fteqw-code/engine/client/r_surf.c:3299:16: error: ‘com_resourcemutex’ undeclared (first use in this function)
Sys_LockMutex(com_resourcemutex); 

cd fteqw-code/engine && rm config.h && make m-rel && release/fteqw -basedir whatever

If that doesn't do it then you're going to need to give more info, like what OS you're trying to compile it for/on. Stuff like that matters when it involves threading. 

Still problems. Ubuntu 16.04 64-bit. 

revision 5134 compiles fine for me on 64bit ubuntu 16.04.

try make clean
revert stuff listed by 'svn status' (or just delete everything listed).
and try 'svn up' to make sure you've actually got the current version (and replace anything you may have deleted above).

other than that, I've no idea - it should just be a case of checking out, switching to the engine dir, and running 'make m-rel'. 

Is there any way to preserve an entity through a "reset" command?

I have an entity called "settings_pig". I'm using him to shuttle settings across saves. It works fine for this, but I need some settings to persist after a reset. Is there any way to do this?

I've been using the temp1 cvar, but I can only use that for a little bit of information. Any thoughts? 

I found a (hacky) solution... I can encode more data than I thought into temp1 

Settings shouldn't persist across a reset.

Type "coop 1; map e1m1" in the console if you want things to remain unchanged after death.

You can even save the games (as long as you don't change maxplayers from 1 to a higher number). 

random()'s purported result is 0-1.

In ID1 the ai_melee function does:
ldmg = (random() + random() + random()) * 3;

Soooo...this means that you could receive 0 damage from a knight slash??? Granted low probability due to having 3 of the random()'s and ai_melee being used on multiple anim frames. 

random()'s purported result is 0-1.

In ID1 the ai_melee function does:
ldmg = (random() + random() + random()) * 3;

Soooo...this means that you could receive 0 damage from a knight slash??? Granted low probability due to having 3 of the random()'s and ai_melee being used on multiple anim frames. 

its worse than that.
the vanilla qcc had a bug that caused any operator with function calls each side to discard one side and use the other for BOTH terms.
so (random()+random()+random()) really just ends up as (random()*3).
(for anyone weak at statistical maths, adding 3 different randoms gives a value biased towards 0.5*3 instead of an even distribution, assuming no qcc bugs)
which is of course more likely to return extreme values like 0 or 1.

whether random() can actually return 0 3 times in a row depends upon the implementation. Some of the simpler ones just use a lookup table and might not have 3 such entries in a row. others might work more on a bit rotation basis and similarly be unable to return the same thing 3 times in a row. true randomness is basically impossible on current hardware, especially if you've no access beyond the userland.

side note: returning 1 is more problematic, at least if you use arrays. array[random()*array.length] = error! you can bias it, but then the last element is less likely which is undesirable too.
so dp+fte never return 1 from random(). they do return 0 though.

other side note: you can die even with health > 0!.. death occurs at health<1 (instead of health<=0). stat truncation results in interesting quirks otherwise. 

Remember that random is really a pseudo-random number generator - a fixed stream of numbers which looks random from one to the next and starts in an unpredictable place. It will once in a blue moon throw out an actual zero, this is the cause of the frozen monster glitch. However, I suspect that the PRNG being used here won't actually have a stream of numbers long enough for three zeroes in a row to ever occur.

Technical details: one of the important details in a PRNG is the "period", how long a string of numbers it produces before it loops back round. While a good quality PRNG like Mersenne Twister has a period with 19936 digits (in binary), that was developed in 1997, a year after Quake came out. Doubt anybody's gone to the effort of upgrading Quake's PRNG in the meantime.

Although I can't be bothered to check what Quake does uses (or gets from the default C libraries) I'd guess it's got a period of 2³². If this was used to generate 23 bit numbers (size of the mantissa in floating point) I would estimate the chance of a particular number appearing twice in a row anywhere within the sequence at 2³² / (2²³ * 2²³ ) = 2⁻¹⁴. So while we'd expect some numbers to occur in pairs, there's no guarantee 0 would be one.

And that's just pairs, when we go up to triples we're looking at something 2²³ times less likely, which is actually exceedingly unlikely to even happen once in the 2³². In other words, chances are the RNG in Quake will never output any single number three times in a row. 

..I take 4 paragraphs to say what spike can cover in just 1 

and usually its me that writes the essays...

one thing I feel I have to point out is that quake uses the libc's rand() function, masked by 65535.
entropy is not spread evenly, and most spread the randomness on its high bits.
so that's bad.

