Minetest logo

IRC log for #minetest-dev, 2014-12-19

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

All times shown according to UTC.

Time Nick Message
00:01 Wayward_One joined #minetest-dev
00:03 diemartin joined #minetest-dev
00:07 gregorycu How does one view fps?
00:07 jin_xi F5
00:09 gregorycu I'm looking at ClientMap::render
00:09 gregorycu More than half of its time is spent in ShaderCallback::OnSetConstants
00:09 gregorycu aka. Half the time it's not even rendering
00:10 gregorycu Isn't this horrible?
00:14 RealBadAngel yes it is
00:15 RealBadAngel but with the current model of how the things goes its necesary
00:15 gregorycu hmmm...
00:15 RealBadAngel thats the main point of moving settings into a class and read its members instead of fiddlling with strings
00:16 luizrpgluiz1 joined #minetest-dev
00:16 gregorycu Is that a project which is happening?
00:16 gregorycu Is someone making that change?
00:16 RealBadAngel yes
00:16 RealBadAngel Zeno started to code that
00:17 luizrpgluiz1 hi devs and guys :)
00:17 gregorycu Can someone tell that dude to hurry the fuck up
00:17 gregorycu Where is he anyway?
00:17 RealBadAngel on the other side of the globe, playing (sleeping?) with kangaroos ;)
00:18 gregorycu I'm on the same side of the globe :P
00:18 gregorycu And I play with the same animals
00:18 RealBadAngel youre playing with kangaroos too? ;)
00:19 RealBadAngel greetings from poland anyway ;)
00:19 gregorycu Yeah, they are awesome
00:19 RealBadAngel but seriously, that change is rather big
00:20 RealBadAngel and definitely wont come in xmas release
00:20 gregorycu That's fine
00:20 gregorycu I'm looking at low hanging fruit to optimise, that is some of it
00:20 gregorycu What can also be done is a cache for bool values
00:20 gregorycu So the conversion happens on save
00:20 gregorycu As opposed to every read
00:21 RealBadAngel there are many spots to make things faster
00:21 RealBadAngel just pick your fav one
00:21 gregorycu Yeah, like the MeshUpdateThread
00:21 gregorycu Not sure what it does, but it does a lot
00:21 luizrpgluiz1 what improvements will put the new version of minetest?
00:21 RealBadAngel it is overloaded atm
00:22 RealBadAngel im working on the code that will require less mesh updates
00:22 VanessaE luizrpgluiz1: http://dev.minetest.net/Changelog#0.4.10_.E2.86.92_.3F
00:22 n4x joined #minetest-dev
00:23 RealBadAngel luizrpgluiz1, theres a thread in forums for that
00:23 RealBadAngel for example meshnodes
00:24 RealBadAngel https://forum.minetest.net/viewtopic.php?f=9&t=10701
00:24 VanessaE what, no link to my test mod? :P
00:25 VanessaE https://forum.minetest.net/viewtopic.php?f=9&t=10428
00:25 RealBadAngel https://forum.minetest.net/viewtopic.php?f=18&t=10621
00:25 VanessaE homedecor uses some mesh nodes also now (mostly starting from that primitives example)
00:25 RealBadAngel VanessaE, i was lookin for links :P just wanted xmas themed in the first place :)
00:26 RealBadAngel btw, can you turn that snowman into a mod?
00:29 VanessaE a snowman?  well, easily done of course..
00:29 VanessaE but not really interested in that honestly
00:29 VanessaE (plus such is not slated for inclusion in minetest)
00:30 nore joined #minetest-dev
00:35 RealBadAngel VanessaE, hmmmm was asking for a model
00:35 RealBadAngel i found one ccbysa 3.0
00:36 RealBadAngel http://www.blendswap.com/blends/view/71110
00:36 VanessaE yeah, I saw that before
00:37 VanessaE Dec 10 03:01:28 <hmmmm> er rather two nodes, a snowman head and a snowman body
00:37 VanessaE Dec 10 03:01:34 <hmmmm> not meshes, just nodes
00:37 RealBadAngel it could be propably split
00:38 RealBadAngel idk blender but it is supposed to do much neater tricks than that
00:38 VanessaE http://irc.minetest.ru/minetest-dev/2014-12-10#i_4056170
00:38 VanessaE seems like he wants a nodebox-based, one-node-high snowman
00:39 RealBadAngel https://www.youtube.com/watch?v=dzm8kTIj_0M
00:39 RealBadAngel theres no "nodebox" :)
00:39 VanessaE heh
00:40 jin_xi https://www.youtube.com/watch?v=Ib2Vl7JEjfc
00:41 CraigyDavi joined #minetest-dev
00:42 RealBadAngel jin_xi, lol
00:43 Selah joined #minetest-dev
00:48 gregorycu Hooray
00:48 gregorycu 5% speed improvement
00:51 luizrpgluiz1 celeron55: hi Celeron :D
01:03 paramat left #minetest-dev
01:05 luizrpgluiz1 left #minetest-dev
01:19 RealBadAngel gregorycu, at what?
01:19 gregorycu Overall, but specifically it's a 10% speed improvement to the MeshUpdateThread
01:20 RealBadAngel can i see the changes?
01:20 gregorycu It's a simple change
01:21 gregorycu I've replaced VoxelArea::getExtent() == v3s16(0,0,0)   with a new function VoxelArea::hasEmptyExtent()
01:22 gregorycu (I would have thought the compiler would have optimised the former to be equiv to the latter, but it did not)
01:22 RealBadAngel simple is a piece of wire, idea behind a piece of code is much more complicated
01:24 AnotherBrick joined #minetest-dev
01:27 gregorycu You are a wise sage
01:36 RealBadAngel gregorycu, make a PR with that code so we can check it
01:39 AnotherBrick joined #minetest-dev
01:43 gregorycu Will do
01:48 proller joined #minetest-dev
01:51 Zeno` joined #minetest-dev
01:51 nore joined #minetest-dev
01:56 gregorycu Zeno
02:02 Zeno` Can #1991 be merged please?
02:02 ShadowBot https://github.com/minetest/minetest/issues/1991 -- Fix visual_scale for plantlike nodes by VanessaE
02:03 Zeno` hi gregorycu
02:04 gregorycu I heard you're going to fix settings
02:04 Zeno` me?
02:04 gregorycu Yeah
02:04 gregorycu Singlehandedly
02:10 Zeno` #1991 is so obviously correct I'll merge it 30 minutes or so unless someone objects
02:10 ShadowBot https://github.com/minetest/minetest/issues/1991 -- Fix visual_scale for plantlike nodes by VanessaE
02:14 RealBadAngel Zeno`, main problem was the "if" that was applying the move of the roots only for scale <1.0
02:14 Zeno` RealBadAngel, correct
02:15 RealBadAngel but jungle grass is bigger for example
02:15 Zeno` Which is why my bad is obvious :)
02:15 Zeno` Also I did a second thing wrong. I was using HDX and for some reason it was working in that :/
02:15 RealBadAngel about the rest, thats my fault, i tried to fix the bug. but others just get used to faulty behaviour
02:16 jin_xi but why flip the signs? just get rid of the if
02:16 paramat joined #minetest-dev
02:17 paramat hmmmm nore https://github.com/minetest/minetest_game/pull/380
02:21 VanessaE for reference:  HDX worked for Zeno` only because his imagery was outdated (there was extra empty space/padding at the bottoms of some images to account for the old vertical offset problem that RBA originally fixed)
02:21 Zeno` it was a trick! Obviously cunningly designed to lure me into a trap
02:22 gregorycu Can you only do one pull request at a time?
