Minetest logo

IRC log for #minetest-dev, 2014-04-25

| Channels | #minetest-dev index | Today | | Google Search | Plaintext

All times shown according to UTC.

Time Nick Message
00:26 luizrpgluiz joined #minetest-dev
00:26 luizrpgluiz left #minetest-dev
00:29 Taoki Hey. Anyone know why finite liquid was removed? I didn't really use it but still sorta sad news.
00:29 Taoki I assume it was too buggy and no one cared to improve it?
00:36 literaryraven joined #minetest-dev
00:47 sfan5_ joined #minetest-dev
00:58 Taoki Well, that seems to be a question no one has an answer to :P
01:15 us`0gb joined #minetest-dev
02:13 ShadowNinja Gah, I'm trying to fix builtin's organization but I keep changing things that will comflict with my async pull.  Can someone look at it? celeron55, hmmmm, kahrl? #1226
02:13 ShadowBot https://github.com/minetest/minetest/issues/1226 -- Remove dependency on marshal and push the Lua error handler only once by ShadowNinja
02:14 ShadowNinja Simply moving files should be handled pretty well by git, but I have to do more than that to tryto unifiy builtin.
02:15 ShadowNinja Eg minetest.get_modpath("__builtin") / engine.get_scriptpath() / SCRIPTPATH.
02:29 dzho joined #minetest-dev
02:46 Ritchie joined #minetest-dev
02:47 CraigyDavi joined #minetest-dev
03:55 CraigyDavi joined #minetest-dev
05:14 CraigyDavi joined #minetest-dev
05:15 OldCoder joined #minetest-dev
06:12 celeron55 Taoki: i don't know, but i think it's for the best; it wasn't going to be fixed in an adequate way so better take it out, and the design of it didn't really fit with minetest's direction in any case
06:13 celeron55 if you want the finite liquid + weather integration, freeminer does provide continued development on that
06:17 VanessaE Taoki: and if all you really need is temperature/humidity controls, plants_lib can do that much already.
06:17 VanessaE (add the snowdrift mod and you more or less have the same damn thing)
06:18 VanessaE (I'm assuming mgv7 is not being used)
06:40 VargaD_ joined #minetest-dev
07:31 diemartin joined #minetest-dev
07:43 darkrose joined #minetest-dev
08:26 BlockMen joined #minetest-dev
08:26 BlockMen i gonna push this fix in a few minutes https://github.com/BlockMen/minetest/commi​t/31af5dc55ad5fefc33c75e94b39323b36274e527
08:39 BlockMen umm, no, this. s8 is enough https://github.com/BlockMen/minetest/commi​t/c5324015bc70488c8b6d68d83ca7d963a5e6c62f
09:22 Hiradur joined #minetest-dev
09:30 PilzAdam joined #minetest-dev
10:13 Hiradur joined #minetest-dev
10:30 smoke_fumus joined #minetest-dev
10:33 celeron55 here's some AI stuff that might interest somene: http://www.altdevblogaday.com/2011/0​2/24/introduction-to-behavior-trees/ http://web.media.mit.edu/~jork​in/gdc2006_orkin_jeff_fear.pdf
10:33 celeron55 someone*
10:39 proller joined #minetest-dev
10:41 Anchakor "My name is Björn Knafla. I am parallelizing computer and video games with a strong focus on artificial intelligence (AI)."
10:41 Anchakor sure does not sound like the computers I know
10:44 rdococ joined #minetest-dev
10:49 celeron55 your english parser is broken 8)
10:53 Taoki celeron55: Well, I can't say I miss finite liquid much. But I really suggest bringing weather back! I don't see any reason why water sources can't be turned to ice blocks when temperature is smaller than 0*, and back to water source again once it's over 0*
10:54 Taoki Of course, we can't have rain generating water on the surface any more. But freezing / unfreezing should really still work
10:55 Taoki Well, even if minetest_game won't re-add that, I'm going to add it in my own games, so it's still fine in a sense
10:56 proller placing one ice you will flood all around
10:57 Taoki proller: But you could do the same by placing a water source there
10:57 Taoki If you mean that someone could use this to troll
10:57 celeron55 Taoki: ask minetest_next about that, or whatever your preferred still-developed game is
10:57 Taoki I mean, someone could take 10 water buckets and place them all over if they're not responsible. Same with ice
10:58 Taoki ok
10:58 celeron55 (and note VanessaE's comments)
11:39 Shardvexz joined #minetest-dev
12:15 proller joined #minetest-dev
12:17 werwerwer joined #minetest-dev
12:26 Megaf left #minetest-dev
12:49 ImQ009 joined #minetest-dev
12:57 pi1 joined #minetest-dev
13:00 Megaf Hi everyone any word on #1243 ?
13:00 ShadowBot https://github.com/minetest/minetest/issues/1243 -- /clearobjects spams errrors
13:03 PilzAdam Megaf, seems to be fixed by https://github.com/minetest/minetest/commi​t/41bc24477b732b74bd18839a9742d32bd85d1b44
13:03 PilzAdam can you confirm?
13:04 Megaf PilzAdam: if you give me a couple of hours I will confirm
13:09 Megaf It's 10:10 AM here, I will get home at 12:00
13:12 grrk-bzzt joined #minetest-dev
13:24 troller joined #minetest-dev
13:26 Zeitgeist_ joined #minetest-dev
13:26 Zeitgeist_ joined #minetest-dev
13:32 PilzAdam ~tell sapier why does pressing Esc while connecting to a server print an error message?
13:32 ShadowBot PilzAdam: O.K.
13:34 troller joined #minetest-dev
13:34 Garmine joined #minetest-dev
13:44 hmmmm joined #minetest-dev
13:59 PenguinDad joined #minetest-dev
14:08 RealBadAngel joined #minetest-dev
14:10 RealBadAngel hi folks
14:13 RealBadAngel PilzAdam, i have finally figured out problems with mergeing pulls, i can do that all in GUI, all i had to do was adding remote branches
14:14 NakedFury joined #minetest-dev
14:14 RealBadAngel now i kinda like it ;)
14:17 Megaf Hi RealBadAngel
14:18 Megaf I'm looking forward to see water reflections
14:18 Megaf that is the only shader I really care for
14:19 Megaf real time shadows would be cool too, but I think that's quite impossible at the moment
14:19 Megaf real time water reflections and real time shadows would brin minetest to a whole new level
14:19 Megaf bring*
14:20 RealBadAngel it is definitely possible, i do have irrlicht code for that, have just to copy that into our code
14:20 Megaf RealBadAngel: Would I be asking too much if you did a minetest branch or fork with that?
