| Time |
Nick |
Message |
| 00:35 |
|
Guest54 joined #minetest-dev |
| 00:36 |
|
kilbith__ joined #minetest-dev |
| 00:50 |
|
tekakutli joined #minetest-dev |
| 00:52 |
|
ShadowBot joined #minetest-dev |
| 01:55 |
|
tekakutli joined #minetest-dev |
| 03:09 |
|
tekakutli joined #minetest-dev |
| 04:00 |
|
MTDiscord joined #minetest-dev |
| 04:58 |
|
calcul0n joined #minetest-dev |
| 06:07 |
|
tekakutli joined #minetest-dev |
| 07:35 |
|
Desour joined #minetest-dev |
| 07:35 |
|
numzero joined #minetest-dev |
| 08:00 |
|
appguru joined #minetest-dev |
| 08:17 |
|
proller joined #minetest-dev |
| 08:20 |
|
hmmmmm joined #minetest-dev |
| 08:20 |
|
appguru joined #minetest-dev |
| 08:25 |
|
appguru joined #minetest-dev |
| 09:06 |
celeron55 |
i'm pretty sure they do. and yes, nice hack, that could solve more problems than it creates |
| 09:33 |
|
appguru joined #minetest-dev |
| 10:58 |
|
appguru joined #minetest-dev |
| 11:22 |
|
appguru1 joined #minetest-dev |
| 11:29 |
|
appguru joined #minetest-dev |
| 11:34 |
|
appguru1 joined #minetest-dev |
| 12:15 |
|
kilbith joined #minetest-dev |
| 12:22 |
|
calcul0n joined #minetest-dev |
| 13:51 |
|
kilbith joined #minetest-dev |
| 14:10 |
|
qur joined #minetest-dev |
| 14:17 |
|
proller joined #minetest-dev |
| 14:30 |
|
Desour joined #minetest-dev |
| 15:21 |
|
jonadab joined #minetest-dev |
| 16:04 |
|
MTDiscord1 joined #minetest-dev |
| 16:11 |
|
MTDiscord2 joined #minetest-dev |
| 17:31 |
|
tekakutli joined #minetest-dev |
| 18:29 |
|
tekakutli joined #minetest-dev |
| 18:31 |
|
numzero joined #minetest-dev |
| 18:36 |
|
kilbith joined #minetest-dev |
| 18:45 |
|
Desour joined #minetest-dev |
| 19:29 |
|
cafblx joined #minetest-dev |
| 19:36 |
cafblx |
I know some of you aren't on Discord, and I'm having issues with reading map blocks, so I'm cross-posting here: https://dpaste.com/8HWRDZ3NA |
| 20:12 |
BuckarooBanzai |
cafblx: i have an implementation in `go` if that helps: https://github.com/minetest-go/mapparser/blob/master/parse.go (used in various other projects) |
| 20:19 |
|
calcul0n joined #minetest-dev |
| 20:26 |
sfan5 |
cafblx: and here's one in python https://gist.github.com/sfan5/52917a0e1338bcda2176edbc416dae60 |
| 20:54 |
|
numzero joined #minetest-dev |
| 21:04 |
sfan5 |
can we get reviews on #13200 |
| 21:04 |
ShadowBot |
https://github.com/minetest/minetest/issues/13200 -- Fix missing particlespawner values by velartrill |
| 21:04 |
sfan5 |
maybe I should set up a macro for this |
| 21:07 |
sfan5 |
merging #13356, #13330 in 10m |
| 21:07 |
ShadowBot |
https://github.com/minetest/minetest/issues/13356 -- Bump protocol version by sfan5 |
| 21:07 |
ShadowBot |
https://github.com/minetest/minetest/issues/13330 -- Update credits for 5.7.0 by sfan5 |
| 21:14 |
|
Guest54 joined #minetest-dev |
| 21:18 |
Guest54 |
celeron55 the waterlogged drawtype hack does not work perfectly right now though, i can't get the corners to align perfectly if param2 is not updated with regards to the surrounding liquid. i am still thinking about how to best handle this – ideally, i'd want it to work for every fluid without having a separate node definition e.g. for water and |
| 21:18 |
Guest54 |
river water and lava “waterlogging”. |
| 21:29 |
Guest54 |
celeron55 i'll experiment a bit with the approach, but if you have any ideas regarding it, please tell. so far i got a few that i have not yet done, but might: 1. steal the special tiles and maybe other things from the highest level surrounding liquid (or something like that). 2. somehow propagate water flow through waterlogged nodes somehow. 3. |
