Time  Nick          Message
05:07 kahrl         I have a patch to reduce the number of needed textures by a ton if the atlas is not used: https://gist.github.com/kahrl/5703714
05:12 VanessaE      oh please do!
05:13 VanessaE      in fact, I thought the idea was to ditch the atlas entirely?
05:18 kahrl         I'm planning to but not now
05:20 VanessaE      right
05:20 VanessaE      now's not the right time for that.
05:20 VanessaE      just trying to save you a little work.
05:23 kahrl         when I remove the atlas I hope to also remove those "Could not open file of texture" errors in ~TextureSource
05:39 celeron55     ummm
05:39 celeron55     what does that patch do
05:39 celeron55     ah
05:39 celeron55     yes, that'll help
05:40 celeron55     i didn't even realize that could be done and wasn't already done
05:58 ShadowNinja   I don't understand how that would make any difference.
05:59 ShadowNinja   Hmmm, only applied to the atlas?
06:01 kahrl         ShadowNinja: most textures used to be added twice (once without and once with ^[forcesingle)
06:01 kahrl         therefore consuming twice the video memory
06:02 kahrl         but if the atlas is disabled, the two textures are identical so this isn't needed
06:02 ShadowNinja   Neccessary with the atlas, not without it?
06:02 ShadowNinja   Too late.
06:03 VanessaE      indeed, that patch cut my memory usage with HDX 512 down to ~5.5GB.
06:03 ShadowNinja   What does ^[forcesingle do?
06:04 kahrl         ShadowNinja: make a copy of the texture that is not part of any atlas
06:40 NakedFury     was that an error or mistake when the game was being coded?
06:40 NakedFury     oversight?
07:10 VanessaE      bbl
07:12 Anchakor      VanessaE: your client eats ~5.5GB?!
07:32 celeron55     NakedFury: it's something that was designed for low-resolution textures
07:33 celeron55     high-resolution ones need different kind of handling
07:33 celeron55     how we're practically dropping all low-resolution specific handling and hoping it doesn't degrade performance on low-end machines using low-resolution textures
07:33 celeron55     now*
07:33 celeron55     (not related to that commit, but rather an overall thing)
12:03 rubenwardy    Does minetest use UDP?
12:12 celeron55     yes
12:15 rubenwardy    just UDP, or TCP as well?
12:37 celeron55     TCP is used only for fetching stuff over HTTP when a server has set up such
12:39 proller       alreasdy used ?
13:06 proller       hmmmm,  #0  0x00000000004990d1 in CaveV6::makeCave (this=0x7fffff7fb8c0, nmin=Variable "nmin" is not available.
13:36 PilzAdam      I noticed something with the chat console: when a chat command opens a formspec the formspec gets the focus, but the chat console is still drawn above it, so everything is messed up
13:36 PilzAdam      this needs to be fixed for Exio's pull request I think
16:22 VanessaE      12:21:06: ERROR[main]: ERROR: An unhandled exception occurred: ServerError: LuaError: error: /usr/share/minetest/builtin/item.lua:339: attempt to call field 'handle_node_drops' (a nil value)
16:22 VanessaE      (git fetched just earlier today)
16:24 PilzAdam      maybe a mod made it nil
16:24 VanessaE      possible.
16:45 NakedFury     will hmmm be done with his world gen by 0.4.8 or .7 upcoming version?
16:46 celeron55     nobody knows; i guess it will take more time than that
16:47 celeron55     umm... except that it is definitely known that no further features go into .7
16:49 NakedFury     so its 0.4.8 the upcoming one?
16:49 NakedFury     .7 I mean
16:49 NakedFury     why am I thinking of 0.4.8
16:49 NakedFury     dammit
16:49 celeron55     yes; it will be released once it is deemed non-buggy
16:49 celeron55     and fixes have been going in all the time now
16:49 Exio          PilzAdam: any "example" mod for that?
16:50 PilzAdam      hm?
16:51 Exio          any mod what opens a formspec with a command? :P
16:51 Exio          i don't know how to write a "testmod" for that
16:51 PilzAdam      rubenwardy's capture the flag game
16:51 PilzAdam      dunno if its uploaded anywhere, tested it on his server
16:52 PilzAdam      just register a chat message and use minetest.show_formspec() in there
17:02 sweetbomber   hmmmm: almost done implementing the fast algorithm for biome distance calculation
17:03 sweetbomber   just two details: - is it bad if the height is a float?   - is it bad if height is defined not by min/max but by height_point as well?
17:04 sweetbomber   for this last one, the height could be given a higher weight, which would have +- the same effect
17:04 sweetbomber   nevertheless, i think it would put the code more clean this way
17:05 hmmmm         i guess it could be, but i think i had a specific reason for doing that
17:05 hmmmm         if there was a height_point instead, doing an O(1) lookup would be absolutely trivial
17:06 sweetbomber   how? with your current algorithm it would take the same time...
17:06 hmmmm         because it'd be a lookup
17:07 * hmmmm       rolls eyes
17:07 sweetbomber   how would you sort 3d points in a lookup table?
17:08 sweetbomber   in a 3d lookup table? that would wast loads of mem, no?