14:21 Megaf so we don't mess with minetest itself
14:21 Megaf I would happily test that
14:21 RealBadAngel check out my shaders branch, i am pushing there all the code
14:21 Megaf ok
14:22 Megaf my girlfriend almost broke with me yesterday because I'm giving more attention to minetest than I'm giving to her...
14:22 RealBadAngel atm theres pushed change that unites nodes shaders, picking shaders per tile, not the node, and introduced water surface shader
14:23 Megaf That's about my gf is a joke...
14:23 ShadowNinja Comments?: http://ix.io/bTD/diff
14:23 troller joined #minetest-dev
14:24 Megaf ShadowNinja: I'm not really sure about that notification "A sapling grows into a tree"
14:24 Megaf Is that something really important that an admin shoud be awere of?
14:24 Megaf aware*
14:25 RealBadAngel ShadowNinja, why not using l-systems birches?
14:25 ShadowNinja Megaf: Yes, it told me what was happening when my server started eating CPU.
14:25 RealBadAngel theyre way nicer
14:26 Megaf do saplings eat CPU now?
14:28 ShadowNinja RealBadAngel: Thet could probably be done.  Can you give me a treedef that is almost exactly identical? (They were a bit bigger last I checked)  I'll also need a jungletree one.
14:30 proller joined #minetest-dev
14:34 RealBadAngel ShadowNinja, sure, hold on
14:36 RealBadAngel https://github.com/VanessaE/moretr​ees/blob/master/tree_models.lua#L1
14:36 RealBadAngel and i meant beeches not birches ;)
14:36 RealBadAngel and thats for jungle trees
14:36 RealBadAngel https://github.com/VanessaE/moretre​es/blob/master/tree_models.lua#L185
14:37 RealBadAngel and yes, beeches are 2-3 nodes higher than default sticks with leaves
14:40 CraigyDavi joined #minetest-dev
14:43 ShadowNinja RealBadAngel: axiom=nil
14:43 ShadowNinja For the jungletree.
14:44 RealBadAngel ouch, thats because there are several possible types
14:45 RealBadAngel https://github.com/VanessaE/more​trees/blob/master/init.lua#L199
14:45 RealBadAngel theyre here
14:49 ShadowNinja RealBadAngel: Which one is most similar to the default one?
14:50 RealBadAngel default one has extra nodes at bottom, none of l-systems one have them
14:51 RealBadAngel i can make a model that will copy default one
14:52 CraigyDavi joined #minetest-dev
14:53 RealBadAngel on the other hand, thats whats done in moretrees is good, it brings variety to jungle trees, they can grow different sizes
14:53 RealBadAngel from single nodes trunks, small to double and very high
14:56 RealBadAngel going out shopping, brb
15:05 Jordach joined #minetest-dev
15:45 tomreyn joined #minetest-dev
15:46 twoelk joined #minetest-dev
15:50 CraigyDavi joined #minetest-dev
15:52 twoelk I think the "default" trees should be harvestable without any help. So no building stairs or such but everything reachable from the ground, afterall they are normally the first tree a newbie encounters
15:55 Calinou joined #minetest-dev
15:59 Zeitgeist_ joined #minetest-dev
15:59 Zeitgeist_ joined #minetest-dev
16:08 Eater4 joined #minetest-dev
16:17 PilzAdam joined #minetest-dev
16:30 jin_xi joined #minetest-dev
16:31 RealBadAngel twoelk, lsystem beech is "reachable" but way nicer
16:31 RealBadAngel it is meant as replacement for default ones from the very start
16:35 twoelk I am pretty much a fan of more_trees, actually that was the mod that made me join a public server for the first time. Was just reading the 2-3 nodes higher part and remembering my first steps in singleplayer long befor I played online and could ask for help
16:35 Megaf PilzAdam: I will try that commit in a couple of minutes, backing up my server now
16:36 pi1 joined #minetest-dev
16:37 Megaf joined #minetest-dev
16:37 rubenwardy joined #minetest-dev
16:39 twoelk Having played on VE-Minetest for a while I keep repairing the jungle trees that are allready too difficult to harvest for the new/casuall/veryyoung player, leaving lots of floating trees around
16:39 RealBadAngel who said cuttin the jungle is easy?
16:42 proller joined #minetest-dev
16:54 Megaf PilzAdam: spamming is gone indeed
16:54 proller joined #minetest-dev
16:54 twoelk the first game should be easy or the newbie will be lost in some dark corner under a jungle tree sobbing, never to be seen again
16:56 Megaf I wish a tree would fall down when we cut it's trunk
16:57 twoelk there are several mods that do that iirc
16:59 Megaf I want to open a new issue but I'm not sure if it's an issue or a "feature"
16:59 Megaf when
17:00 Megaf when you ran /clearobjects server do stop responding to any request. I couple of folks attemped to connect to my server now while It's doing the clearing and they could not complete the request
17:00 Megaf 13:57:42: ERROR[ConnectionSend]: con(4/1)RunTimeouts(): Peer 5 has timed out. (source=peer->timeout_counter)
17:00 Megaf 13:58:53: ERROR[ConnectionSend]: con(4/1)RunTimeouts(): Peer 6 has timed out. (source=peer->timeout_counter)
17:01 Megaf I don't understand this behaving
17:02 troller joined #minetest-dev
17:03 Megaf http://paste.debian.net/95720/
17:03 Megaf troller: does clearobjectes freezes freeminet too?
17:03 troller yes
17:04 troller its stupid feature
17:05 troller useless
17:05 Megaf The way it works is not right
17:05 troller why you use it?
17:05 celeron55 https://github.com/minetest/minetest/pull/490
17:05 Megaf troller: becuase a lot of objects floating around wil decrease performance
17:06 celeron55 does this conflict with something in the works at the moment? if not, is there a reason to never merge it if it is rebased?