02:22 VanessaE haha
02:22 VanessaE gregorycu: no
02:22 VanessaE as long as you use a separate branch for each feature you're coding, you can have more than one pull
02:22 gregorycu Oh, I don't have that
02:23 gregorycu So one PR per branch then
02:23 Zeno` merging
02:24 gregorycu My changes? Hooray!
02:25 paramat as far as i know if those temporary biomes get merged i will have to tell those using the biome api to use 'clear registered biomes' and 'register biome' within 'register on mapgen init'
02:26 kaeza gregorycu, even if you are working in a single feature, it is better to use a separate branch, and leave the "master" branch as a "clean" copy of Minetest
02:26 gregorycu Isn't the minetest/minetest branch a clean copy?
02:26 hmmmm this is disasterous
02:28 hmmmm i wonder how it would work out if I created the ServerEnvironment before scripts initialize, and then start processing environment steps only after they finish
02:33 paramat its a bit messy having to do everything in 'on mapgen init' so i really don't mind if thoe simple biomes don't go in, to give us more time to work on a solution
02:36 hmmmm you don't need to register biomes in on_mapgen_init though
02:36 hmmmm there's absolutely no harm in having biomes registered if the mapgen used is v6
02:36 hmmmm it would be preferable if you could register biomes outside of on_mapgen_init
02:37 gregorycu Are biomes a new feature?
02:39 paramat yes i agree, i did this because registered decos grief mgv6
02:39 paramat perhaps place decos should be removed from mgv6
02:39 Tesseract gregorycu: Yes, fairly new.  Not yet considered a public API.
02:40 Tesseract paramat: Nah, decos can be usefull in v6 too.  You can still register mapgen-specific decos too.
02:41 paramat yes, but it would solve one of these problems
02:41 paramat this indeed needs rethinking
02:45 paramat if place decos is in v6 we need to disable v5/v7 deco registration, to do that mapgen must be detected and this has to be done n 'on mapgen init'
02:45 paramat *in
02:47 paramat wotamess
02:48 paramat mapgen will need to be detected in 'on mapgen init' to enable/disable mgv6's on-gen function, to avoid that griefing mgv5/v7
02:53 paramat i'll withdraw my pull request then for now
02:53 hmmmm ?
02:54 hmmmm you don't need to register decorations if mapgen v6 is used
02:54 hmmmm just the biomes
02:54 gregorycu Who knows a lot of stuff about the MeshUpdateThread ?
02:59 paramat how do i register mapgen-specific decos? didn't know about this
03:00 hmmmm register the biomes in the main init, decos in on_mapgen_init
03:04 paramat yes
03:05 hmmmm i'm gonna try to see if i can move environment init before scripts start to run
03:05 hmmmm this has a lot of indirect consequences, though
03:05 paramat but then it seems players will have to clear registered decorations in 'on mapgen init'
03:06 hmmmm if i can remove on mapgen init completely, that would be excellent
03:07 paramat a while ago i tested 'clear biomes' and discovered it had to be done within 'on mapgen init' to have any effect
03:07 hmmmm not so
03:08 paramat .. if biomes were registered within 'on mapgen init'
03:08 hmmmm any biomes that were registered before it are cleared except for the internal builtin all stone biome
03:08 paramat i may have misremembered
03:08 paramat need to test again
03:09 hmmmm i need to ask celeron if there is any actual reason why scripts run before environment init
03:09 MinerDad joined #minetest-dev
03:10 hmmmm but this is stuff that needs to wait until after release
03:10 paramat yes, since we're so close to release perhaps we go with hmmmmm's default biome in builtin
03:10 hmmmm i don't *like* it, but it'll have to do
03:10 hmmmm so listen
03:11 hmmmm with clear biome and register biome you're able to add biome groups
03:11 hmmmm each mod can add its own set of biomes
03:12 hmmmm selecting a biome set makes the builtin functions do a raw biome clear and a raw biome add of all the biomes registered to that group
03:12 paramat okay
03:13 hmmmm the idea is to do as much of the work as possible in the lua
03:13 hmmmm features like these are higher level and certainly don't need to be part of the core
03:13 hmmmm the core is overbloated the way it is
03:14 paramat this could be solved short term by removing 'place decos' from mgv6?
03:15 hmmmm it's getting too close for comfort
03:15 hmmmm we should probably let it go until after release
03:16 paramat yes certainly, by short term i meant after release
03:16 paramat i need to test and play with all this more, and edit my biome branch
03:18 paramat sorry i should have discussed all this a while ago :) i'll close my MTgame pull
03:20 paramat i still have lots of testing to do on new noise stuff for release so will focus on that
03:33 paramat left #minetest-dev
03:44 harrison joined #minetest-dev
03:46 n4x joined #minetest-dev
04:28 hmmmm not sure why it took me so long to realize it but our noise frankly sucks
04:28 hmmmm it's low-quality
04:29 hmmmm it has sort of a roundish-cube look
04:29 hmmmm especially the 3d noise
04:30 Zeno` pics?
04:30 hmmmm see any mapgen output ever
04:30 hmmmm this is one of the reasons why I want to add simplex noise as an option
04:43 kahrl let's rename minetest to roundcube! oh wait... :P
04:44 kaeza kahrl, https://github.com/minetest/minetest/issues/1959#issuecomment-67597942
04:45 kahrl kaeza: hrm
04:46 kahrl since I don't have windows: any idea in what function it locks up?
04:46 gregorycu Thank you, I'll take a look right now
04:46 VanessaE kahrl: got a moment?
04:46 kahrl not really
04:46 VanessaE mmh
04:46 kahrl need to finish some stuff for uni
04:47 VanessaE well I just wanted to point out while you're here....   regarding https://github.com/minetest/minetest/commit/17dde5ddd0387c66aefc77e210c676e39c3b653c
04:47 kahrl glad it's the last day of uni for a while ;)
04:47 VanessaE there's an XCF for that, http://digitalaudioconcepts.com/vanessa/hobbies/minetest/images/infinite16.xcf
04:48 kahrl oh... does it produce different results?
04:48 VanessaE kilbith mentioned that the image you ended up with has bad contrast compared to my original file
04:48 VanessaE he seems to think so but I haven't had time to verify.
04:49 kahrl well of course it looks like that with the white github background
04:49 kahrl but it looks fine to me in the main menu
04:50 VanessaE oh, nope he's right
04:51 VanessaE hm, in practice it'll look the same though against a black background
04:51 VanessaE well almost the same, the difference is subble
04:51 VanessaE subtle*
04:52 VanessaE http://digitalaudioconcepts.com/vanessa/hobbies/minetest/screenshots/Screenshot%20-%2012182014%20-%2011%3a52%3a40%20PM.png
04:53 VanessaE (left is yours, both loaded from their original files into GIMP and then flattened against black)
04:53 gregorycu kaeza: I am still not able to repo
04:54 gregorycu I'm using 4.11 rc1
04:55 VanessaE kahrl: I'd much appreciate it if you'd use my original file ;)  also there's a 64px version for high-DPI screens if you're interested.
