| Time |
Nick |
Message |
| 00:20 |
|
Kimapr joined #minetest-dev |
| 00:27 |
|
appguru1 joined #minetest-dev |
| 00:41 |
|
Seirdy0 joined #minetest-dev |
| 00:44 |
MTDiscord |
<Jordach> there we go, universal binary built with freetype, openal and sqlite |
| 01:33 |
|
specing_ joined #minetest-dev |
| 04:00 |
|
MTDiscord joined #minetest-dev |
| 04:40 |
|
TC01 joined #minetest-dev |
| 05:32 |
|
absurb joined #minetest-dev |
| 08:00 |
|
ShadowNinja joined #minetest-dev |
| 09:37 |
|
calcul0n joined #minetest-dev |
| 09:48 |
|
hlqkj joined #minetest-dev |
| 09:51 |
|
proller joined #minetest-dev |
| 10:08 |
|
calcul0n_ joined #minetest-dev |
| 10:36 |
|
Fixer joined #minetest-dev |
| 10:56 |
|
absurb_ joined #minetest-dev |
| 11:21 |
|
calcul0n_ joined #minetest-dev |
| 11:28 |
sfan5 |
merging #11160, #11161, #11165 in 6m |
| 11:28 |
ShadowBot |
https://github.com/minetest/minetest/issues/11160 -- Devtest: Remove testnodes_show_fallback_image by Wuzzy2 |
| 11:28 |
ShadowBot |
https://github.com/minetest/minetest/issues/11161 -- Reserve vectors before pushing and other code quality changes by sfan5 |
| 11:28 |
ShadowBot |
https://github.com/minetest/minetest/issues/11165 -- Game: Scale damage flash to max HP by SmallJoker |
| 11:49 |
|
proller joined #minetest-dev |
| 11:51 |
|
x2048 joined #minetest-dev |
| 12:21 |
x2048 |
Hey, anyone with good knowledge of the engine here and how the decisions are made around rendering of liquids? |
| 12:22 |
sfan5 |
just ask here, maybe someone knows |
| 12:26 |
x2048 |
OK, so I'm working on transparency, and how block meshes are generated. There is function face_contents() in mapblock_mesh.cpp that decides what to render on a face, and it uses ContentFeatures.isLiquid() to make decision about liquids. isLiquid() in turn looks at liquid_type ("source" or "flowing") which controls flow, and not drawtype ("liquid") |
| 12:26 |
x2048 |
which controls appearance. Is there a reason or is it just overlooked? |
| 12:28 |
x2048 |
The result is that anything with drawtype="liquid" and liquid_type="none" would render incorrectly, so my take is that it's a bug. |
| 12:29 |
sfan5 |
historically the assumption that liquidtype matches drawtype was probably made, https://github.com/minetest/minetest/commit/7f25823bd intended to change that and missed this |
| 12:33 |
x2048 |
Good, looks like I can change to drawtype test then, thanks! |
| 12:55 |
x2048 |
What is the best way to prevent z-fighting when I e.g. have transparent water adjacent to leaves? Wuzzy mentioned there was something in Irrlicht for that. |
| 12:57 |
sfan5 |
no idea what he could have meant |
| 12:58 |
|
hlqkj_ joined #minetest-dev |
| 13:00 |
x2048 |
Taken from here: https://github.com/minetest/minetest/pull/11130#issuecomment-813153324 |
| 13:23 |
sfan5 |
pushing http://sprunge.us/p4E22u?diff in 5 mins |
| 13:27 |
MTDiscord |
<appguru> x2048: pretty sure there's nothing in Irrlicht for that, offsets are and always have been the way to go - no idea what Wuzzy / Wuzzy2 could have meant |
| 13:28 |
|
Taoki joined #minetest-dev |
| 13:28 |
|
appguru joined #minetest-dev |
| 13:33 |
|
specing_ joined #minetest-dev |
| 13:45 |
Krock |
will merge #10323 and #11156 in 10 minutes |
| 13:45 |
ShadowBot |
https://github.com/minetest/minetest/issues/10323 -- Add vector.to_string and vector.from_string by Desour |
| 13:45 |
ShadowBot |
https://github.com/minetest/minetest/issues/11156 -- Make edit boxes respond to string input (IME) by yw05 |
| 13:56 |
Krock |
done |
| 14:08 |
|
x2048 joined #minetest-dev |
| 14:17 |
pgimeno |
the only solution to Z fighting I'm aware of is to not use a depth buffer, https://computergraphics.stackexchange.com/questions/95/avoiding-z-fighting-with-coincident-surfaces |
| 14:29 |
sfan5 |
hm the broken android build is kinda bothering me |
| 14:30 |
sfan5 |
guess I'll clean up my awful irrlicht building script a bit and finally push it |
| 15:05 |
|
hlqkj_ joined #minetest-dev |
