Time Nick Message 00:02 kaeza I would like to be able to create yet another FPS based off minetest (a la AOS/OpenSpades), so crouching would be useful for e.g. taking cover from enemy fire 00:04 paramat ah i see 00:04 pitriss kaeza: it sounds good if you bind it to some new key..:) 00:05 kaeza pitriss, currently, it adds a new "crouching" key (default is LCTRL), which BTW can also be used by mods in the same way `sneak` is used now 00:05 proller and walking to 1x1 holes 00:06 pitriss ahh ok.., then it sounds good.:) 00:11 paramat on the subject of 1x1 holes, i discovered if you attach a player to a vehicle, and reduce the visual scale of the vehicle, it makes the player smaller too, so you can drive through 1x1 passages 00:12 kaeza I recall that bug being reported, but didn't know it also affected the collision box 00:13 kaeza https://github.com/minetest/minetest/issues/419 00:15 paramat lol that image mauvebic posted 00:19 paramat hmmmm, celeron seems happy for someone else to finish the job of bringing mgv5 back, i might attempt this. it needs ores, biomes and trees added. so what biomes would you like to see in mgv5? i'm thinking mgv6 biomes plus a snow biome with snowy pinetrees. i would also like splizard's needles node, and my pinetree and pinewood nodes added to MTgame, to complete the default snow biome nodes. your opinions? 00:21 proller paramat, you need port all your mapgent to core ;) 00:23 paramat well i'm now getting my head around core mapgen code, you were right (!) c++ is not much more difficult than lua 00:23 proller you got it! 00:24 proller next point: its 100x faster 00:25 paramat yes, lua mapgen is still too slow for some servers, okay for singleplayer though 00:25 proller if you cant observe world - its not ok 01:08 paramat wow mgv5 is much better than mgv6, especially the caves. it also uses eased 3D noise 02:46 Sokomine paramat: i'm not sure about that. some servers seem to work ok with plantlife in its fullness, and that's *slow* due to the huge amount of plants it spawns and the requirements thes plants have for spawning 02:46 Sokomine i mean - if even that works ok, then most servers are probably on the faster side 05:44 RealBadAngel paramat, you could try those amplified biomes from mc 05:45 RealBadAngel they look pretty neat 06:42 hmmmm https://www.shadertoy.com/view/Ms2SD1 06:42 hmmmm hue hue hue 06:42 hmmmm shader perlin noise 06:43 hmmmm paramat, going forward I'd like all new mapgens to use the biome infrastructure 06:43 hmmmm biomes are supposed to be a thing that, going forward, mapgens all use 06:43 hmmmm like they'd all share that piece of code 06:44 RealBadAngel btw, if lua mapgens are slow, core are faster, why not make mapgens DLLs ? 06:44 hmmmm because that's a windows-specific solution 06:44 hmmmm in any case, zeno` wants that 06:45 RealBadAngel i just used windows name 06:45 RealBadAngel this way we could have fast mapgens and dont have to maintain them in mainstream 06:47 hmmmm i think it would be better overall if lua mapgens could generate without an envlock 06:48 Hunterz How many dedicated servers are hosted on windows? 06:48 RealBadAngel http://luajit.org/ext_ffi.html 06:48 Hunterz I think many server owners use linux for ther dedicated servers. 06:49 RealBadAngel this could be the way to do that 06:49 Zeno` ? 06:49 Zeno` What do I want? 06:49 RealBadAngel drop our bundled lua, embeed stable luajit and start to use its features 06:50 Zeno` Oh C mapgens? Yeah I already have those 06:50 Zeno` Not ever (likely) to push the code though because nobody seems interested 06:51 RealBadAngel Zeno`, i meant mapgen libs, which user can download and use with game 06:51 RealBadAngel this way they can be way faster and flexible 06:52 Zeno` RealBadAngel, I use on the server what I will "extensions"; the are .so files that are a hybrid of Lua and C 06:52 Zeno` So, yes, a little bit different probably 06:52 RealBadAngel technic computer is using libs since a long time already 06:53 RealBadAngel libs are compiled for both linux and winows 06:54 RealBadAngel https://github.com/minetest-technic/technic_game_2/tree/master/mods/forth_computer 06:54 Zeno` I started off just