04:56 kahrl VanessaE: I'll change it this weekend
04:56 VanessaE ok.
04:56 kahrl keep the .xcf around in case we need it again at some point
04:56 VanessaE sure
04:56 sol_invictus joined #minetest-dev
05:00 exio4 joined #minetest-dev
05:04 MinerDad joined #minetest-dev
05:16 gregorycu I have this situation where terrain fades into view, the disappears, then fades into view again
05:16 gregorycu The same terrain, and I'm not moving
05:17 VanessaE view range auto tuner.
05:17 VanessaE yeah, does that to me too, sometimes
05:18 gregorycu This happens endlessly
05:18 gregorycu Why does it happen?
05:20 VanessaE no idea
05:20 pro__ joined #minetest-dev
05:26 kaeza okay... gregorycu, kahrl, the issue seems to be somewhere in async event handling
05:26 kaeza I have no idea how to debug this further
05:27 gregorycu When you freeze, what does the console window say as the last line?
05:29 kaeza currently, I modified builtin/mainmenu/common.lua, around line 183 (anon functions in handle_async call), and the second function is never called
05:29 kaeza (the function that should handle the result of the async call)
05:30 kaeza I also noticed that those async functions are being called twice in succession. not sure if that is intended
05:33 gregorycu hmm
05:35 gregorycu Well, I chucked some dodgy Lua in that file, and indeed, I crashed
05:35 kaeza from a quick look, looks like whoever coded that ignored the first rule in menu_lua_api.txt and is using globals from inside the functions
05:36 kaeza maybe kahr-l's patch just introduced some timing differences that makes it fail now
05:37 kaeza (I'm assuming some sort of race condition or something)
05:38 gregorycu In any case, Bad lua shouldn't cause a crash
05:39 gregorycu I don't know how to handle it gracefully however
05:40 kaeza I don't know how async calls work really, but in any case it should be some kind of hack as Lua is not really thread-safe
05:40 kaeza and too sleepy to check out
05:40 kaeza me, not Lua :P
05:41 n4x joined #minetest-dev
05:43 gregorycu Before you go
05:43 gregorycu You got no output in your console?
05:43 kaeza just the regular intro
05:44 gregorycu Interesting
05:44 gregorycu If there is a lua error, it is meant to print to the console
05:44 gregorycu Maybe there is a race condition where terminate is called before it gets the chance
05:47 gregorycu This is a Lua issue, and I don't know enough
06:09 neoascetic joined #minetest-dev
06:14 Megaf joined #minetest-dev
06:14 n4x joined #minetest-dev
06:15 exio4 joined #minetest-dev
06:29 paramat joined #minetest-dev
06:30 paramat hmmmm i have noticed the rounded-cube look of eased 3D noise, not unpleasant though, would tuning the easecurve help?
06:31 hmmmm i don't think it would have a significant impact
06:31 paramat non-eased 3D noise is still my favourite, looks amazing
06:31 paramat i'm now testing lacunarity and flags in a lua mapgen, good so far and interesting
06:32 hmmmm the rounded cube look i think is caused by having a low number of octaves
06:32 paramat absvalue is indeed different to abs(2d noise result)
06:32 Zeno` lunacy!
06:33 hmmmm combined with a low persistence value
06:34 paramat okay i'll test eased 3D noise next and try tuning it
06:34 paramat what i mean of course is i noticed the rounded-cube look in mgv5
06:35 hmmmm is the output identical to the old mapgen v5?
06:35 paramat perhaps slightly different
06:36 paramat not sure
06:38 paramat one thiing i did notice is new mgv5 seems to have more extreme values: higher mountains, more floatlands
06:38 MinerDad joined #minetest-dev
06:39 hmmmm that's not necessarily good
06:39 paramat uhuh, epic though
06:40 Hunterz joined #minetest-dev
06:40 paramat highest point found in new mgv5 so far is y = 175
06:42 paramat https://cdn.mediacru.sh/R/R-Kj-hNdFDra.png
06:45 gregorycu Is that good or bad?
06:47 paramat good unless new mgv5 is significantly different from old mgv5
06:48 paramat i might re-compile c55's branch to compare again since i'm so unsure about this
06:55 paramat mgv5 has terrain persistence 0.55, i'm tryng 0.63 in this lua mapgen and it looks awesome
06:59 hmmmm aghh
06:59 hmmmm i have to fix the irrlicht u64 problem every time i skip to a different old commit
07:00 paramat the rounded-cube look is still subtly there but perhaps it's inherent with noise, not a problem to my eyes
07:00 hmmmm it's inherent with that type of noise
07:00 hmmmm simplex noise does not have this problem
07:02 paramat interesting i'll research simplex
07:03 paramat lacunarity 1.1 creates very high quality smooth rolling hills
07:04 hmmmm with 3d noise?
07:06 paramat 2d
07:06 hmmmm i tried modifying lacunarity a couple days ago, nothing below 2 though
07:07 hmmmm with 3 it tends to make many small bumps in the terrain
07:07 hmmmm so yeah, i guess it's the logical opposite of 1.1
07:08 hmmmm i can't help but wonder if the rounded cube look would be reduced if we used cubic interpolation rather than linear
07:09 paramat lacunarity 1.1 with high octaves and persistence makes high quality 3d noise, very un-cubic
07:09 paramat but heavy processing
07:13 paramat ahhh its so good being back in lua mapgen, core is stressful
07:14 paramat 3d noise eased, absvalue is bubbly like merging bubbles with inverted ridges between
07:15 paramat all checked and working then, thanks for these new possibilities
07:31 Megaf_ joined #minetest-dev
07:33 gregorycu Ahh damnit
07:33 gregorycu Minetest searches for mods in ..\mods UNLESS it's under your home directory...
07:36 gregorycu No, that's not accurate
07:36 paramat okay simplex noise sounds excellent from research, really looking forward to that
07:40 MinerDad joined #minetest-dev
07:42 hmmmm i can really understand when people say they miss the way 0.3.x was
07:42 hmmmm there's a beautiful simplicity in it
07:42 hmmmm super low tech but it works well
07:44 gregorycu The Lua C API doesn't appear to make much sense
07:47 hmmmm ...?
07:54 gregorycu For the record, minetest searches for mods relative to the binary if RUN_IN_PLACE is set, otherwise in user directory
07:55 hmmmm that's the point of RUN_IN_PLACE
07:55 gregorycu Yes... it would appear to be
07:55 hmmmm paramat:  http://i.imgur.com/ScVcjFm.jpg   http://i.imgur.com/ScVcjFm.jpg
07:55 hmmmm I'd say v5 is pretty different
07:56 hmmmm er http://i.imgur.com/AZDj7w7.jpg
07:56 gregorycu wow
07:56 gregorycu What's your FoV?
07:56 hmmmm what?
07:56 gregorycu Field of view
07:57 hmmmm paramat:  most notably, 3d noise isn't eased in 0.3.3
07:57 hmmmm I don't know..
07:57 gregorycu It's large
08:24 kilbith joined #minetest-dev
08:24 * paramat looks
08:25 paramat but eased in later 0.3.x?
08:26 hmmmm this is the latest 0.3.x
08:26 hmmmm btw here's the same thing with not-eased noise:  http://i.imgur.com/rPb2QsH.jpg
08:26 paramat so mgv5 only became eased in c55's branch...?