17:06 troller Megaf, in freeminer it not decreases performance, and no need to run it
17:07 troller and it works only with 4096 blocks - no sense on big server
17:07 Megaf celeron55: That's a little scary actually
17:07 celeron55 Megaf: it's unarguably useful for certain interesting things
17:08 celeron55 i think it needs a server-wide configurable whitelist and in singleplayer games it should interactively ask for permission
17:08 celeron55 after that, the security issues are small enough
17:08 Megaf What about some official thing from minetest.net? I mean, a whitelist on minetest.net, like, whitelist.minetest.net where the mods could get data from
17:09 Megaf it could be something integrated with mods githubs
17:09 celeron55 the mods are likely to want to eg. post schematics and other creations to somewhere, and download similar things
17:10 celeron55 and there could be uses for eg. synchronizing something between server instances (whatever an admin can come up with)
17:10 Megaf I'm not sure, that could generate much overhead, in network and cpu
17:11 Megaf Don't get me wrong, that's really useful and quite fantastic, but so can be the drawbacks
17:12 celeron55 useful and fantastic drawbacks 8)
17:14 troller joined #minetest-dev
17:16 rsiska joined #minetest-dev
17:24 CraigyDavi joined #minetest-dev
17:25 troller joined #minetest-dev
17:31 Calinou https://forum.minetest.net/v​iewtopic.php?f=15&t=9116 → the term “Vote” is not that good
17:31 Calinou it's a consensus
17:32 BlockMen im moving the rest of /old to "/doc/old" now
17:41 Megaf Calinou: that voting should take longer, my game is not ready yet
17:41 Megaf I'm making a game for Raspberry Pis and really low end computers
17:42 Calinou "Minetest 8086"
17:43 Megaf heh
17:43 Megaf That could be a good name
17:43 RealBadAngel celeron55, https://www.youtube.com/watch?v=hubBBzh_190
17:43 Megaf It's called rpi at the moment
17:43 iqualfragile joined #minetest-dev
17:43 RealBadAngel ^^^ up to date shaders
17:45 RealBadAngel Megaf, offtopic, can RPI run android?
17:45 Calinou I don't think so
17:45 Calinou it would need suitable drivers
17:45 sfan5 it can
17:45 Megaf RealBadAngel: Yes it can
17:46 Megaf there are somebuilds for it
17:46 RealBadAngel cool
17:46 Megaf but they are really slow due the GPU blob
17:46 RealBadAngel i was thinkin about some business
17:46 sfan5 that has nothing to do with gpu being in a blob
17:47 RealBadAngel putting rpi into keyboard case, giving it an os and selling it
17:47 BlockMen fuck, why is the doc folder not removed?
17:47 Megaf RealBadAngel: http://androidpi.wikia.com/wiki/Android_Pi_Wiki
17:47 BlockMen oh, it is
17:47 BlockMen lol
17:47 BlockMen *and i ment the old folder^^
17:47 Megaf RealBadAngel: what for? It's really slow as desktop when running Linux
17:48 Megaf It's pretty fast using ReactOS
17:49 celeron55 22
17:49 celeron55 oops
17:50 Megaf RealBadAngel: http://www.raspberrypi.org/android-4-0-is-coming/ http://androidpi.wikia.com/wiki/Android_Pi_Wiki
17:50 Megaf That's it
18:02 Megaf the /clearobjects still running...
18:03 Calinou your PC works with ReactOS? :o
18:03 Megaf Calinou: oops
18:03 Megaf wait a minute,
18:04 Megaf s/It's pretty fast using ReactOS/Raspberry Pi as desktop is pretty fast using RiscOS
18:11 Megaf Calinou: and yes, a computer can work with ReactOS
18:11 Megaf but earlier on I meant RiscOS on Raspberry Pi
18:48 BrandonReese joined #minetest-dev
18:52 Megaf joined #minetest-dev
19:01 BrandonReese joined #minetest-dev
19:12 RealBadAngel celeron55, have you saw the video?
19:19 celeron55 RealBadAngel: the water should get an orange in place of the white when the sun is setting
19:20 celeron55 or in place of blue
19:20 VanessaE follow the tint of the tone map
19:20 VanessaE maps*
19:24 celeron55 (in reality practically all of water's color comes from the sky and clouds and sun and whatever else that it reflects)
19:24 celeron55 in minetest it tends to be painted blue no matter what's above it 8)
19:25 RealBadAngel this is temporary
19:25 RealBadAngel water will have reflections
19:26 RealBadAngel now reflection texture is just set to all blue
19:27 VanessaE did you fix that glitch with animated water textures throwing off the speed of the water rippling shader?
19:28 RealBadAngel huh?
19:28 Anchakor I wonder how crappy will flowing water look compared to this :)
19:29 RealBadAngel crappy ;)
19:29 VanessaE remember, the thing where the water rippling shader would go apeshit under minetest_game (except when paused) but not under minimal because _game has animated textures?
19:29 Shardvexz joined #minetest-dev
19:29 RealBadAngel but i started to read about Bezier curves to implement real flowing liquids
19:31 RealBadAngel VanessaE, with the changes i made this has no effect on shaders
19:31 VanessaE ok good.
19:31 VanessaE also, did you fix the issue with the *bottom* surface of the water being moved around via the "waving" shader?
19:31 kaeza joined #minetest-dev
19:31 RealBadAngel when water surface is detected the tile for it uses own shader with disabled animations
19:32 VanessaE (I mean the -Y side of the block)
19:32 RealBadAngel and pause doesnt stop shaders animations timers
19:33 RealBadAngel hold ur horses
19:33 RealBadAngel shaders now work per tile, not the node
19:33 VanessaE ok
19:34 RealBadAngel water surface shader is supposed to work only on [0] tile
19:34 VanessaE ok ok, do this:
19:34 VanessaE build a 3x3x3 cube of glass.  hollow the center out and fill it with water.
19:34 VanessaE turn on the waving water shader and observe the bottom/-Y surface of the water
19:35 RealBadAngel you mean that water node that have another one over it shouldnt be affected by waving?
19:36 VanessaE no.  I mean that the bottom-most -Y surface of the water node, under the current version of the shader, will also move up/down to match the top +Y surface
19:37 VanessaE what should happen is that the bottom shouldn't move at all if there's no more water below it.