extending Lua, but that was not nice. My current solution is a "hack"; If I was to push the code I'd really want to clean it up first 06:55 Zeno` yeah that's basically one half of my hack 06:56 Zeno` this is why my server version isn't up-to-date. I need to make the hack "nice" 06:56 Zeno` one day 06:57 Zeno` I could get rid of the whole hacky part altogether if there was a fast way to populate the Lua stack with node ids (and vice versa) 06:58 Zeno` But they're not contiguous in memory, so... *shrug* 06:58 Zeno` I did mean to add something (a light C pointer) but got sidetracked 07:16 RealBadAngel https://github.com/minetest/minetest/commit/f48f6869303ab331600efaf9ec5796c238180106 07:16 RealBadAngel this has to be changed 07:16 RealBadAngel it breaks old clients 07:17 RealBadAngel adding 3rd optional drawtype for two already existing ones is just wrong 07:19 RealBadAngel setting for connected glass shall be read just in content_mapblock.cpp 07:19 RealBadAngel without adding new drawtype for that 07:22 RealBadAngel after case NDT_GLASSLIKE: check the setting and move to case NDT_GLASSLIKE_FRAMED: 07:25 RealBadAngel atm thx to adding that new drawtype minetest_game is not useable with stable clients, only current dev 11:08 jin_xi here is an idea: why not add a modifier to textures so dynamic ones can be deleted instead of leaked? like [dynamic: , so whenever a dynamic texture with a name is created dynamic ones with the same name would be deleted. name could be modname+position hash or so 11:21 Zeno` because it looks like a hack? 11:21 Zeno` A hack that doesn't address the real problem 11:24 Zeno` And I don't mean hack in the good sense of the word 11:24 Zeno` What makes a texture "dynamic"? 11:25 Zeno` And, why should a mod care what is dynamic and what is not (whatever dynamic means)? 11:25 jin_xi lol, do you know what the problem is? 11:26 Zeno` yeah... a leak 11:27 Zeno` Do you? 11:27 jin_xi ok, let me spell it out for you a bit: some mods need to change the appearance of a thing, by changing textures 11:27 jin_xi screens and such 11:28 jin_xi currently every frame of such screen is leaked 11:29 sfan5 no 11:29 sfan5 it's not leaked 11:29 sfan5 just saved for later usage 11:29 jin_xi it will probably not be used again and its a problem 11:30 jin_xi also it currently can not be deleted at all 11:31 Zeno` but adding a modifier? 11:31 Zeno` That's not a solution :/ 11:31 jin_xi would allow to delete textures if a mod wants to not keep them 11:32 Zeno` When does the lifetime of the texture expire? 11:32 jin_xi never 11:32 Zeno` never? So it will never be released? 11:32 Zeno` What's the point then? 11:33 jin_xi the point is to allow use of mods that use texture combining for dynamic content without the need for quitting your client so often 11:34 Zeno` How will minetest know when to release the memory for the texture? 11:34 sfan5 it won't 11:34 sfan5 minetest will cache every texture 11:35 Zeno` Yes, but how will adding a modifier clarify that situation? 11:35 jin_xi well, textures are created by name. by adding a name to a dynamic texture only one (current) could be kept 11:38 Zeno` so why the need for a modifier? 11:38 celeron55 the way to solve that is to add reference counting to every usage of textures in the client 11:39 celeron55 which is a rather big task but possibly can be done assuming there are no irrlicht interfaces that actively fight against doing that 11:41 jin_xi but has this to be done for all textures really? updating ones are a very specific thing. is the texture caching a problem somewhere else? 11:42 jin_xi because ideally i still think there should be a drawtype which pastes a texture over a node, to allow for sings and such without extra entities 11:43 gravgun in which way would it "pastes a texture over a node" ? 