08:26 hmmmm you're doing something waay different
08:27 hmmmm mgv5 as in the mgv5 in 0.3.1, 0.3.2, and 0.3.3
08:27 hmmmm did he base his v5 revival off an earlier version?
08:28 paramat don't know
08:29 paramat his input on this will be interesting =)
08:30 paramat all i know is right from the beginning of reviving v5 it was chosen to have eased noise
08:30 paramat we could make it an option
08:31 paramat when i compiled his branch it was almost certainly eased (from memory)
08:31 paramat fairly sure
08:32 paramat i would heve noticed the spikiness
08:36 paramat non eased v5 looks good to me, should be an option, especially since it's true to some/all versions of 0.3.x
08:37 hmmmm guess what
08:37 hmmmm it's already an option
08:37 hmmmm you can set mapgen noise parameters in the config :)
08:38 paramat oh yeah
08:38 paramat kewl
08:39 hmmmm heh
08:39 hmmmm looking at the commit where genuine perlin noise was dumped in favor of the simplified version
08:40 hmmmm "// This is too slow"
08:40 hmmmm i'll have to give it some love and bring that back too, since the noise it creates is typically of higher quality
08:41 hmmmm humm
08:41 hmmmm sorry to break it to ya but historically minetest never had eased 3d noise
08:41 hmmmm https://github.com/minetest/minetest/commit/48a772b3c6388cbe7f9003557f27cc2cf51b8fe8  this is where 3d noise gets added
08:42 MinerDad joined #minetest-dev
08:42 paramat in early 0.4.x
08:44 hmmmm noise.cpp doesn't seem to be touched in 0.4.x
08:46 Zeno` touch noise.cpp
08:46 Zeno` problem fixed
08:46 hmmmm https://github.com/minetest/minetest/commits/0.4.4/src/noise.cpp
08:46 hmmmm so yeah
08:46 paramat 26 feb 2011 seems to be v0.2.x looking at releases
08:47 Zeno` wow, I didn't realise it would have been that long
08:47 hmmmm after adding the mindless "flip xz" and "constant 1" noise options, celeron didn't actually touch that file
08:48 paramat O/
08:48 paramat left #minetest-dev
08:48 n4x joined #minetest-dev
08:52 GrimKriegor joined #minetest-dev
09:04 neoascetic joined #minetest-dev
09:17 FR^2 joined #minetest-dev
09:27 OldCoder joined #minetest-dev
09:37 OldCoder joined #minetest-dev
09:37 OldCoder joined #minetest-dev
09:44 ImQ009 joined #minetest-dev
10:31 MinerDad joined #minetest-dev
10:33 T4im joined #minetest-dev
10:41 exio4 joined #minetest-dev
10:51 AnotherBrick joined #minetest-dev
10:56 SudoAptGetPlay joined #minetest-dev
10:57 AnotherBrick joined #minetest-dev
11:06 jin_xi joined #minetest-dev
11:16 exio4_ joined #minetest-dev
11:54 pro__ joined #minetest-dev
12:14 ImQ009 joined #minetest-dev
12:20 MinerDad joined #minetest-dev
12:27 exio4_ joined #minetest-dev
12:36 exio4 joined #minetest-dev
12:42 nore joined #minetest-dev
12:46 ImQ009 joined #minetest-dev
12:52 kilbith does someone could update http://dev.minetest.net/Changelog  please ?
12:53 Zeno` 0.4.11 is not released yet though
12:53 kilbith it can me made since the freezing appliance
12:54 Zeno` hmm, probably yeah
12:55 selat joined #minetest-dev
12:58 gravgun joined #minetest-dev
13:23 sol_invictus beeheizeitiesaijahbu
13:24 gregorycu_ joined #minetest-dev
13:31 roniz joined #minetest-dev
13:34 ImQ009 joined #minetest-dev
13:46 Ritchie joined #minetest-dev
13:57 n4x joined #minetest-dev
14:03 MinetestForFun joined #minetest-dev
14:10 shadowzone joined #minetest-dev
14:17 luizrpgluiz joined #minetest-dev
14:19 luizrpgluiz hi
14:25 Amaz joined #minetest-dev
14:39 shadowzone joined #minetest-dev
14:42 roniz http://a.pomf.se/vrnjvs.png what is wrong with the glass?
14:44 sfan5 roniz: 1) you should ask this in #minetest 2) the drawtype of glass should be set to glasslike
14:45 roniz okay, can you help in #minetest sfan5
14:45 sfan5 roniz: see 2)
14:46 exio4 joined #minetest-dev
14:46 n4x joined #minetest-dev
14:48 MinetestForFun joined #minetest-dev
14:53 Hunterz joined #minetest-dev
14:55 MinerDad joined #minetest-dev
14:56 sfan5 celeron55: has there been a decision about including subgames into 0.4.11?
14:59 celeron55 none of the menu stuff that postponed it from 0.4.10 was done, so the situation is quite the same
14:59 sfan5 so.. no subgames?
14:59 celeron55 i.e. there still are no game descriptions in the menu
15:01 celeron55 what's the planned release day?
15:01 sfan5 I think hmmmm said something about christmas
15:01 sfan5 I wrote around 24th in the RC post on the forums
15:02 sfan5 s/around 24th/"around the 24th"/
15:07 nore joined #minetest-dev
15:08 proller joined #minetest-dev
15:14 celeron55 there's a serious polishedness issue in skytest for example
15:14 celeron55 because the helicopter lags like literal hell due to the very heavyweight map generator
15:15 sfan5 nore: https://github.com/minetest/minetest_game/pull/379 ok? [yay/nay]
15:15 celeron55 it's just... not worth releasing to the wide public, practically everyone would hate it
15:16 kilbith Carbone is mature for it, though
15:18 celeron55 guys, it would be kind of a good idea to focus development on these things that that subgames are trying to do but fail at due to engine issues
15:18 Zeno` any comments on #1967?
15:18 ShadowBot https://github.com/minetest/minetest/issues/1967 -- Add display_gamma option for client by Zeno-
15:18 celeron55 instead of anything else
15:18 celeron55 nothing else is worth it
15:18 celeron55 just so that you know what to do for 0.4.12
15:20 pirilon joined #minetest-dev
15:20 pirilon left #minetest-dev
15:24 pirilon joined #minetest-dev
15:27 Zeno` celeron55, what kind of things?
15:27 celeron55 this is still a very tough question
15:27 celeron55 Zeno`: that isn't a tough question
15:27 celeron55 Zeno`: the things that for example skytest and tutorial need are quite obvious
15:28 celeron55 skytest needs a way to asynchronously generate the custom world so that the helicopter does not suck, and tutorial needs support for pre-made worlds which can be instantiated from the menu
15:28 celeron55 and both need a description in the menu
15:28 Zeno` I thought we were not allowed to add extra Lua functions
15:29 celeron55 anything that is required for a clean implementation for each of those features is allowed, because how else they could ever be made?