19:38 celeron55 there should also be some kind of logic for not making water underground be in the sky's colors but instead something more like stone and dirt
19:38 RealBadAngel without it we wont get "waves at beachside" effect
19:38 celeron55 and not wave around
19:38 VanessaE celeron55: +1
19:39 RealBadAngel theres such logic already
19:39 RealBadAngel sunligt_seen or so
19:39 VanessaE I thought the same
19:39 celeron55 does it work? that's good
19:39 VanessaE RealBadAngel: wait nonono, you're talking about the +Y side of the node
19:39 VanessaE that's fine
19:40 VanessaE I'm talking about the -Y side, or really the -Y edges of the X/Z sides since there is no -Y face on a water source anyway
19:40 VanessaE those edges should *not* move if there's no more water below them
19:40 VanessaE but the top surfaces and edges should move.
19:40 RealBadAngel btw, lava: https://www.youtube.com/watch?v=zxUXNNFpPpc
19:42 BlockMen joined #minetest-dev
19:42 Anchakor like there is no sense in having waves and tiny pocket of air at the bottom of glass of water
19:43 VanessaE Anchakor: exactly.
19:43 Anchakor (when it is full)
19:46 PenguinDad RealBadAngel: just noticed that waving water is broken on your shaders branch http://ix.io/bVL
19:48 RealBadAngel PenguinDad, i know, i have this fixed on local already
19:48 VanessaE RealBadAngel:  I mean this issue:  http://digitalaudioconcepts.com/vanessa/extra/Scre​enshot%20-%2004252014%20-%2003%3a47%3a27%20PM.png
19:49 VanessaE the red wavy lines should imply the motion of the water, the }--- implies how much overlap there should not be with the space the water sits in :P
19:49 RealBadAngel that was an issue with #define variables not being able to use with #if
19:50 RealBadAngel VanessaE, splitting face shaders should do the trick i think
19:50 VanessaE ok
19:50 RealBadAngel so bottom cannot wave
19:50 VanessaE but it's a solvable issue without taking a performance hit?
19:50 RealBadAngel am i correct?
19:51 VanessaE yes, bottom should never wave
19:51 VanessaE and if possible, bottom should be exactly aligned to the bottom of the node where it belongs.
19:51 RealBadAngel actually sorting out shaders before game starts speeds things up
19:51 VanessaE because honestly, this looks weird :P
19:52 RealBadAngel and thats the main change in the forthcoming shaders update
19:52 RealBadAngel shaders are being compiled on textures build
19:52 RealBadAngel ONCE
19:53 RealBadAngel not per every mesh update, which was just sick
19:53 RealBadAngel also i have trashed the code that allowed for asynced shaders requests
19:54 RealBadAngel so, no more timed out errors
19:55 RealBadAngel sorry karhl, but gfx is not as easy as two threads calculating something
19:56 RealBadAngel current situation can be compared to the situation when you went outside and waiting for a miracle to get dressed
19:56 RealBadAngel before someone noticed youre naked
20:00 RealBadAngel atm compiling of all nodes shaders takes circa 1.5s, thats 5(materials)x16(drawtypes)
20:00 RealBadAngel so many instances
20:01 RealBadAngel shaders are not using drawtypes by now but ive allowed them to do so
20:02 RealBadAngel so torches, rails or fences can have own shaders
20:02 RealBadAngel within one source shader ofc
20:05 RealBadAngel celeron55, you havent commented the vid... and im really curious what do you think about it
20:07 celeron55 i did comment
20:08 RealBadAngel oh, indeed
20:09 RealBadAngel but youre next that expected water reflections
20:10 RealBadAngel shit i cant code everything at once ;)
20:13 us`0gb joined #minetest-dev
20:13 hmmmm joined #minetest-dev
20:20 Guest38529 joined #minetest-dev
20:23 Guest38529 RealBadAngel: You are doing well, you can code all at twice
20:23 Shardvexz joined #minetest-dev
20:30 RealBadAngel yeah sure. but seriously changes ive made lately let shaders work on the nodes.
20:30 RealBadAngel and the other shaders work on smthg else
20:31 RealBadAngel so, in the near future all kind of postprocess shaders will become possible
20:35 Megaf RealBadAngel: I'm looking forward for water reflections and real time shadows
20:37 Megaf RealBadAngel: If we do realtime shadows, would those shadows be shadows from anything that emits light or only sun shadows?
20:37 RealBadAngel oh, another qest given...
20:38 RealBadAngel Quest
20:38 RealBadAngel but seriously. i was thinkin about implementing ray marching here
20:39 RealBadAngel but atm im not quite sure what will be the best in sense of performance
20:39 Megaf RealBadAngel: I'm talking so much about water reflections and real time shadows because they will make the game looks much cooler and real and beautiful and less artificial, and you dont like that, you could just disable it
20:40 Megaf RealBadAngel: ray marching could be interesting too
20:40 RealBadAngel taoki suggested using current lights as shadows map
20:41 RealBadAngel which makes a lot of sense, since server have to calculate light level anyway
20:41 Taoki RealBadAngel: My suggestion is using the current lighting (param2 geometry coloration map) as a black and white image, as a pass in deferred lighting. You can then extract dynamic lights from it
20:41 Taoki As in, use it as a mask for lights
20:42 RealBadAngel so deffered light map
20:42 Taoki But yeah. Using it is needed both because visible light must match the value of param2 (eg: fetching light value at node position in Lua), and so sunlight doesn't shine in caves
20:42 Taoki Yeah
20:42 Taoki I think deferred lighting is best choice
20:43 Taoki Oh... and I think two deferred maps like this need to be made. One for sun / moon light, and another for local lights (possibly one per local light, not sure how much FPS that would cost though). Depends on what's needed however
20:44 RealBadAngel we will see
20:44 Megaf WHat about player shadows?
20:45 RealBadAngel thats pretty trivial
20:45 RealBadAngel it is post process effect
20:45 RealBadAngel once sun is here it can be made
20:46 RealBadAngel notice that in my branch sun is already there as light source
20:46 RealBadAngel always 900 nodes higher
20:46 RealBadAngel thats used for bumpmapping
20:47 VanessaE Ok, Dreambuilder Game is updated.
20:47 VanessaE so now if there's any inclination to add it (my) game to the official builds, well at least it finally has a name now :)
20:47 VanessaE I just need to make a title card for it.