| 21:29 |
Guest54 |
figure out the complete state machine for a hypothetical 5-state automaton rule with air, flowing liquid, source liquid, waterlogged flowing liquid, waterlogged flowing source. 4. make sure a param2 waterlogged nodebox or rails node copies the param2 from placement in a liquid automatically (no reflow necessary after placement). |
| 21:34 |
Desour |
Guest54: afaik, your hacky thing did just use the liquid corner heights of the neighbors when rendering, right? (it's probably quite ugly watching how liquid then flows down on a node and immediately jumps into the node below) |
| 21:37 |
Desour |
iirc, liquid param2 just stores the liquid height. so, using the few remaining bits of some paramtype2s might work. i.e. you have 1/16th nodes instead of 1/256th nodes |
| 21:38 |
Guest54 |
Desour i basically just render the liquid stuff into a node if f->param_type_2 == CPT2_WATERLOGGED |
| 21:39 |
Desour |
oh, so you the information from the last node? eww |
| 21:40 |
Guest54 |
and no, it does not look ugly actually, as you can see from the picture. what i have done in the picture is fiddled with param2 of the waterlogged nodes manually (basically that adjusts the height of the water level) |
| 21:40 |
Guest54 |
well, i can show you how primitive it is, wait |
| 21:41 |
Guest54 |
i really was just playing around |
| 21:41 |
Guest54 |
and it is horrible i know |
| 21:41 |
Guest54 |
i am not using the correct values |
| 21:41 |
Desour |
what I described as ugly was the liquid flowing into it, a picture can't show this |
| 21:42 |
Guest54 |
Desour ignore the blatant cheating please ;) https://mister-muffin.de/p/W88j.diff |
| 21:43 |
Guest54 |
you see it was only a quick idea i had that i sketched in like 20 min or so yesterday, on a train |
| 21:43 |
Desour |
ehem, checkPrivilege 👀 |
| 21:45 |
Guest54 |
well i'm only using that it on anarchy servers obv (also, a more practical reason is i hate to input “/grantme all” in every fresh singleplayer world) |
| 21:47 |
Desour |
there's a setting for default privs |
| 21:47 |
Guest54 |
oh |
| 21:47 |
Guest54 |
anyways, you can see the waterlogging hack is very primitive, but if you play around with it, you might agree that probably the only thing needed to make it look good at this level of complexity is to actually update the level of the waterlogged nodes properly and have it propagate to surrounding liquid |
| 21:48 |
Desour |
(I'm seeing now that you're not reusing the liquid stuff from the last node (as I earlier thought), but hard-coded it for the waterlogged node) |
| 21:48 |
Guest54 |
i'd like to claim i'm not that dumb, but once during testing i filled my devtest world with self-replicating waterlogged source nodes, so i doubt you'd buy such a claim :P |
| 21:49 |
Desour |
it wouldn't be dumb, just more hacky |
| 21:50 |
Guest54 |
anyways, as with a lot of other things i'd might to reiterate that there is an easy and obviously wrong way to develop this thing further and it results in having to define a waterlogged node for each fluid type |
| 21:50 |
Guest54 |
like to |
| 21:50 |
Guest54 |
(i sleepy) |
| 21:50 |
Guest54 |
have you dealt with fluid flow in the past desour? |
| 21:50 |
Desour |
no |
| 21:54 |
Desour |
hm, the liquid types could be stored in a sparse mapping, for example a unique_ptr<array<unique_ptr<array<content_t, 8>>, 16*16>> |
| 21:55 |
Guest54 |
uh what |
| 21:55 |
Guest54 |
why |
| 21:56 |
Desour |
aka content_t (*(*)[16*16])[16] (outer parens might be wrong) |
| 21:56 |
Desour |
for less mem usage |
| 21:56 |
Desour |