I ended up writing my own rand() function to avoid shitty random numbers provided by emscripten's rand... quake's particle system uses a few of the low bits, and emscripten's was just far far too predictable, instead of a random mix of directions half of the particles in each puff went the exact same direction!

its also worth noting that a real random number generator should be happy to return the exact same value repeatedly some times, but generally humans actually want their random number generators to be LESS random. So rand()&0xffff being 0 three times in a row is actually quite unlikely with such a screwed up human-friendly generator.

either way, if its any value < 1 then the player might not notice, so yeah, 0 vs any damage value <1, not that much difference really.

(is that better, Preach?) 

In order to implement seeds in my (progs only) monster randomizer I ended up using random number tables. Doom style. 

The damage function uses ceil to round up the damage so most of the time any values less than 1 get rounded to one. Since it is nigh impossoble to get 0, 0, 0 (using latest FTEQCC so I expect to have all 3 random()s), and 0, 0, >0 is more likely, I would expect the damage range for an ai_melee call to be 1-9 instead of 0-9. 

(talking out my buttcheeks)
What if one just loaded up the knight's animation frames then determined the length of the longest animation swing, like where the night's sword was FULLY extended. Then that that distance and do a trace from the player to the night. If the distance is greater than the distance of the sword then damage = 0. If the distance is within range then full damage. If the animation frame is between fully extended and the wind-up then take a % of damage based on which frame the knight's attack was in. Like if there are 4 frames between backswing to full extension, then determine the distance, (and if in range), base the damage on closeness and attack frame. this would be consistent and still random by player movement.
Does this make sense? just gibberjabbin over here... :) 

*edit: "Then that that distance " = Then take that distance 

What does crandom() do? Do I get cranberry sauce numbers from it? 

Found it:
return 2*(random() - 0.5);

Now I want cranberries. 

As promised about 8 weeks ago, I've finally got round to reposting my teleporting monster flag code which was lost when Inside3d went down.

https://tomeofpreach.wordpress.com/2017/10/08/teleporting-monsters-flag/

I couldn't resist a spot of refining and fixing it. There's a single line fix here which took several days of bug hunting to work out when it affected Quoth, so benefit of accumulated wisdom I guess. The code's also a lot more self-contained - apart from the three entry points which call the new function, there's only one (optional) change to existing code which needs to be made. That should make it very easy to customise and extend... 

When and if I make time to get into QuakeC this sound dandy. 

Hope this is the right place to ask.

GLHexen 2 has an annoying long standing issue when upscaling the HUD with -conwidth, some HUD elements are tiled and the seams become visible, like so:

https://i.imgur.com/aff3Auj.png

Any idea how can this be fixed? I mostly want to submit a patch for the Hammer of Thyrion source port, I've contacted the author and he's definitely interested in fixing this, but he's no idea what's causing this. We've tried changing GL_CLAMP to GL_CLAMP_TO_EDGE, and it helps with some images, but the others still have those thin lines. I've noticed that FTE seems to have this fixed, how was this done? 

clamp-to-edge is good, but it doesn't affect anything when you're atlasing the texture. In such cases, the atlas should just have the border pixels doubled such that it gets the same value instead of averaging in the texel from the neighbouring image.
beware of mipmaps - eg the smallest mipmap is a 1*1 image that averages the ENTIRE texture instead of just the interesting part. you can use a few mip layers by aligning and capping the mip levels (on desktop gl anyway). or just prescale and avoid mips entirely (with 1 pixel borders).
alternatively you can pinch texture coords inwards by half a texel instead of padding with borders, but beware that the half-texel depends upon the mip level that will be used, so that's still messy.
(sidenote: gles supports npot textures only with clamp-to-edge and no mipmapping)

the other thing to do, especially for the hud images themselves, is to use premultiplied alpha (aka: rgb*=a [or equivalent] before uploading, then draw with blendfunc one,one_minus_src_alpha, you'll need glColor4f(rgb*a,a) for variable transparency). This avoids all discolouration on the edges between opaque/transparent texels, how this works should be well documented elsewhere but basically premultiplied alpha is awesome - USE IT!

also, if you're in contact with the author, can you get him to respond to my bsp2 patch (ideally by merging part of it)?
he doesn't really need to bump all the limits that my patch bumped, so long as the clipnodes are no longer so horribly limited... 

If I was to make a mod to create ambient sound entities that could play custom sounds from a point, how would I go about this?

Also what's the best program to use for this kind of thing?

Can custom ambient sounds be done with Quoth? if so, how?