| 15:37 |
x2048 |
OK, I've avoided Z-fighting for my case by forcing backface culling when water is adjacent to anything with visual_solidness == 1, the same behavior as for solid water. |
| 15:40 |
appguru |
Well, that just means you don't draw one of the faces anymore |
| 15:40 |
x2048 |
Yeap, and it kind of works in that case (water <-> glass / leaves). |
| 15:41 |
pgimeno |
it's a reasonable solution |
| 15:41 |
x2048 |
BTW, does anyone know how minetest avoids z-fighting when drawing overlay tiles? |
| 15:42 |
x2048 |
I did not see any universal offset for node faces or anything, so there must be some specific solution. |
| 15:45 |
MTDiscord |
<appguru> Maybe it just overlays the textures |
| 15:50 |
sfan5 |
rubenwardy: https://github.com/minetest/minetest_android_deps_binaries/blob/master/Android/Build/Irrlicht.sh |
| 15:50 |
sfan5 |
if you have some time I wouldn't mind if you tested it (and maybe committed the result) |
| 15:54 |
x2048 |
appguru, I found that MeshCollector simply loops over layers: https://github.com/minetest/minetest/blob/07500479191ed927ab661b3758ffcd2fd43158c5/src/client/meshgen/collector.cpp#L28 |
| 16:06 |
|
DS-minetest joined #minetest-dev |
| 16:14 |
|
Seirdy0 joined #minetest-dev |
| 16:16 |
|
x2048 joined #minetest-dev |
| 16:21 |
Pexin |
> <+sfan5> hm the broken android build is kinda bothering me |
| 16:21 |
Pexin |
thank |
| 16:31 |
x2048 |
Moving on with transparency, I have another problem: in order to draw 6-face transparent liquids correctly, I need to create 6 meshes for each node (which is a lot). I made a shortcut so that if a liquid defines just one tile, I merge the node into the relevant mapblock mesh via MeshCollector, but, for example, MCL2 defines all 6 faces for |
| 16:31 |
x2048 |
mcl_core:water_source, all identical, which results in 200K+ drawcalls even on a small sea. |
| 16:32 |
x2048 |
Is it something to handle in the engine or is it better to document that 6-face transparent liquids are a performance problem and should be avoided (like overlay tiles)? |
| 16:44 |
sfan5 |
hold on aren't defined tiles expanded so that there's always 6? |
| 17:05 |
x2048 |
This only worsens the problem, and I just found out that it replicates the TileSpec on loading from lua :( |
| 17:05 |
x2048 |
Not TileSpec, but TileDef |
| 17:17 |
|
calcul0n_ joined #minetest-dev |
| 17:20 |
|
x2048 left #minetest-dev |
| 17:22 |
|
x2048 joined #minetest-dev |
| 18:16 |
MTDiscord |
<Jordach> possibly an issue with model[] in formspecs: |
| 18:16 |
MTDiscord |
<Jordach> UNSUPPORTED (log once): POSSIBLE ISSUE: unit 0 GLD_TEXTURE_INDEX_2D is unloadable and bound to sampler type (Float) - using zero texture because texture unloadable |
| 18:16 |
Krock |
but minetest generates fallback textures?! |
| 18:17 |
MTDiscord |
<Jordach> no fallback either |
| 19:01 |
|
x2048 joined #minetest-dev |
| 19:20 |
|
DS-minetest joined #minetest-dev |
| 19:28 |
|
TechDude joined #minetest-dev |
| 19:53 |
|
x2048 joined #minetest-dev |
| 20:09 |
|
tech_exorcist joined #minetest-dev |
| 20:09 |
Pexin |
the 5.4.1 backport does not contain the minimap normaltexture fix? |
| 20:11 |
sfan5 |
please comment so we don't forget |
| 20:13 |
|
pauloue joined #minetest-dev |
| 20:29 |
|
tech_exorcist joined #minetest-dev |
| 20:31 |
|
tech_exorcist joined #minetest-dev |
| 20:32 |
|
tech_exorcist joined #minetest-dev |
| 20:34 |
|
tech_exorcist joined #minetest-dev |
| 20:36 |
|
tech_exorcist joined #minetest-dev |
| 20:46 |
|
tech_exorcist joined #minetest-dev |
| 20:48 |
|
tech_exorcist joined #minetest-dev |
| 20:49 |
|
tech_exorcist joined #minetest-dev |
| 21:26 |
|
proller joined #minetest-dev |
| 22:28 |
|
jonadab joined #minetest-dev |
| 22:41 |
|
hlqkj joined #minetest-dev |
| 22:42 |
|
olliy joined #minetest-dev |
| 23:05 |
luk3yx |
Is it possible for https://github.com/minetest/minetest/commit/3edb1ddb8127aa7e8de867be50c695271091cb94 to be included in 5.4.1? |