or how would you map to the liquid type? |
| 21:58 |
Guest54 |
btw, i actually wanted to auto-update the levels in the waterlogged nodes from adjacent fluids in my hack properly and do some more stuff, but two things made me not do that, first “dear passengers, in a few minutes we'll be reaching [redacted] main station” and second because i got frustrated with how minetest rebuilds a lot of stuff when you |
| 21:58 |
Guest54 |
even change the timestamp on nodedef.h |
| 21:59 |
Guest54 |
or when you switch branches in git |
| 22:00 |
Guest54 |
Desour i don't understand why you are talkign about liquid type mapping at all |
| 22:00 |
Desour |
use ccache, then such recompiles are fast |
| 22:01 |
Desour |
because having a separate nodedef for each pair in waterlog-nodes x liquids would be meh |
| 22:06 |
Guest54 |
look, i have investigated build issues in the past and i know how to do do fast and reliable incremental rebuilds of minetest fast (use a recursive top down build system that checks file hashes when timestamps mismatch instead of the make-style toposort timestamp mess). i just did not have the necessary files with me. |
| 22:07 |
Guest54 |
back to liquids, why do you think about this in terms of nodedefs at all? |
| 22:07 |
Guest54 |
i mean, i just want a stateless rendering hack that gets turned on by paramtype2 = waterlogged |
| 22:08 |
Guest54 |
everything relevant – liquid type, liquid level, special tiles etc. pp. should ideally be inferred from the surroundings |
| 22:08 |
Guest54 |
so you can just plop rails in swamp water or river water or lava and have them do the thing |
| 22:09 |
MTDiscord |
<x2048> Merging #13352 now |
| 22:09 |
ShadowBot |
https://github.com/minetest/minetest/issues/13352 -- Return 'loops' occlusion culler under a setting by x2048 |
| 22:10 |
Guest54 |
Desour i definitely don't want the waterlogged node to “decide” anything besides it's param2 (which should be updated from surroundings). i am thinking about this from the POV of “how can i make it so that any idiot could use it”. |
| 22:10 |
Desour |
>everything relevant [...] should ideally be inferred from the surroundings <- and how do you want to do this fast and reliably? |
| 22:11 |
Guest54 |
and my answer to it is that “any idiot” probably would only want to change the param2 type to waterlogged and be done with it |
| 22:11 |
Guest54 |
like make a chest or rails waterlogged and it should ”just work” |
| 22:12 |
MTDiscord |
<x2048> ...merged |
| 22:12 |
Guest54 |
Desour what's slow about treating waterlogged as adjacent liquid from the POV of any liquid and stealing special tiles from adjacent nodedefs (e.g. the liquid node right above a node) |
| 22:13 |
Desour |
the problem comes if you want to flow through more than one node |
| 22:14 |
Desour |
(if you don't want flow-through, you don't even need a param2 value) |
| 22:15 |
Guest54 |
oh, so the issue is that because you need to figure out where the fluid comes from you actually want the waterlogged node to adopt a fluid type |
| 22:16 |
Desour |
yes |
| 22:16 |
Guest54 |
Desour so i would guess you might be a bit unhappy with my “every liquid treats waterlogged node as the same liquid for purposes of liquid level updates” idea as it is on a level of HAHA STUPID |
| 22:18 |
epoch |
would it work to have liquids treat water-loggable blocks as air for purposes of liquid movement? |
| 22:19 |
epoch |
might be interesting to have waterlogged sand and gravel |
| 22:19 |
Guest54 |
no that would not work, because the liquid replaces air |
| 22:20 |
Desour |