15:29 celeron55 but! the tough questions is subgames for this release
15:30 hmmmm joined #minetest-dev
15:30 celeron55 question*
15:31 shadowzone good thing my subgame is simple and doesn't have all these extra features
15:31 hmmmm oh hey celeron
15:31 hmmmm why is the ServerMap created after script init
15:32 celeron55 shadowzone: that means it's also quite identical to minetest_game and thus unneeded
15:32 celeron55 hmmmm: i don't know
15:32 hmmmm good
15:32 hmmmm i'm going to flip that around and remove on_mapgen_init
15:33 celeron55 maybe you will find out while doing it
15:34 hmmmm oh another thing
15:34 shadowzone celeron55: I am trying to make different from minetest_game and I am young and still learning so I could use some encouraging.
15:35 hmmmm what version was your mapgen v5 revival branch based on
15:35 celeron55 shadowzone: that means you will need these extra features 8)
15:35 shadowzone Which I hope to end up with
15:35 shadowzone I am working on a HUGE mod for it
15:35 hmmmm it certainly doesn't match the output of 0.3.3 (which has incredible terrain, by the way) and minetest never had eased 3d noise
15:37 celeron55 i am guessing 0.3.2 because i have that code around
15:37 hmmmm the difference between 0.3.2 and 0.3.3 is two commits and neither of them are related to map generation
15:37 celeron55 i would guess the noise parameters have somehow got modified in the process of porting it
15:37 celeron55 either their interpretation or the actual parameters itself
15:38 zat joined #minetest-dev
15:40 celeron55 or that they just haven't changed; obviously all screenshot of v5 over the years have been from the absolute best stuff it has ended up creating
15:42 hmmmm i did a comparison last night with the same seed at the same position
15:46 Calinou joined #minetest-dev
15:47 celeron55 flying around in a current mgv5 world, i would say yes, this is the exact same stuff we had back then
15:48 celeron55 dunno about exact seeds, doesn't seem like a big issue
15:48 celeron55 various things have changed since then anyway
15:48 hmmmm http://i.imgur.com/ScVcjFm.jpg    http://i.imgur.com/AZDj7w7.jpg
15:50 celeron55 i will say though that current texture packs are boring as hell compared to the bright colors of 0.3
15:53 Zeno` did 0.3 have stuff (land) floating in the air?
15:54 exio4 joined #minetest-dev
15:55 Megaf_ Zeno`: Im getting a billion of this
15:55 Megaf_ 10:53:57: ERROR[ServerThread]: ServerEnv: Trying to store id=2824 statically but block (-6,2,-13) already contains 128 objects. Forcing delete.
15:55 Megaf_ like, 200 of them in one second
15:56 Megaf_ I increased the number to 4000
15:57 Zeno` that has nothing to do with anything I've done AFAIK
15:57 VanessaE I've gotten the same thing on my worlds before, and I've been unable to explain it
15:58 Zeno` I get it on MTZ quite commonly :(
15:58 VanessaE go to the mapblock in question and there are NONE in the block (not even the 128 allegedly already there)
16:03 Megaf_ yep, there's nothing there
16:03 sol_invictus reported block position is false
16:03 VanessaE make sure you're in the right block (multiply the coords by 16 and teleport to there)
16:04 VanessaE sol_invictus: oh?
16:04 sol_invictus VanessaE: ah, right
16:05 sol_invictus well, I remember I added logging to spawn entity function and the reported position of spawn entity and the error message were different
16:06 sol_invictus anyway, why spawn entity is not logged?
16:06 sol_invictus I'd help tracking down problems like this
16:07 sol_invictus s/I'd/it'd
16:21 pro__ joined #minetest-dev
16:23 MinerDad joined #minetest-dev
16:24 Calinou can be added in builtin code
16:25 hmmmm so it's friday, supposed to be the day of the 0.4.11 release
16:25 hmmmm what's the status on the blockers?
16:29 Calinou wasn't it December 24?
16:29 Zeno` 2 blockers fixed, 3 to go
16:30 hmmmm no, there's a week of feature freezes
16:30 kilbith hmmm: the week-end is the period when the devs are the most available for reviewing the PRs, better releasing after that
16:31 hmmmm so
16:31 hmmmm #1896 has a patch already - nobody committed it?
16:31 ShadowBot https://github.com/minetest/minetest/issues/1896 -- font_size setting broken
16:32 pro joined #minetest-dev
16:32 hmmmm #1860, RBA said he's working on it yesterday
16:32 ShadowBot https://github.com/minetest/minetest/issues/1860 -- Wield item is always fully bright
16:32 hmmmm the root cause for #1959 was found i think, kaeza is the one working on that i think?
16:32 ShadowBot https://github.com/minetest/minetest/issues/1959 -- Checking public serverlist checkbox causes minetest to freeze.
16:33 hmmmm we could theoretically get this out today
16:33 Zeno` #1896 has a workaround committed that you and I discussed, hmmmm
16:33 ShadowBot https://github.com/minetest/minetest/issues/1896 -- font_size setting broken
16:33 hmmmm then perhaps the blocker label should be removed
16:33 Zeno` well, I dunno
16:35 Zeno` 5c55738 is related
16:35 Zeno` https://github.com/minetest/minetest/commit/f7f7fec497f83da0d9f6ea2e83796720484033ff is the one that bypasses the issue
16:37 Zeno` Since sapier seems to be away maybe the blocker should be removed
16:37 hmmmm yeah it was
16:38 hmmmm as far as i understand, the other two are in progress
16:41 Zeno` Since the issue no longer manifests because the offending code is commented maybe the blocker should be removed
16:42 Zeno` it still needs to be fixed but it doesn't look like anyone is working on it and *shrug* the current version works
16:43 VanessaE hmmmm: the release date IS december 24.
16:43 VanessaE that was the target date.
16:44 hmmmm who came up with friday as the feature freeze then
16:44 VanessaE beast me
16:44 VanessaE beats*
16:44 hmmmm a one week feature freeze started last friday
16:44 hmmmm so this friday at whatever time, there should be a release
16:44 VanessaE er, today is friday you know...
16:45 Zeno` not here! It's saturday here
16:45 VanessaE sorry, your choice of words just seems odd
16:45 hmmmm but today is not the 24th
16:45 VanessaE anyway, the 24th is the target date.
16:45 hmmmm aha
16:45 VanessaE it was supposed to be out by then as a "christmas release"
16:45 hmmmm i don't know who came up with that
16:45 hmmmm christmas release doesn't mean it needs to be released on or right before christmas
16:46 VanessaE um, yeah it does
16:46 VanessaE that's basic marketing
16:46 hmmmm ..
16:46 hmmmm well look
16:46 hmmmm as soon as these two blockers get resolved, there's going to be a release
16:46 Zeno` the release date as far as I know is the 24th
16:46 hmmmm point me to where that was said
16:46 Zeno` no, leave it until the 24th hmmmm
16:46 hmmmm could you explain why
16:47 VanessaE Dec 19 10:01:55 <sfan5> I think hmmmm said something about christmas
16:47 VanessaE Dec 19 10:02:06 <sfan5> I wrote around 24th in the RC post on the forums
16:47 Zeno` because that is the date that was published and that's the purpose of a feature freeze even if no new issue crop up
16:47 VanessaE but that's far too recent, lemme search back
16:47 hmmmm I said we should add a snowman block that gets added to maps around christmas time
16:48 hmmmm and stops getting added after christmas
16:48 VanessaE hmmmm: that ^^^ was in response to c55 asking what the release date was.