20:50 Taoki RealBadAngel: Any idea when there will be any code for basic dynamic lighting? Don't wish to rush, just eager... it will probably be the most awesome thing in Minetest since Minetest was created :)
20:51 RealBadAngel well, engine allows that already
20:51 RealBadAngel it is our code that misses that feature
20:51 Taoki Irrlicht uses dynamic lighting, yes. It's actually the Minetest code that currently uses an ugly hack for lighting (vertex colors)
20:52 RealBadAngel actually, our code can be disabled easily
20:53 RealBadAngel oneliner in mapblock_mesh
20:53 RealBadAngel and adding sun
20:53 RealBadAngel i mean set all the material's lighting to true
20:54 RealBadAngel and put there omnidirectional lightsource
20:55 RealBadAngel ive used fixed one for testing purposes
20:55 Taoki Not that easy sadly. First we need to disable vertex colors set by param2... this part is easy itself. Then we need to add an Irrlicht light source for both sun / moon and any light node like torch... also not the hardest thing. However, we need to bake the geometry with param2 for each light, and send it to the shader core to extract the dynamic light from it. Which must then be added on top of..
20:55 Taoki .. the textured geometry
20:55 VanessaE ok, title card created and uploaded.
20:55 VanessaE I think that should do it for my game.
20:56 RealBadAngel Taoki, sun is easy
20:56 RealBadAngel local lighsources are pain in the ass
20:56 Taoki Yeah
20:56 Taoki Well,
20:57 RealBadAngel but hell, for that we do already have the logic
20:57 Taoki We can make it so only sun / moon light are masked by the param2 map. But then light sources will shine through walls, and might not always match the value returned by getLight() in Lua (or what that function is)
20:58 Taoki After all, sun and moon are lights with infinite radius... local lights would have finite distance
20:58 RealBadAngel yeah
20:58 Taoki Still, that means local lights will shine through walls. I don't think that's very acceptable
20:58 RealBadAngel and for finite and local ones current logic will do
20:59 Taoki So best way is to generate the param2 geometry map image for each local light, and use that to mask it... the light itself having an infinite distance past that.
20:59 Taoki But what if there are 20 local light in view? How much can the video card handle mixing black & white images per frame without very low FPs
20:59 RealBadAngel that would cost us too much
20:59 Taoki yeah
21:00 RealBadAngel 2-3 lights sources are enough
21:00 Taoki Best way might be two param2 passes. One for sun / moon light, other for local lights if any are present. Local lights might still shine in each other's raduis, but it's a much cheaper way
21:00 Taoki If there will be a finite number of lights, it should certainly be a setting IMO
21:01 Taoki Peple can set it based on video card performance
21:01 RealBadAngel hold on
21:01 Taoki yes
21:01 RealBadAngel my idea is a bit different
21:01 * Taoki listens
21:01 RealBadAngel use real, hardware light from irrlicht
21:01 RealBadAngel sun/moon
21:01 RealBadAngel as basis
21:02 RealBadAngel then add ours, but not as real light sources
21:02 RealBadAngel but usual way, as gl_color
21:02 Taoki What do you mean ours though?
21:02 RealBadAngel current lighting system
21:03 Taoki Irrlicht light sources are the way to do dynamic lighting. Though it might be possible to do it in shaders entirely... still would tend toward Irrlicht's system
21:03 RealBadAngel which could be brought down just to lightlevel
21:03 Taoki Oh, yes. I think you mean the same thing as me
21:03 RealBadAngel i mean no soft lighting or whatever its called
21:04 RealBadAngel just the light level at pos
21:04 Taoki Soft lighting should still work
21:04 Taoki ah
21:04 Taoki Hmmm...
21:04 proller joined #minetest-dev
21:04 Taoki param2 value at position would be cheaper and better than per-vertex brightness, yes
21:04 RealBadAngel hi proller
21:05 proller hi
21:05 RealBadAngel i can use that value as light
21:05 Taoki Still, I'm not sure what the shader code will do with it.
21:05 Taoki We need to render a black & white version of the geometry based on the value of param2 at each position. Different render-pass... then post-process that as a modifier to light
21:05 Taoki There would be at least 3 render passes overall
21:06 RealBadAngel proller, have you rebased my sun/moon patch?
21:06 Taoki Or maybe 2
21:06 RealBadAngel its pretty different from what you have merged before
21:06 RealBadAngel Taoki, no
21:07 RealBadAngel light calculations on the server side ~= render passes
21:07 EvergreenTree joined #minetest-dev
21:07 RealBadAngel !=
21:08 proller RealBadAngel, merged current master in freeminer
21:08 Taoki RealBadAngel: http://1.bp.blogspot.com/-G6xtoRnB2x8/Uf8u4V7HusI​/AAAAAAAAAX0/x-QFBdEAYq0/s1600/Display+sheet.jpg Pass 1 would be untextured geometry, similar to "Ambient Color" in that image. Pass 2 would be untextured geometry with param2 map, similar to "Depth". Pass 3 would be lighting only, similar to "Ambient Occlusion"
21:09 Taoki Then in the shaders code, Pass 1 is rendered, with Pass 3 using ADD blending on top of it, but with Pass 2 SUBTRACTED from it. Or something like that I think
21:09 Taoki I think add and subtract are the right names for image blending too
21:10 Taoki At least what I was thinking... my ideas might not always be the best
21:10 Taoki erm... *Pass 1 would be TEXTURED geometry :P
21:12 Taoki RealBadAngel: Actually, I don't really get how this would work out. Would the render passes have to be *created* server-side, or shader-side? I mean, the server knows the light value at each node posibion... but could the shader code use such information to generate the lighting map?
21:15 RealBadAngel in fact i do not care
21:15 RealBadAngel engine provides me color per vertex
21:15 RealBadAngel which contains both celestial lights and artificial
21:16 Taoki Hmm... that is good
21:16 Taoki Still, how can you tell local lights from one another, and how much each contributes to the light value at a given location?
21:16 RealBadAngel my idea is just to disable celestial there and let irrlicht handle them
21:16 RealBadAngel ^^ as above
21:17 RealBadAngel make engine ones local
21:17 RealBadAngel or treat them as local
21:17 Taoki Still can't understand... sorry if I'm stupid >_<
21:17 Taoki Both sun and moon as well as local lights should use hardware lighting, so both would have to work the same way
21:17 Taoki Only that sun / moon light have infinite radius
21:18 RealBadAngel sun/moon will use hardware lighting
21:18 RealBadAngel to lit the scene
21:19 Taoki What about light caused by torches?