there aren't just liquid updates, but also draw. and liquid can maybe flow out of waterlogged nodes. so treating all liquids in waterlogged nodes the same wouldn't be perfect |
| 22:20 |
Guest54 |
Desour obviously i need to figure out some smoke and mirrors so not more state is necessary |
| 22:20 |
Guest54 |
epoch waterlogged sand and gravel would be just … sand and gravel? i don't get it |
| 22:21 |
Guest54 |
i mean the main thing i wanted to have is rails leading into the sea and fences underwater not making these ugly air pockets |
| 22:21 |
epoch |
it is like the difference between hoe'd dirt and waterlogged hoe'd dirt |
| 22:21 |
epoch |
visually they'd be the same, but they'd leak water through I guess |
| 22:21 |
Guest54 |
the way i have implemented it you can even have waterlogged water hehe |
| 22:21 |
Desour |
water++ |
| 22:23 |
epoch |
probably wouldn't want water logged stairs to leak water from all sides, just the sides with holes. >_> |
| 22:23 |
Guest54 |
2 wet 2 flowy |
| 22:24 |
Guest54 |
oh yeah that was the one point i did not bring up so far i want a solution that shuts up everyone who worries about waterlogged stairs |
| 22:24 |
Guest54 |
i thought it just seemed petty to mention it, but the waterlogged stairs thing comes up every time and i'm tired of it. |
| 22:25 |
epoch |
this is more worry than I want atm, so do whatever |
| 22:25 |
Guest54 |
nah, i am fishing for ideas |
| 22:25 |
Guest54 |
i mean you could feasibly figure out if the entire side of a nodebox is blocked, but it would complicate things beyond “solution fits in head” |
| 22:34 |
Guest54 |
just in case i might have missed it, did anything happen in the “let's make obby glass invisible and break optionally framed glasslikes in various ways for non-16×16 texture packs“ saga? haven't seen anything and i just want to know if the issue is disregarded or not and if it is disregarded what the reason is. (in before: “glass textures |
| 22:34 |
Guest54 |
that do not have at least 1/16 of frame were never guaranteed to work, it's the fault of people making 7×7 or 512×512 texture packs lol”.) |
| 22:34 |
Guest54 |
Desour if not through a param2, how would you handle the non-flow-through visual-only hack? |
| 22:35 |
Guest54 |
Desour i thought using a param2 kinda makes it possible to add some neat stuff later |
| 22:35 |
Guest54 |
i am of sleep. good night everyone! |
| 22:35 |
epoch |
g'night |
| 22:36 |
Guest54 |
Desour and if you come up with some interesting variation on my thing, show and tell pls. |
| 22:40 |
Desour |
if you have no flow-through, you can just take the neighboring param2 and use the param2 that would be at the waterlogged pos if the liquid had already flown into it (i.e. just subtract something from the neighbor with max param2) |
| 22:41 |
Guest54 |
well, it looks pretty ugly tbh |
| 22:42 |
Guest54 |
because then every liquid ends in the waterlogged node |
| 22:42 |
Desour |
if there are different liquid types, just don't flow in |
| 22:42 |
Guest54 |
e.g. stacking fences underwater makes useless extra water-flows-into-this-thing faces |
| 22:42 |
Guest54 |
unless i am mistaken |
| 22:43 |
Desour |
if you have water above, you don't draw the top face |
| 22:44 |
Guest54 |
well, if you have water above a waterlogged nodebox and no flow-through, then a waterlogged nodebox below will only be flown into from the sides |
| 22:44 |
Desour |
for stacking fences you could maybe check the node above the liquid neighbor |
| 22:47 |
Guest54 |
just save liquid type in node meta, problem solved hahahaha |
| 22:47 |
Guest54 |
(it does not work, don't do it) |
