off if it is too long time? 

The engine does have a limit, not strictly on time spent, but on number of instructions. You can only run 100,000 instructions in a single think, touch, spawn (etc) function before the QC concludes that there's a runaway loop and the engine packs it in. 

regarding locals getting clobbered in recursive functions, you can try '#pragma warning enable F302' to get fteqcc to print warnings when it detects that locals were not initialised before use.

fte/dp have a higher limit than 32.

you can get around the instruction limit by splitting up your routine into multiple thinks. in general your mod will be unplayable if you do not do this in the first place due to the stalls induced by running the qcvm.
in fte, there's some obscure 'sleep' builtin which will defer the current invocation/coroutine until a later time. doing this will preserve locals but not globals/fields. the instruction limit will be reset in the process. hurrah for infinite loops that won't lock up the game. just make sure nothing is harmed if some other think or whatever function removes the entities in the mean time... 

I thought it was 100000 instructions per frame. 

You get a new counter every time PR_ExecuteProgram is called, which means every time the engine calls a QC function of some kind.

They can even nest - suppose you are in a monster's think function, and use 50000 instructions before you call walkmove and collide with a player. You get a new lot of 100000 instructions to run the monster's collision function, then yet another set of 100000 to resolve the player's collision function, THEN you go back to the think function and get the remain 50000 instructions to wrap up there. 

ohhhh, nice to know i was being overly conservative then. this is like finding 20$ in my couch! 

Is it possible to return a string to a function that does traceline and returns the texturename of the wall that traceline hits. I understand that the entity returned would be world, but is it possible to get the texture name from the bsp surfacr? Like "brick2_4"? 

 

probably darkplaces has this. 

Check for extension DP_TRACE_HITCONTENTSMASK_SURFACEINFO

Then you can define:
string trace_dphittexturename;
then when you do a standard traceline, the texture name gets assigned to that variable.

One of the nice properties here is that there's no extra built-in function involved, so the code is very easy to make safe in engines which don't support the extension. In those engines, the texture name will always be "". So you can do stuff like special footsteps or impact particles on particular textures, and the other engines will gracefully degrade to the defaults for free! 

//----------------------------------------------------------------------
// Check for contact with sky brushes
//----------------------------------------------------------------------
float(vector targ_origin) check_pointcontent =
{
local float surfnum;
local string texturename;

surfnum = getsurfacenearpoint(world, targ_origin);
if (surfnum >= 0) {
texturename = getsurfacetexture(world, surfnum);
if (strncasecmp(texturename, "SKY", 3) == 0) return TRUE;
}
return FALSE;
}; 

float(vector targ_origin) check_pointcontent =
{
����return (pointcontents(targ_origin) == CONTENT_SKY);
} 

The latest version of DP does sky pointcontent detection different, hence the function, but a working example of detecting surface information can be adapted for what Qmaster wanted. ;) 

nt 

As ever yall are awesome. Thank you! 

That one was too easy for the masters. Here's a real stumper for you.

I want to simulate a halucinagenic side-effect for one of my power-ups in game. Here's how I want to achieve it, in psuedo code:

float TRIPPY_EFFECT_INTERVAL = 5.000 //5 seconds
// string "gfx/palette_trippy1.lmp"
// string "gfx/palette_trippy2.lmp"
// string "gfx/palette_trippy3.lmp"

void () trippydrug_think = {

if (trippydrug > FALSE)
{
setpalettefile("gfx/palette_trippy", trippydrug, ".lmp");
trippydrug++;
self.owner.trippy_power = (trippy_power * 2)
self.think = trippydrug_think;
self.nextthink = (time + TRIPPY_EFFECT_INTERVAL);

}

};

The 1st palette file will have the colors shifted over 4 colors so most things are similar to what they should be, then mixed up a little bit in the 2nd, with the 3rd palette file looking like crayola barf making it hard to see.

So... I can't find out how to change the palette on the fly within the engine code? I can override palette.lmp in the gfx folder, but... more than once, during runtime...? 

host_basepal or something stores the palette.lmp file in memory. mix that up with the "bf" command to figure out how the engine applies a new palette in a software renderer. or something.
in gl renderers things get much more messy, figure that one out on your own, if you dare. 

Its like a treasure hunt! 

I've got two questions, and I suspect that spike might be able to answer for fteqcc if nobody else can.

1) Is it possible in QC to have a function that returns a function - and if so, what is the syntax for that?

2) If 1) is possible, I would like a function that can return itself (or other functions with the same signature as itself). Is this possible, and how? 

if in doubt, typedef it.

typedef void() fnc;
fnc bar = {}; //or: void() bar={}; depends which you consider more correct/easier
fnc() foo = {return bar;};
void() test = {local var fnc t;t = foo();t();};

foo()(); isn't supported for some reason.
nor is void()(). bah.
seems the return statement doesn't do strict checks beyond basic types, must fix some day.
needing to define a local function as var is also worrying, one day I'll want to be able to define functions inside functions, I believe it should be possible to parse nested functions automatically, but I'm paranoid enough to not fix that until then.
on the plus side, void()foo()={return bar;}; also works! yay for mixing notation.


a function that returns itself ad infinitum is problematic. probably best to define the function as returning a __variant, seeing as you'll need to store it anyway before you can call it. qc can then suffer from the infinite type recursion itself. :P
I'm not sure what the advantage is though. 

Thanks for that, the plan is to do some contination-passing, state machine stuff. Thinking it might be easier to use a global rather than passing a return value, but the __variant trick would probably make it work... 

http://pecope.co/experiments/quake_web_tools/QuakeWebTools.html

I haven't really been working on this until the last couple of days, so it hasn't changed that much, but I added camera controls to the MDL viewer and you can now preview BSPs in 3d (though there is no lighting currently) with first person controls.

Seems to handle anything I throw at it too, which is surprising. Thought it would be easier to break :)

Anyway, I'm still planning to implement a proper UI at some point, but I have been concentrating on core file type support, which is about done now, but there is still a lot of cleanup that needs doing.