21:19 RealBadAngel mt engine will calculate just local lightsources as glowing stuff, torches etc and pass them as vertex colors to the shaders
21:20 RealBadAngel so, the scene will sum real lights and artificial ones
21:20 Taoki But not use hardware lights for local lights too?
21:20 RealBadAngel no
21:20 Taoki I see. Was hoping that could be done as well
21:21 RealBadAngel we do need light as a node property
21:21 RealBadAngel light level
21:21 RealBadAngel for that current code is enough
21:21 Taoki Torches can't cast directional light then. Unless they're used very smartly in the shader code. And even so I don't think they can then cast dynamic shadows
21:22 Taoki Also, I was hoping this would finally allow for colored local lights
21:22 RealBadAngel coloured lights?
21:22 RealBadAngel no problemo
21:23 Taoki Yeah. So torches can also have a color assigned to the lighting... for example torches that cast green light
21:23 Taoki If that's possible it's a good thing :)
21:23 RealBadAngel irrlicht allows that
21:23 RealBadAngel just we are not using it
21:23 Calinou hardware lighting: the notion of light level dies?
21:23 Calinou don't forget about gameplay.
21:23 Taoki Yes, but for dynamic lights. You said you don't plan to add HW light sources for these
21:24 Taoki Calinou: That's what we're saying. We still need light to represent param2 (light value of node)
21:24 Taoki Which is the tricky part
21:24 RealBadAngel color can be also a parameter
21:25 RealBadAngel atm its just white
21:25 Taoki yeah
21:25 RealBadAngel Calinou, we will leave light level for gameplay
21:25 Taoki But then the engine / shaders must also tell light sources from one another. A torch casting red light might be near one casting green light
21:25 RealBadAngel and use it to lit the scene
21:26 Taoki The red and green light sould combine
21:26 Taoki *lights would
21:26 RealBadAngel we wil see
21:26 Taoki yeah
21:27 RealBadAngel one thing i have to say
21:27 RealBadAngel engine allows that for sure
21:27 Taoki RealBadAngel: Either way, local lights and light colors can come after sun and moon HW lights are in first. You could make celestial lights hardware only at first, then later make code for locals too
21:27 RealBadAngel its just us to adapt
21:27 Taoki Irrlicht yes :) Though shaders are needed to change intensity based on light level
21:28 Taoki As Irrlicht isn't meant to work with our contept of lighting. It doesn't understand our light levels, smooth lighting, etc
21:28 Taoki Se we need to take the whole lighting pass from irrlicht, and add the light level pass as a mask to it
21:28 RealBadAngel thats what i said already
21:28 Taoki Right... sorry
21:28 Taoki yes
21:28 RealBadAngel we just need light level at a position
21:29 RealBadAngel smooth lighting code is obsolete
21:29 proller and very slow and buggy
21:29 RealBadAngel and that is needed for things like light sensors etc
21:29 RealBadAngel nothing else
21:30 RealBadAngel or maybe plants growing code
21:30 RealBadAngel but definitely not litin up the scene
21:31 RealBadAngel also server should not care about lights
21:31 RealBadAngel it makes it slow
21:32 RealBadAngel longer i develop GFX stuff, more stupid things i find
21:32 Taoki HW lights should still match the light level, so the server returns realistic and correct values
21:32 RealBadAngel thats the real problem
21:33 RealBadAngel how to bound HW lights and lightlevel
21:33 markveidemanis joined #minetest-dev
21:34 RealBadAngel my idea is to use existing code
21:34 RealBadAngel and cut from there everything that have impact on lighting
21:35 RealBadAngel just leave there light level calculations
21:35 RealBadAngel but im not sure how the extra light sources will behave then
21:36 proller player with torch must be light source too
21:36 RealBadAngel so atm i would rather mix current system with celestial and omnidirectional ones
21:37 RealBadAngel proller, i am aware of that pull. but as c55 said, for those who cant use shaders, thats disadvantage
21:37 kahrl joined #minetest-dev
21:37 RealBadAngel and thus cannot be allowed
21:37 proller wich pull?
21:38 RealBadAngel wield light
21:38 RealBadAngel until its aviable for players without shaders too, i wont agree to merge it
21:38 markveidemanis left #minetest-dev
21:39 RealBadAngel hi kahrl
21:44 kahrl hi RBA, what's up?
21:46 RealBadAngel kahrl, im makin a mess in shaders.cpp, is that your code or taken from somewhere else?
21:47 kahrl I took tile.cpp and deleted most of it and changed the rest to work with shaders instead of textures
21:47 kahrl not sure how much it changed since I wrote it
21:48 RealBadAngel lemme explain what im doing right now
21:48 RealBadAngel code allowed to compile shaders at runtime, on mesh updates
21:49 RealBadAngel i restricted that only to startup
21:49 RealBadAngel and main thread only
21:50 RealBadAngel so ive trashed threads aware code
21:51 kahrl I don't remember, are all required shaders always known on startup?
21:51 RealBadAngel yes
21:51 kahrl if so that sounds like a good idea
21:51 RealBadAngel on texture refreshing
21:51 RealBadAngel in nodedef.cpp
21:51 kahrl unless it causes huge startup delays like preload_item_visuals does
21:52 RealBadAngel no it doesnt
21:52 kahrl if there's a huge number of items
21:52 RealBadAngel we do have 5 materials and 16 drawtypes
21:52 kahrl ok good
21:52 RealBadAngel compiling 5*16 shaders takes circa 1s
21:53 RealBadAngel and even drawtypes are made in excess
21:53 kahrl doesn't the current shader code answer most shader request through the material cache?
21:53 RealBadAngel no use yet, but shader can be aware of dratype used
21:53 kahrl requests*
21:54 kahrl so doesn't it only do cross-thread requests mostly on startup?
