brush entities do count towards the overall .bsp file limit, but func_wall vs func_detail will have different resource usage because the func_detail has to slice up the map geometry that it is overlapping (leafs/nodes, faces). 

On my current mapping adventure I'm creating an outdoor map with terrain-esque features. I find that in Quakespasm and Winquake I get stuck when walking over that edge:

http://maikmerten.de/base1/terrainstuck.jpg

All fine in FTE and Darkplaces. Map compiled with the awesome ericw-tools. All brush vertices on integer coordinates.

I'm pretty sure a random vertex manipulation should resolve the issue and I guess this is just a side-effect of how Quake handles collision (I think FTE and DP basically have their own ways of doing things there).

Is there a lazy way to resolve such issues reliably other than pushing vertices until things are resolved?

BSP and .map source: http://maikmerten.de/base1/terrainstuck.zip 

It can even happen out of nowhere after a compiling in areas where nothing has been changed since several compilings before.

Move some vertex around a bit and that will disappear. 

Yeah, I already fixed it - once you discover such problems a workaround does seem to be mostly trivial. Just wondered if this is an indication that I'm doing something wrong or if this is just some charming oddity of Quake in general. Looks like it's the latter and I shall embrace the madness Quake awakens in our hearts and minds.

Glad I'm testing with multiple engines, though ;-) 

If one is mapping using trenchbroom on an unsaved map and TB crashes, is the map lost?? 

Ive found them hiding in C:/users/MYNAME.


"Trenchbroom-crash.map" 

 

this is just some charming oddity of Quake in general

Yes, and that's why there is not much terrain in Quake custom maps, and one of the reasons there is none on the official maps. 

niccce. happy to help, drow! 

I'm using ericw-tools to compile my maps but the lightmaps always look like shit. They are too blurry. Is there a way of increasing the lightmap resolution? 

you can do it on a per entity basis by adding an '_lmscale' field set to eg 0.25
on the command line you can use '-lmscale 4' with ericw's light util.
I'd also recommend the '-bspx' light util argument, if only so that you don't need the external lits.

alternatively compile it as a q3 bsp via q3map2 (using q1 ents still so that mods don't break). A _lightmapscale field set to 0.25 or so for that, I believe. this'll give you a few more options that are not available with q1 bsps.

that's the theory, anyway. note that each of these have different results when viewed in other engines. 

There's a good chance I inadvertently broke one of _lmscale/-lmscale/-bspx since I haven't tested them since Spike contributed the features, but a ton of other things have changed since then. If that's the case, there are old releases available at: https://github.com/ericwa/ericw-tools/releases and feel free to file a bug.

Anyway I would +1 using q3bsp. AFAIK all of the _lmscale/-lmscale/-bspx options will produce a q1bsp that only looks right in FTEQW. 

iirc the per-entity _lmscale + -bspx will generate redundant data for engines that don't support it.
obviously other engines will get inferior lighting, but it should at least otherwise work (unless you use -novanilla for smaller bsp sizes, in which case other engines will glitch, but shouldn't crash).

I really ought to add this stuff to QSS too, same as many other things. 

QSS now supports lmscale too, in case anyone cares. 

Won't you lose lightstyles by going q3bsp? I always considered that a feature-gap that might have a sound technical explanation (*) but just robs single-player maps of a simple way to make things appear more dynamic and dramatic.


(*): I guess lightstyles just don't go well with the light-grid, but then I'm not a fan of that anyway because its size just balloons quickly as you extend map dimensions unless you lower the resolution so much you could just as well sample from the lightmap. 

the rbsp variant (and derivatives like fbsp) supports lightstyles (4 per surface, like quake).
and yes, 4 styles per lightgrid node too (it uses some compression scheme).
FTE supports them (if you've custom shaders then lightmap passes need to be first due to format weirdness), but dp doesn't.

(side note: it might be nice to calculate model lighting for the centre of each leaf, and interpolate between those, using surface data in place of neighbouring solid ones.)

even with q3bsp itself, you can get lightstyles with rtlights.
alternatively you can get quite creative with custom shaders. 

The problem with q3bsp is that i can't use TrenchBroom and Radiant is pathetic. :( 

Can you provide a screenshot of the light/shadow situation?

In Quake (when not going to q3bsp) it's pretty much not possible got get sharper shadow/light contrast than shown, e.g., on https://maikmerten.de/base1/randomscreens/happy-with-those-cones.jpg - all one can achieve is to make sure you don't get ugly stairs by using some advanced light options (in my case, for final compiles I use -extra4). 

The problem with q3bsp is that i can't use TrenchBroom
Does the compiler not accept q1 map format?

You could use the -convert flag on my qbsp (see http://ericwa.github.io/ericw-tools/doc/qbsp.html ) and hook that in as a compile step. You can convert vanilla q1 map to q2 (which I think q3bsp compilers accept). Or, valve to brush primitives, etc. 

In Quake (when not going to q3bsp) it's pretty much not possible got get sharper shadow/light contrast than shown
There is the "scale your textures up 2x and use texutre scale 0.5" hack that gets you higher resolution. Downside is it causes more qbsp subdivision of your level so the faces still fit in Software Quake's surface cache. 

Downside is it causes more qbsp subdivision of your level...

Would QS/QSS have any issues with that?

I don't think I asked this specifically in a previous post, if it has been answered already my apologies ;) 

It makes a mess of linear filtering.
Replacement textures should at least fix that, but anyone using the default settings+textures will find it looks a bit ugly.
Software renderers will have less mipmaps available (this also affects fte's r_softwarebanding).

note that you can also double the -subdivide qbsp arg too which will reduce the subdivisions needed.
QS supports up to (128-1)*16-precision. glquake actually supports up to (18-1)*16-precision.
FTE/DP/QSS(now) support up to (256-1)*16-precision. However that doesn't mean you must use lightmaps that big, as it kinda makes a mess of the lightmap allocator resulting in excessive wasted space, with more texture switches/batches (doubling won't hurt much though).