| 22:48 |
Guest54 |
Desour could you point me to a document that explains the sparse map trick you were alluding to? |
| 22:49 |
Desour |
there is no document |
| 22:49 |
Desour |
it was just an idea |
| 22:49 |
Guest54 |
well, as far as i see it you only need to save one value for liquid type per waterlogged node, so where does the sparse anything come in? is it because not everything is waterlogged usually? |
| 22:50 |
Desour |
it's for flowing more than one node far, and flowing out of waterlogged |
| 22:50 |
Guest54 |
so what would be stored in the map |
| 22:51 |
Desour |
it's one value per node, but not one value for every node in the block, hence sparse |
| 22:51 |
Desour |
content_ts |
| 22:53 |
Guest54 |
ok i see where you want to go with this i think, but i am somewhat sceptical of if it is worth it at all. like what is the worst that could happen with the “treat waterlogged as adjacent liquid of the same type and just update from there” trick? having a bunch of waterlogged nodes where water flows in and lava flows out? |
| 22:55 |
Desour |
do you now want to make flow? |
| 22:56 |
Guest54 |
probably not until i have a proper model of it |
| 22:57 |
Guest54 |
which is why i am asking what's the worst that you can imagine |
| 22:57 |
Desour |
when making the mapblock mesh, you need to find out the node where it came from |
| 22:58 |
Desour |
s/it/the liquid/ |
| 22:58 |
Guest54 |
yeah but obviously that might not be solvable for the general case, even without flow-through |
| 22:58 |
Guest54 |
consider the case W - N - L, where W is water source, N is waterlogged nodebox, L is lava source |
| 22:59 |
Guest54 |
where does the liquid come from eh? |
| 22:59 |
Desour |
I'd not draw any liquid in that case |
| 22:59 |
Desour |
or just choose one based on a hash of the nodepos |
| 23:00 |
Guest54 |
in any case, i just wanted to point out that these kind of problems occur without flow-through |
| 23:01 |
Guest54 |
and i am really not sure what to do in this case anyway lol |
| 23:01 |
|
lhofhansl joined #minetest-dev |
| 23:02 |
lhofhansl |
I think #13352 wasn't quite done when it got merged. I left two comments. |
| 23:02 |
ShadowBot |
https://github.com/minetest/minetest/issues/13352 -- Return 'loops' occlusion culler under a setting by x2048 |
| 23:02 |
Desour |
with flow-through you have harder issues. if one node is somewhere inside a big cube filled with waterlogged nodes, how do you find the liquid? |
| 23:03 |
Desour |
and note that such a cube can be multiple mapblocks large |
| 23:04 |
Desour |
the meshmaker doesn't have any nodedata out of the mesh chunk |
| 23:06 |
Guest54 |
increasingly makes it seem like flow-through might not be worth it hehe |
| 23:08 |
Guest54 |
i mean maybe just like just let mods deal with it like they deal with the stuff where water nodes randomly tp themselves through sand and gravel or so |
| 23:10 |
Guest54 |
or maybe just handle flow-through “properly” for the ”from above” case, as then you have no conflicts regarding which liquid it could be? anyway, nighty nighty |
| 23:25 |
|
proller joined #minetest-dev |
| 23:25 |
|
kilbith joined #minetest-dev |
| 23:44 |
|
Thresher joined #minetest-dev |
| 23:45 |
Thresher |
how do i force an area around an entity to stay loaded? |
| 23:45 |
Thresher |
im having a problem where an entity despawns when its too far away from somebody |
| 23:46 |
Thresher |
ive tried using `minetest.load_area` in a 20-block radius around it |
| 23:46 |
Thresher |
i also tried `emerge_area` with the same radius |
| 23:46 |
Thresher |
but no luck |