21:54 RealBadAngel not really
21:55 RealBadAngel i do compile all the possible matches
21:55 RealBadAngel material/drawtype
21:55 RealBadAngel and then call just for ID
21:56 RealBadAngel by now, it was possible to time out
21:57 RealBadAngel i saw situations that game started, shaders were timed out and applied on next mesh update like 10-15s later
21:58 kahrl hmm
21:58 RealBadAngel more complex mods, more media, larger lags
21:59 RealBadAngel my approach eliminates that
21:59 RealBadAngel no go when theyre not ready
21:59 kahrl I thought sapier made it so it loops until the shader is received
21:59 kahrl no timeout essentially
22:00 RealBadAngel i saw that many times
22:00 RealBadAngel he failed then
22:01 RealBadAngel anyway, shader should be compiled just once
22:01 RealBadAngel and then just fetched
22:01 RealBadAngel and theres absolutely no point doing that in game
22:02 Taoki RealBadAngel: What if we will someday be able to turn them on or off from the pause menu? Like it should be possible with most settings really
22:02 RealBadAngel why not?
22:02 Taoki Perhaps add an exception to be ready for that if it will ever happen
22:02 RealBadAngel just set the flag
22:02 Taoki So shaders aren't applied only at process startup
22:02 Taoki Sure :)
22:02 RealBadAngel code is aware of the flags
22:02 Taoki Awesome then
22:03 RealBadAngel i was talkin about applying/compiling shaders at runtime
22:03 RealBadAngel which was just insane
22:03 kahrl you'll have to recreate all the meshes though when you change the setting
22:03 RealBadAngel no
22:04 RealBadAngel i wont do anything
22:04 kahrl what about the vertex colors and day-night transition?
22:04 RealBadAngel just call mesh update with new state of the flag
22:05 RealBadAngel which is done all the time
22:05 RealBadAngel shaders doesnt modify them
22:05 RealBadAngel so turning them on/off doesnt harm
22:07 RealBadAngel kahrl, plz check the current code: https://github.com/RealBadA​ngel/minetest/tree/shaders
22:07 kahrl any part in particular?
22:08 RealBadAngel nodedef.cpp
22:08 RealBadAngel thats something new here
22:09 RealBadAngel ive moved shaders generation from mesh updates to textures refresh
22:10 RealBadAngel here, because i am able to set shader there per tile, not the node
22:10 kahrl I see
22:10 RealBadAngel also the code refreshes all the possible ones
22:11 RealBadAngel so leaving this part makes me sure i do have everything compiled
22:13 Taoki proller: https://forum.minetest.net/viewtop​ic.php?f=5&amp;t=6418&amp;p=139374 Are you (still?) someone who could attempt to help with fixing this and getting that code in? Hoping for an year to see that in master
22:13 Taoki RealBadAngel: ^ Might wanna look at that too sometime, since you are good at understanding the engine and graphics system
22:14 kahrl RealBadAngel: I think that's good, don't see any obvious problems with your approach anyway
22:14 Taoki And also don't avoid merging something excessively I guess :P
22:14 kahrl as long as we don't merge meta_set_nodedef ;)
22:14 proller Taoki, it works acceptable in freeminer
22:14 Taoki proller: No more memory leaks?
22:14 proller maybe small
22:15 Taoki Is the branch still compatible with latest master also?
22:15 Taoki ok
22:15 Taoki Really need to get someone to merge it in master too (disabled by default)
22:15 sapier joined #minetest-dev
22:15 proller firs you must kill Pilz**** ;)
22:16 RealBadAngel as a workaround for meta_set_nodef i think some new drawtypes will do
22:16 Taoki Does PilzAdam still care about it (as in being against it)?
22:16 PilzAdam proller, small? they were quite significant on my system
22:16 RealBadAngel wirelike, tubelike
22:16 sapier ShadowNinja: I don't know exactly but I guess it's because celeron didn't want to wait for regular shutdown on local host
22:17 Taoki IIRC, latest version of the VBO branch had much smaller memory leaks
22:17 sapier that's why shuting down sets peer timeout to 0.1 second causing immediate peer droping
22:17 RealBadAngel but imho ability to change nodedef on the fly could be the best solution for modding
22:18 Taoki Actually, not leaks. From what I remember it's just high memory usage due to how mapblocks are handled or something
22:18 sapier smaller leaks ... :-) a leak is a leak no matter how big
22:18 Taoki Which could be reduced if chunks of the world were stored for a smaller amount of time client-side.
22:18 RealBadAngel hold on a sec
22:18 Taoki sapier: Yeah, just remembered it's not a leak after all. At least not with a simple fix that was added in the latest version. Just high memory usage
22:19 RealBadAngel tiles do already have special fields
22:19 Taoki There was a memory leak at first, because the mesh wasn't removed with the object. That was solved
22:19 sapier aren't we talking about those engine memory leaks?
22:19 RealBadAngel why the fuck nodes cant?
22:19 Taoki sapier: We're talking about the legendary VBO code again :)
22:20 PilzAdam Taoki, ideally the memory (RAM) should go down, since we are storing the mesh in vRAM now
22:20 sapier those not beeing leaks as in lost memory but as in use even more and more memory by engine
22:20 kahrl RealBadAngel: special fields?
22:20 PilzAdam +usage
22:20 Taoki PilzAdam: I think it does once a chunk of world is unloaded from the client at least.
22:20 RealBadAngel kahrl, used for animations
22:20 Taoki But server has to re-send it then
22:20 Taoki When the player re-explores an area
22:21 kahrl RealBadAngel: and what do you mean by adding that to nodes?
22:22 PilzAdam Taoki, I am talking about mesh data
22:22 Taoki PilzAdam: But given that proller and users of the Freeminer branch (as well as myself when I last tested VBO I believe) are noticing a more acceptable memory usage, maybe it can still be added with a big warning on it or something? Myself and others would still love to be able to use it. As long as it won't crash my system, I'm fine with the extra memory usage :)
22:22 Taoki Ah. Mesh data isn't being removed?
22:22 RealBadAngel kahrl, nodeboxes? textures?
22:22 Taoki IIRC that was fixed
22:23 kahrl RealBadAngel: I don't get it
22:23 RealBadAngel kahrl, anything that could ovveride definition
22:24 kahrl RealBadAngel: that's essentially meta_set_nodedef, isn't it?