16:48 Zeno` you don't "feature freeze" until x/x/x and then decide before that date there is nothing left to resolve and release anyway
16:48 VanessaE I'm checking logs
16:48 hmmmm Zeno`, the feature freeze was defined to be a week lon
16:48 hmmmm long
16:48 hmmmm that's the problem i'm having
16:49 hmmmm so if the release date were actually going to be the 24th, why wasn't the feature freeze on the 17th?
16:49 Zeno` I'm certain the 24th was specified. If not then *shrug* (but a week is not long enough; that's a different issue)
16:49 VanessaE hmmmm: why does anything around here lag behind?
16:50 Zeno` it was/is a 2 week feature freeze IIRC
16:50 hmmmm because people like yourself and zeno think that we should wait longer
16:50 hmmmm that's why
16:50 Zeno` me?
16:50 hmmmm you're clearly a proponent of a 24th release
16:51 Zeno` Yes, I do think 2 weeks is more appropriate but I was not involved in the discussion at all
16:51 Zeno` So don't look at me :P
16:51 hmmmm the development process as a whole is fubar
16:52 hmmmm waiting some arbitrary amount of time doesn't make the software better
16:52 VanessaE um...
16:52 hmmmm this is like Debian
16:52 Zeno` it does. It gives people time to TEST
16:52 hmmmm "if we have REALLY old packages, our OS will be more stable"
16:52 Zeno` you really think 2 weeks is a long time?
16:53 VanessaE we should "wait longer" because we have a BAD habit of making a pending-release announcement when there are nasty bugs left on the table, and not focusing on fixing those nasty bugs before people start talking about releasing...
16:53 hmmmm 0.4.11 should have its own branch already
16:53 hmmmm any bugfixes should be retroactively applied to that
16:53 hmmmm development should never stop though
16:53 VanessaE I don't care if it's released at christmas, channukah (8th night, since the 1st night is long past), new years, easter, or any other major holiday
16:54 hmmmm this has got to be the only project in existence that doesn't have a distinction between -RELEASE and -CURRENT
16:54 VanessaE I care that a release is planned when there are few big, user-facing bugs on the table, such that devs who are capable of fixing those bugs can find the time and can be bothered to fix them
16:54 Zeno` it's probably one of about 2 projects that also has functions that are 500 lines ore more (commonly)
16:55 Zeno` seriously, what is wrong with a feature freeze?
16:56 hmmmm nothing, but you need to set a boundary somewhere
16:56 Zeno` yes, the boundary is the 24th
16:56 hmmmm 1 week was the agreed upon boundary
16:56 hmmmm this better be the least buggy release ever done
16:56 jin_xi joined #minetest-dev
16:56 Zeno` 1 week to check for bugs and properly test?
16:57 hmmmm we shouldn't have to do so many god damn bugs
16:57 Zeno` even 2 weeks is not long enough
16:57 hmmmm from now on if someone adds a new feature, they need to add unit tests for it
16:57 Zeno` you cannot unit test many of the things in minetest
16:57 VanessaE hmmmm: oh, also, the start-of-freeze was delayed because kahrl was working on adding some icons
16:57 VanessaE Dec 12 16:16:46 <Tesseract>     I'm just waiting on kahrl for icons before start
16:57 VanessaE ing the freeze.
16:57 VanessaE Dec 12 16:17:15 <Tesseract>     hmmmm: Really?  Is anyone going to make substant
16:57 hmmmm then perhaps your software is poorly designed
16:58 Zeno` unit tests for anything but the lowest level of functions are a waste of time
16:58 VanessaE (oops, ignore the second line)
16:58 Zeno` hmmmm, my software?
16:58 VanessaE and before that, Dec 12 13:14:40 <PilzAdam>      ok, so we freeze today?
16:58 Zeno` no, I don't think so :P
16:58 hmmmm minetest as a whole
16:58 VanessaE Dec 12 13:15:22 <Tesseract>     PilzAdam: +1
16:59 VanessaE so yeah, freeze was announced and began on the 12th
16:59 hmmmm yeah
16:59 hmmmm you and zeno both indicated you want it to be longer than a week
16:59 hmmmm so the date which the feature freeze went into effect is irrelevant
17:00 Zeno` hmmmm, a week is not long enough. But I based all of my comments on the fact that I *thought* 24th was the planned released dat
17:00 Zeno` date*
17:00 hmmmm zeno-, have you ever heard of test-driven design
17:00 Zeno` those remaining two blockers will not be fixed today anyway
17:00 Zeno` hmmmm, I have
17:00 hmmmm well then
17:01 Zeno` hmmmm, you can not convert the minetest codebase to a test-driven design because of the crap load of functions that do 25428592 different things
17:01 Zeno` instead of ONE
17:02 n4x joined #minetest-dev
17:02 Zeno` well, you can. But not this month
17:02 hmmmm and that's fine
17:02 hmmmm good things take time
17:02 hmmmm a lot of end users are starting to complain about overall quality being more important than having new features
17:02 Zeno` there are good arguments against test-driven design as well
17:02 VanessaE AH
17:02 VanessaE found the reference
17:02 VanessaE Dec 12 10:08:23 <sfan5> there are about 12 days left until we plan to release, w
17:02 VanessaE ould not be a good time to release an RC build for people to test?
17:02 hmmmm the reason why we focus on new features is because that's what everybody wanted initially
17:03 VanessaE that equates to dec. 24
17:03 sfan5 VanessaE: that's a guess
17:03 sfan5 s/that's/that was/
17:03 sfan5 ugh
17:03 sfan5 s/.*/I was not sure when we planned to release./
17:04 sfan5 VanessaE: try searching before that date
17:04 celeron55 i'm testing some subgames to be up to date with them
17:04 VanessaE and besides, according to my logs, *I* was the one who suggested start-of-freeze on Dec 10 and release day on Dec 24
17:04 VanessaE sfan5:  ^^
17:04 celeron55 voxelgarden is a very solid start at something slightly different than mt_game, but it's lacking content; i'm guessing it will be good once the author finds a way to add mobs in a solid way
17:05 VanessaE well not suggested but rather put 2-and-2 together
17:05 celeron55 everything it contains is of high quality
17:05 hmmmm speaking of mobs
17:05 hmmmm mobv2 is broken in 0.3.3
17:05 hmmmm thanks to an irrlicht interface change
17:05 celeron55 carbone is a bit like a lower quality less different version of voxelgarden with more content
17:06 Zeno` One of the best arguments against test-driven design is that all the things it's suppose to do can be in the actual code
17:06 Zeno` except assert() is redefined
17:06 Zeno` in mt
17:06 celeron55 skytest is awesome but the helicopter sucks due to the engine
17:06 celeron55 tutorial needs management functionality in menu
17:06 hmmmm assert() was made to be redefined
17:07 celeron55 none of this is very good
17:07 Zeno` hmmmm, it is not, however, meant to persist in release builds
17:07 hmmmm it doesn't
17:07 Zeno` it does in minetest
17:07 hmmmm if that's true, then it's a problem
17:08 celeron55 is everyone going to hate me if i will still say that subgames need to wait for the engine to catch up with them
17:09 hmmmm no, it's true
17:09 hmmmm but damn
17:09 hmmmm people want new features
17:09 hmmmm people want a good interface for their subgames
17:09 celeron55 we need to add features to the engine based on what subgames want to do
17:09 hmmmm people want minetest to be stable and backwards compatible
17:09 hmmmm people want better quality control
17:09 Zeno` hmmmm, https://github.com/minetest/minetest/blob/master/src/debug.h#L80
17:10 Zeno` read and weep
17:10 kaeza joined #minetest-dev
17:10 hmmmm for the people to get what they want, minetest would need to have full time developers
17:10 VanessaE http://irc.minetest.ru/minetest-dev/2014-11-30#i_4043879
17:10 VanessaE that's enough.