11:44 jin_xi idk ;) but it seems silly to need an extra entity per sign just for holding a textures. in some cases it might be needed for interaction 11:49 gravgun I already have thought about dynamic/custom(generated) textures, but what I can propose would be a big change in the Lua API and how we handle "textures" from there 11:49 gravgun (cf http://irc.minetest.ru/minetest-dev/2014-10-03#i_3961920 ) 11:49 gravgun BTW sfan5, still working on the VM thing 11:50 jin_xi with pasting over i mean something similar to what firelike does 11:54 Zeno` sfan5, why remove the bug fix tag for a fix that prevents uninitialised variables (objects) being used? 11:55 sfan5 did I remove that tag? 11:55 Zeno` looks like it 11:55 sfan5 oops 11:55 sfan5 I deleted that label because I didn't see it still has a bug 11:56 sfan5 we don't even need a "bug" label 11:56 sfan5 a "bug" label on a pull request automatically means it's a bug fix 11:58 Zeno` yeah true 11:59 Zeno` it's kind of a confusing tag if it's a PR :( 11:59 Zeno` wait, the fix it now a bug? 11:59 Zeno` this is confusing lol 12:00 Zeno` sfan5 removed the Bug fix label 18 hours ago; sfan5 added the bug label 4 minutes ago maybe I just need sleep 12:00 Zeno` o/ 12:49 Krock any comments to 1731 ? 12:50 Zeno` none further than I made last night 12:51 Krock yes, I'm waiting for dev comments and (maybe) a +1 12:51 Zeno` I am a dec 12:51 Zeno` dev 12:52 Krock I mean those with merge privileges 12:52 Zeno` pfft :p 13:54 OldCoder Any reason to use mapgen 7 vs 6 ? 13:55 sfan5 no 13:55 sfan5 there are some mapgens that use v7 13:55 sfan5 but those mods will enable it by themselves anyway 14:11 Dragonop OldCoder, Ethereal mod uses v7, and makes a lot of biomes 14:12 OldCoder shadowzone says buggy but Krock disagrees 14:12 Dragonop is't buggy 14:12 OldCoder kk 14:12 Dragonop *isn't 14:13 Dragonop It's a mapgen that can generate biomes 14:13 OldCoder kk 14:13 Dragonop but you have to mod it 14:13 OldCoder Go on 14:13 Dragonop Ethereal adds a lot of biomes 14:14 Dragonop try it 14:14 OldCoder Wait 14:14 OldCoder So if I don't have a special mod is it useful? 14:14 Dragonop You can use it to build stuff, but it's only stone 14:14 OldCoder Got it 14:14 OldCoder So if I use it 14:15 OldCoder without knowing what to do with special mods, bad 14:15 Dragonop yep 14:15 OldCoder kk 14:16 Dragonop Maybe for a creative server, but there are mods that creates a flat world 14:16 OldCoder kk 14:17 sfan5 OldCoder: v7 is not buggy but just not finished 14:17 OldCoder kkty 14:20 Zeno` It'd be hard for it to be buggy.... /me hides 14:20 OldCoder kk 14:27 Zeno` stuff like float height_mod = (float)(y + 17) / 2.5; is a bit of a worry though 14:29 Zeno` suggests that the programmer had NFI about implicit casts and promotion 14:31 celeron55 or maybe 2.5 was 2 previously and it was needed then 14:31 celeron55 there's nothing wrong explicitly casting stuff in places like that if you ask me 14:32 Zeno` there is 14:32 Zeno` it's as bad as doing a = (b * c) * d; 14:32 Zeno` it calls into doubt the rest of the code 14:33 Zeno` anyway 2.5 is not float 14:34 OldCoder 14:36 Dragonop 14:39 celeron55 also, yes, probably most of minetest's contributors do not have a fucking idea about implicit casts and promotion 14:40 celeron55 or at least some of them 14:40 OldCoder Well, the code will stabilize over time 14:40 OldCoder Simply curious about the current state for practical reasons 14:52 gravgun celeron55, wiki's "code style guidelines" says "Use STL", does that mean I should avoid implementing custom conainers? 14:53 celeron55 gravgun: unless actually needed, yes; it mostly refers to the fact that minetest originally used irrlicht's containers instead of STL 14:53 celeron55 (which could be mentioned in there...) 14:56 gravgun I ask cuz' I implemented a custom container that holds many structs of the same type, on top of C realloc'd arrays, because STL is sometimes darn slow and this is somehow performance-critical code (a VM aimed at generating vertices info for completely custom renderers) 15:45 celeron55 well sounds like std::vector would work just fine 15:46 celeron55 std::vector (and others) can be used without copying stuff by just using resize() and getting a reference to the new element, if that's what you were looking for 15:46 celeron55 (altough, maybe it doesn't get rid of all of it) 15:49 gravgun Yes, although std::vector is potentially slower at moving things given it's designed for C++ OOP, and the struct I use was made a POD struct specially to avoid that 17:42 