22:24 Taoki But hell... Second Life uses 3GB of RAM for me sometimes, and it still works perfectly well. I can totally live with Minetest reacing 1GB, especially with a setting I choose to enable for greater FPS
22:24 RealBadAngel yes
22:25 RealBadAngel thats needed as hell, no matter how
22:25 Taoki Though in my last test I remember I think VBO had Minetest use less than that... depends how much you explore is all
22:25 kahrl unfortunately it turned out to be too slow, iirc
22:25 kahrl not surprising if it has to check metadata of each node on mesh generation
22:26 RealBadAngel its done in minecraft
22:26 RealBadAngel and works
22:26 RealBadAngel in JAVA
22:27 Taoki Sssh... never say something we want to do in Minetest is done in Minecraft. We might get into trouble ;)
22:27 RealBadAngel shall i paste link to minecraft sources>
22:27 kahrl really? when was that added?
22:27 RealBadAngel ?
22:27 RealBadAngel its on github
22:27 kahrl (I didn't really follow minecraft development after the official release)
22:28 RealBadAngel at least server's code is on github
22:28 RealBadAngel https://github.com/RealBadAngel/mc-dev
22:29 kahrl well the server doesn't render stuff ;), at least I hope it doesn't
22:30 RealBadAngel reading how others have done something is not violating any license
22:32 kahrl as long as you don't subconsciously write the code you've read before
22:32 kahrl which has happened before... some of the mingw guys avoid reading any part of the official platform SDK for that reason
22:34 kahrl anyway... any link that explains the meta_set_nodedef thing in minecraft?
22:35 kahrl I could maybe try it out (without reading the code) in the game
22:38 BlockMen left #minetest-dev
22:39 VanessaE what about the thing that Nore already did?
22:41 kahrl #1118?
22:41 ShadowBot https://github.com/minetest/minetest/issues/1118 -- [WIP] Meta set nodedef by Novatux
22:41 VanessaE that's the one
22:41 VanessaE I've played with it
22:41 VanessaE it worked surprisingly well at the time
22:42 kahrl I guess the [WIP] is why it wasn't merged :)
22:43 VanessaE if I have the right package in mind, this mod http://digitalaudioconcepts.com/vane​ssa/hobbies/minetest/microblocks.zip  would be the one I used to "demo" 1118.
22:44 RealBadAngel it could solve many issues we have
22:45 RealBadAngel and reduce nodes definitions greatly
22:45 VanessaE the mod has a bug or two but it demonstrates the concept quite well
22:46 RealBadAngel mesecons/pipeworks/technic could reduce node definitions by thousands
22:48 kahrl but if each node of wire or pipe has its nodebox stored in metadata, how big will mapblocks be? (mostly thinking of bandwidth problems here, though map storage could be a problem too)
22:48 VanessaE I'm not sure that's too much of a concern
22:49 VanessaE the biggest pipelines I've ever seen on a server are still a pittance compared to the full size of even one mapblock anyway
22:49 kahrl I'd rather allow different types of wire to be distinguished by param2
22:49 VanessaE a new drawtype is best
22:49 * VanessaE looks at RealBadAngel
22:49 RealBadAngel i can code that
22:49 RealBadAngel but that wont be universal
22:50 VanessaE different types as in, HV vs LV, vs TTL?
22:50 RealBadAngel as nores/c55 work could be
22:51 RealBadAngel see, if i code wires, still pipes wont be single defed
22:51 RealBadAngel then i can code pipes
22:51 RealBadAngel but what if we want another kind of wires/pipes/whatever?
22:51 kahrl how about allowing multiple node definitions for each node
22:52 kahrl which one is used is selected by param2
22:52 VanessaE I'd think the node name alone would be sufficient for that?
22:53 VanessaE multiple node defs?
22:53 VanessaE hm
22:54 kahrl I thought you wanted to reduce node ID usage
22:54 VanessaE hmmmm....
22:54 VanessaE well sure, of course
22:54 VanessaE but you have to be able to texture them
22:54 VanessaE and define them to do different things
22:55 VanessaE an LV cable looks and behaves differently from a HV cable, for example
22:56 kahrl well it depends on what the nodes do
22:56 VanessaE but to have a single "LV cable" node that can change shape automatically instead of having to define dozens of shapes, that's the benefit I was thinking of
22:56 kahrl I mean ABMs would still just match on node name and not on node name + param2
22:56 VanessaE just as you have one "fence" node that changes shape depending on its surrounding fence nodes.
22:56 kahrl but if the different behaviour is implemented in lua, of course lua can check param2
22:57 sapier http://de.tinypic.com/r/n6ffuu/8 what do you think about adding download rate to media download progress bar?
22:57 VanessaE right, but param2 is numeric.  that could get confusing in the code.
22:57 VanessaE sapier: +1000000000^(inf)
22:58 kahrl true
22:58 sapier but will only work for mintest download not for http mode
22:58 proller instead of speeding up connection will print download animations...
22:58 VanessaE sapier: you could calculate it for http too, as you already know when you requested a media object and when it arrives
22:58 VanessaE the speed would just be more coarse
22:59 sapier yes but that's gonna be quite tricky
22:59 sapier and to be honest I don't have real interest in http mode
22:59 VanessaE proller, if you think printing a speed to the screen is gonna slow down a network connection, you need to go back to Intro to Computing 101.
23:00 sapier well proller those numbers are real numbers not some fake pictures
23:01 VanessaE sapier: that said, it might be sufficient to just print the current speed and not the average speed.
23:01 sapier and it's not mega bit but megabyte
23:03 sapier yes seems to be a little bit big to me too
23:03 PenguinDad left #minetest-dev
23:04 sapier right now rate is only updated once per 10 seconds to cause quite less cpu usage ... means it's gonna be 0 KB at begining for 10s ... while progress bar already increases ... looks a little bit strange
23:04 sapier for this picture I had 160mb texture data to have a chance to take it
23:05 proller shut up and commit
23:05 sapier it's localhost as my internet connection can't handle more then about 300-400 kb/s
23:05 sapier average and current or current only?
23:06 proller curr
23:06 VanessaE current only I think.
23:06 sapier ok current is 10s
23:17 sapier https://github.com/sapier/minete​st/tree/add_media_download_rate it's still WIP
23:18 sapier you have to either compile non curl client or disable http fetching by adding "disable_remote_media_server = true" to config
23:20 iqualfragile joined #minetest-dev

| Channels | #minetest-dev index | Today | | Google Search | Plaintext