17:11 celeron55 hmmmm: what i am saying is that everything else can have lower priority
17:11 Zeno` assert() is used incorrectly all throughout minetest, so there is this horrible hack there
17:11 hmmmm okay then
17:12 celeron55 i'm okay with changing debug.h's assert to not override the standard assert
17:12 hmmmm priority is a murky concept when you're dealing with people not on the hook to get this piece software working
17:12 hmmmm I'll certainly respect priorities though
17:12 celeron55 (and leave it there for where it's explicitly wanted)
17:12 hmmmm is there a list of subgame requests?
17:13 VanessaE https://forum.minetest.net/viewtopic.php?f=15&amp;t=9103
17:14 celeron55 umm
17:14 Krock joined #minetest-dev
17:14 celeron55 there is a thread on the forum but i cannot find itt
17:14 celeron55 it*
17:14 celeron55 about modder's requests directed to engine development
17:14 celeron55 where the hell is it?
17:15 Krock Already checked the forum "Inofficial Developement"?
17:16 celeron55 here i think; it has devolved into something not very useful though https://forum.minetest.net/viewtopic.php?f=3&amp;t=9179
17:16 celeron55 i'll consider starting to collect such a list myself
17:18 VanessaE this?  https://forum.minetest.net/viewtopic.php?f=5&amp;t=9709
17:22 VanessaE eh, beat me to it.
17:24 MinerDad joined #minetest-dev
17:28 hmmmm different topic... what do you guys think about running abms in a coroutine so that we can lua_yield() after a certain timeout
17:28 hmmmm this way Server consumers don't get starved from a mod that's taking too long
17:29 VanessaE how much effort would it take?
17:29 Krock hmmmm, how long would be that timeout?
17:31 Krock #1987 any thoughts? I tested it and haven't got any problems with it
17:31 ShadowBot https://github.com/minetest/minetest/issues/1987 -- Change TileSpec::frames to be std::vector not std::map by gregorycu
17:31 Krock ..yet
17:32 Krock Yes, it could be labeled as feature..
17:32 Krock but it would increase the speed (looking on his pictures)
17:33 Zeno` There is no problem with it as far as I can see and, perhaps, std::vector is the better type but profiling on Linux yields no difference
17:33 Zeno` That said, I would agree to it because std::vector seems the more appropriate data type
17:35 MinerDad joined #minetest-dev
17:35 Zeno` and maybe it does help with MSVC build... I can't profile that
17:35 Krock hmmmm, If I may ask - What would you say about the PR above?
17:36 hmmmm it's fine but it can't get committed until after the release
17:36 Krock too bad :(
17:37 Zeno` why?
17:37 Zeno` it probably makes no difference with an optimised build
17:38 Zeno` it's better but not something that is "too bad" :(
17:39 Du_Draig joined #minetest-dev
17:41 iqualfragile joined #minetest-dev
17:45 Krock Zeno`, I'm not sure if it's the commit but I have about 10% less CPU usage now
17:45 VanessaE celeron55: here's an important one:
17:45 VanessaE https://forum.minetest.net/viewtopic.php?f=5&amp;t=9709#p147812
17:46 VanessaE pretty much every complex mod out there aside from his list suffers this problem.
17:46 kaeza related to that ^, I've been thinking of a way to allow mods to override the description for a particular ItemStack
17:47 VanessaE yep, that too
17:48 kaeza I have a hacky proposal: if the ItemStack metadata is valid JSON, use some fields from it for special things
17:49 kaeza like "image", "description", etc, fields
17:49 kaeza if it can't be parsed as JSON, just do nothing
17:49 kaeza but... not sure if metadata is sent to client in the first place anyway :P
17:49 hmmmm okay in order to do that you need to read map metadata
17:50 hmmmm for every single node written to a mesh
17:50 hmmmm this is sllllow
17:50 kaeza "node"?
17:50 hmmmm "node"
17:50 ImQ009_ joined #minetest-dev
17:50 hmmmm ahh itemstack
17:51 kaeza yeah
17:51 VanessaE well there is still that old meta-set-nodedef branch...... ;)
17:51 hmmmm ha ha ...
17:51 VanessaE (it could be used to solve this issue)
17:51 hmmmm nope
17:51 VanessaE I never said it was efficient ;)
17:51 ImQ009 joined #minetest-dev
17:53 rubenwardy joined #minetest-dev
17:55 kaeza how to translate from node meta to stack meta is left up to mods, and the JSON parsing and interpretation of fields is done fully client-side
17:56 kaeza and since stack meta is just a regular string it should be backwards compatible
17:58 Anchakor_ joined #minetest-dev
18:00 VanessaE and here's another one:  https://forum.minetest.net/viewtopic.php?f=5&amp;t=9709&amp;start=50#p150944
18:00 VanessaE (summary:  liquids should just stop flowing at all when they hit CONTENT_IGNORE, rather than trying to spread)
18:04 pro joined #minetest-dev
18:05 n4x joined #minetest-dev
18:06 gravgun joined #minetest-dev
18:08 VanessaE https://forum.minetest.net/viewtopic.php?f=5&amp;t=9709&amp;start=75#p155705
18:09 VanessaE I've noticed that middle-click-to-craft-10 slowness as well, even in singleplayer
18:10 VanessaE celeron55: you wanted to make a list, this thread (same one I pointed you at) does have some good stuff.
18:24 pro joined #minetest-dev
18:25 Sokomine joined #minetest-dev
18:26 VanessaE bbl
18:40 kaeza the middle-click-to-craft-10 issue is because the client or server makes 10 "pick one" requests instead of a single "pick ten" request
18:42 jin_xi srs
18:43 kaeza or at least, the log outputs 10 times "player foo crafted bar:baz 1", instead of a single "player foo crafted bar:baz 10"
18:44 kaeza each "request" possibly has to make a slow look up with get_craft_recipe or whatever
18:51 MinerDad joined #minetest-dev
19:07 monte joined #minetest-dev
19:13 VargaD joined #minetest-dev
19:31 hmmmm hey celeron
19:31 hmmmm if you're around I was just wondering what the rationale was behind the restrictions on PseudoRandom()
19:31 hmmmm why does min and max have to be within a certain range?
19:32 celeron55 it happens to have a rather limited implementation
19:32 hmmmm but what was the whole purpose
19:32 hmmmm did you just decide to add some arbitrary restrictions to it
19:32 celeron55 what restriction exactly?