paramat nore, please can my boats mod pull be merged? https://github.com/minetest/minetest_game/pull/328 tested and fine tuned for hours in my 'canyon' river mod 17:55 paramat nore, sfan5, MTgame's new grass (of dirt-with-grass) texture is really bad, dark and miserable, this one is much better as an improved-tiling version of what we had before https://github.com/minetest/minetest_game/pull/181 17:56 paramat the new dirt texture is bad too =) 18:05 Calinou >bad 18:05 Calinou the new dirt texture we have is quite good 18:05 Calinou the grass is not supposed to be super bright 18:05 Calinou this was the problem of Minetest 18:05 Calinou the new one has a good (not perfect) colour 18:06 Calinou it would need a recolour to fit anyway 18:06 Calinou grass is supposed to be quite smooth 18:20 paramat grass is the most visible texture of MTgame, and sets the mood, it used to be bright, fresh, with contrast and crisp detail, as if under a sunny sky. the new one is depressing, looks like grass under a dark sky 18:21 kilbith well, the new one is a good balance between bright /dark 18:22 paramat rich fertile dirt is dark brown, almost black, the new dirt is also depressingly lacking in contrast, and is a horrid light brown 18:22 paramat *rah!* 18:52 Calinou is the dirt of Minetest always wet though? 18:52 paramat well ... good point 18:53 VanessaE actually it is somewhat moist, otherwise the grass would be brown. 18:53 VanessaE in theory anyway 19:00 paramat there were actually complaints (quite rightly) about the new textures looking too dark and 'medieval' before they were merged. i wasn't too keen on the previous dirt (too light brown) but at least it had some contrast and detail 19:00 VanessaE I don't have too much opinion on the textures one way or another. it's too subjective 19:03 Calinou the new cobble is good actually 19:04 Calinou you can spam it without penalty ;) unlike all the other old ones 19:04 paramat yes i love the new cobble 19:05 paramat i'm an artist and designer and have strong reactions to visual stuff ... subjectivity goes out the window =) 19:08 Calinou everyone can callthemselves an artist, even programmers 19:08 Calinou same goes for designers 19:08 kilbith Calinou: the new cobble is too dark relative to the stone texture 19:09 kilbith you can't have darker broken stone if you break the 'raw' stone 19:09 kilbith (irl) 19:09 Calinou https://cdn.mediacru.sh/9pt71Ybwkp4o.png 19:09 Calinou looks fair to me 19:10 Calinou almost the same 19:10 kilbith you're not in exterior 19:10 kilbith compare them under the sunlight 20:52 Krock kilbith, I agree (2 hours later) 20:53 kilbith the current grass is perfect 20:53 kilbith same for the dirt 20:54 kilbith sorry, you talked about the cobbles 20:54 Krock Calinou, the dirt is dry by default (you can see that when using a hoe on dirt) 20:55 PenguinDad kilbith: the current cobble looks really nice imo http://a.pomf.se/fyelaa.png 20:56 Krock still, too dark. 20:56 kilbith yes, unappropriatly dark 20:57 Krock well, I prefer Cisoun's dirt and dirt_side textures 20:57 PenguinDad Where is this cobble too dark? o_O 20:57 kilbith I vomit when I see the Cisoun's dirt/grass 20:58 kilbith hm, wait 20:58 kilbith confused 20:58 kilbith Cisoun textures = current ones ? 20:58 Krock kilbith, you've got the wrong textures in your mind 20:58 kilbith ^ 20:58 Krock https://github.com/SmallJoker/base_game/blob/master/mods/default/textures/default_dirt.png 20:59 Krock https://github.com/SmallJoker/base_game/blob/master/mods/default/textures/default_grass_side.png 22:04 paramat 'hmmmm: going forward I'd like all new mapgens to use the biome infrastructure'. i assume you mean the biome API of mgv7 (which you want to change), okay that's a relief, less work choosing and creating biomes myself 22:18 paramat one thing lacking in register decoration is the way it fixes decoration density, i have found that forests only look good if noise is used to vary the density. this means my lua mapgens will always look just that little bit better than core mapgens =) 22:19 paramat though i understand the API is meant to be simple to use, and noise variable density is probably to complex for it