19:32 hmmmm if(max-min > 32768/10)
19:33 celeron55 if you have eg. max-min=25000, you get much higher chance of the result being <7768 or something than any other value
19:33 shadowzone joined #minetest-dev
19:33 celeron55 because it's just % basically
19:35 hmmmm huh
19:35 celeron55 it's embarrassingly bad really but this is again one of those things that has worked well enough that nobody noticed it wasn't really up to scratch
19:35 hmmmm it's really bad actually
19:35 celeron55 and now people rely on it giving the same values
19:35 celeron55 (well, some people; not that many i would guess)
19:35 hmmmm for example blockseeds are always even or odd depending on the least significant bit of the map seed
19:36 hmmmm so anything that does range(0,1) == 0 or whatever would always fail for a specific map
19:36 hmmmm or always be true for a different map
19:36 hmmmm a better random is definitely needed, but i don't completely understand what you mean by this
19:36 hmmmm [02:33 PM] <celeron55> if you have eg. max-min=25000, you get much higher chance of the result being <7768 or something than any other value
19:36 hmmmm did you find that out via experimentation?
19:37 electrodude512 joined #minetest-dev
19:37 celeron55 it randomizes a value that has a proper distribution between 0 and 32767 or something like that (i assume here that it does exactly this, it might differ a bit)
19:37 celeron55 so then when you feed that value to %25000, the distribution shifts
19:38 hmmmm ahhh
19:38 hmmmm i get what you mean
19:38 celeron55 you no longer get >25000 values and they add to the <7000 values
19:38 hmmmm the proper way to do this would not to use modulo but rather floating point division by the max
19:38 hmmmm yes I totally understand
19:38 hmmmm that is a rather obvious consequence of using such a mechanism
19:38 celeron55 well to begin with, i think it should generate enough bits to not fall into this problem at all no matter how it does it
19:39 hmmmm (float)next() / RAND_MAX) * (max - min) + min
19:39 celeron55 that would then consistently prefer some values over others still
19:39 hmmmm it generates 32 bits but only uses the lower 16 of them
19:39 hmmmm how would that?
19:39 celeron55 it should just detect that it needs a large value and do the equivalent of rand()<<whatever+rand()
19:39 celeron55 and then continue from there
19:39 hmmmm you're supposed to use the middle-ish 16 bits of m_next
19:40 hmmmm the lower bits in a LCG are notorious for having low enthropy
19:40 kilbith joined #minetest-dev
19:40 hmmmm oh sorry
19:40 hmmmm you use the top 16 bits
19:41 hmmmm so really it's m_next[16:31]
19:43 hmmmm how would what I proposed prefer some values over others, assuming the rand algorithm generates uniformly random bits (which it doesn't in practice)?
19:43 ImQ009 joined #minetest-dev
19:43 kilbith joined #minetest-dev
19:44 celeron55 because of the rounding
19:44 kilbith joined #minetest-dev
19:44 celeron55 if you squeeze 3 into 2, you'll get 2 in one bucket and 1 in the other
19:45 hmmmm is C-style truncate-to-zero rounding the issue?
19:46 Amaz joined #minetest-dev
19:46 hmmmm it seems so
19:47 hmmmm (int)((float)next() / RAND_MAX) * (max - min) + min + 0.5f)
19:48 hmmmm with this ^, pr.range(2,3) should be 2 50% of the time
19:52 MinerDad joined #minetest-dev
19:52 ImQ009 joined #minetest-dev
20:10 celeron55 hmmmm: but pr:range(0,32766) will be some value 2 times more often than the other values
20:10 celeron55 this was my original point
20:10 celeron55 it will happen no matter how you round it
20:10 hmmmm ahhh
20:11 hmmmm pidgeonholled
20:11 celeron55 precise random generators actually generate more bits than needed and *throw away* results that are out of the wanted range
20:13 celeron55 but i guess we can get away by generating a bunch of extra bits and just scaling or taking modulo of the result
20:15 n4x joined #minetest-dev
20:16 exio4 joined #minetest-dev
20:18 hmmmm that's how pseudorandom already works...
20:18 hmmmm it wouldn't hurt to have a better algorithm like the mersenne twister
20:22 kilbith Tesseract: you could add paramat as active contributor in credits
20:50 exio4_ joined #minetest-dev
20:51 y joined #minetest-dev
21:03 AnotherBrick joined #minetest-dev
21:16 casimir joined #minetest-dev
21:19 n4x joined #minetest-dev
21:19 exio4 joined #minetest-dev
21:26 casimir I already asked this in the forum, but will subgames be included in 0.4.11 ?
21:31 celeron55 the situation with them hasn't changed at all since 0.4.10
21:31 celeron55 so sadly and due to engine development being done in the wrong parts, i don't think they will
21:32 celeron55 there is only one type of subgame that is not nonsensical to consider given the situation of subgames and the engine
21:33 celeron55 that type is the "mt_game with way more content" type
21:33 celeron55 which basically holds dreambuilder and carbone
21:33 celeron55 (if you want to know more you could read today's logs)
21:35 MinerDad joined #minetest-dev
21:36 Miner_48er joined #minetest-dev
21:43 celeron55 i would like to hear it if someone has a good reason for objecting this decision
21:45 celeron55 i like where the tutorial, skytest and voxelgarden are going but each of them is limited by the engine right now
21:46 celeron55 and i am very dulled out by the mt_game-with-simply-more-content subgames
21:47 celeron55 does someone think that officially recommending carbone to be included with mt_game in builds would be a good idea?
21:48 celeron55 well, it's a fact that some people do
21:48 celeron55 does someone have a reason to object that being done?
21:52 Megaf_ The only game that should be officially included is minetest_game and the user must me told about choice and alternatives
21:54 Megaf_ hmmmm:  How hard would it be to implement Proxy support, like Socksv5 in Minetest?
21:54 kaeza that is exactly what we have now, and it's working marvelously /s
21:55 Megaf_ kaeza: that what?
21:55 kaeza Megaf_> The only game that should be officially included is minetest_game and the user must me told about choice and alternatives
21:56 Megaf_ kaeza: the one that is included is minimal
22:15 celeron55 informing the user would be good indeed
22:15 celeron55 doing it in the current menu or game would be very clumsy
22:15 celeron55 this stuff simply isn't up to scratch now
22:16 celeron55 i guess i will personally implement this for 0.4.12
22:16 celeron55 everyone else already had the chance, now i decide how it's done 8)
22:18 Anchakor_ joined #minetest-dev
22:18 celeron55 i for sure hope implementing it isn't worse than with irrlicht's native UI system
22:22 exio4 joined #minetest-dev
22:23 MichaelRpdx joined #minetest-dev
22:25 hmmmm wow
22:25 hmmmm it is so frustrating how sensitive noise parameters are
22:28 hmmmm tada!  http://i.imgur.com/Fu7YJEx.png  ore veins
23:02 pro joined #minetest-dev
23:10 luizrpgluiz joined #minetest-dev
23:10 luizrpgluiz hi
23:19 Bratmon joined #minetest-dev
23:21 MinerDad joined #minetest-dev
23:30 n4x joined #minetest-dev
23:31 luizrpgluiz when will end the test session of the new version of minetest?
23:36 kaeza I published an Inno Setup script to make an installer a while ago; I can modify it so it installs only minimal, and optionally download and install other games as well.
23:37 kaeza FWIW, CMake has the capability to create installers with such a feature too
23:37 kaeza Inno Setup is Windows-only I think; not sure about CMake-made installers
23:40 kaeza well, for Linux it doesn't matter anyway, as that work is carried by the relevant package manager
23:43 luizrpgluiz left #minetest-dev
23:58 celeron55 i don't think any distributions have minetest and games listed separately though
23:58 celeron55 at lest fedora and ubuntu don't
23:58 celeron55 least*

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