17:08 hmmmm         i figured it would at first, but it wouldn't be that bad
17:09 sweetbomber   even with infinite height?
17:10 hmmmm         the height ranges could be based off a tree structure
17:10 hmmmm         or something
17:11 hmmmm         er, but if it were just a regular lookup table, i guess you could still do a lookup with height; out of 65536 possible values, only a small portion will be really used, so they can be weighted like you said
17:11 sweetbomber   hmm...65536*100^2?
17:12 sweetbomber   anyway, cant you remember the reason for height min/max?
17:13 hmmmm         say, 64 divisions... 65536/64 = 1024, so without weights, each discrete segment would represent increments of 1024; with weights you can increase the granularity of a certain range (and necessarily decrease the granularity of some other range, like those higher up and much less significant)
17:15 hmmmm         i believe i wanted features such as beaches to start occuring at specific heights, otherwise it would look too ugly
17:15 celeron55     that sounds like something that can't be known without trying
17:16 sweetbomber   ill try giving more weight to height and see what happens
17:16 hmmmm         good idea
17:17 sweetbomber   imagine, if weight was infinite, it would mean a mandatory height.
17:17 sweetbomber   something between infinite and 1 would give the desired result
17:17 hmmmm         i wanted to add in weights but i never got around to it
17:17 sweetbomber   ill think if it is possible to give different weights to different biomes
17:18 hmmmm         different biomes?  of course it is
17:18 hmmmm         er, oh, you mean with your algorithm
17:19 hmmmm         with mine it'd be as simple as multiply the distance by a number
17:19 hmmmm         multiplying*
17:19 sweetbomber   meanwhile ill commit the code, if you can say if my coding is within minetest's standards
17:19 sweetbomber   i would appreciate
17:24 sweetbomber   https://github.com/sweetbomber/minetest/commits/fastCalcBiome
17:25 sweetbomber   for now just ignore error handling and printf's
17:27 sweetbomber   dunno if you have a big biome lua script, here is one randomly generated: http://pastebin.com/uGqM5P0g
17:27 sweetbomber   generates 300 biomes
17:27 Exio          PilzAdam: https://github.com/EXio4/minetest/commit/b18e5e4939a3865251d8089ff65ccbc82d0084a7
17:30 PilzAdam      Exio, seems good
17:31 hmmmm         300 biomes
17:31 hmmmm         yikes
17:31 sweetbomber   lol
17:31 hmmmm         erm
17:31 hmmmm         big problem
17:31 hmmmm         we don't accept spaces as indentation at all
17:32 sweetbomber   oh, no prob me neither
17:32 sweetbomber   that was just tweaking on gedit
17:32 sweetbomber   usualy i use vim with noexpandtab
17:32 hmmmm         from what it seems it's okay
17:32 hmmmm         couple stylistic things though
17:32 sweetbomber   thats what i wanna hear
17:32 sweetbomber   shoot
17:32 hmmmm         for variable names we don't use camelCase, camelCase is for method names
17:33 sweetbomber   ok
17:33 hmmmm         pointers ought to be type *foo;
17:33 hmmmm         not type* foo; or type * foo or any other variant
17:33 hmmmm         or at the very very least, keep it consistent throughout the source file
17:33 sweetbomber   ill use the one you pointed
17:34 hmmmm         we don't like lines above 80, and 90 chars is the hard limit
17:34 hmmmm         for macros just break them off onto a new line
17:34 hmmmm         see noise.h for examples
17:34 sweetbomber   sorry, didnt noticed that. i usually use wrap
17:34 sweetbomber   *will fix
17:35 sweetbomber   anyway, this code was just to get me warm
17:35 hmmmm         lol +  // Are there any clusters?
17:35 hmmmm         kind of obvious from the code
17:35 sweetbomber   i like to comment code with a bit of salt :p
17:36 hmmmm         also, not having whitespace between control keywords and the conditions is a celeron thing
17:36 sweetbomber   (no bad language! relax :)
17:36 hmmmm         most of us discourage that however
17:36 sweetbomber   that was a typo, i guess... troughtout the rest of the code i think i kept that convention
17:37 hmmmm         coding style is basically Linux kernel style
17:37 hmmmm         oh, you haven't seen this have you
17:37 sweetbomber   where is that?
17:37 hmmmm         http://dev.minetest.net/Code_style_guidelines
17:37 Exio          linux kernel style? camel case? tabs? wat ;P
17:38 hmmmm         modified a bit
17:38 hmmmm         linux kernel style is for C, so it can't be exactly the same
17:39 hmmmm         i wrote "something that resembles the Linux Kernel code style"
17:39 sweetbomber   no, "not having a whitespace between control keywords"
17:39 sweetbomber   :)
17:39 hmmmm         oh, er, that specific rule,  i don't think i have it written down
17:40 sweetbomber   ill read the style guide. did missed it
17:40 sweetbomber   just tried to reverse-engineer other code
17:41 hmmmm         ah
17:41 hmmmm         https://www.kernel.org/doc/Documentation/CodingStyle  section 3.1 Spaces
17:43 hmmmm         the code itself is very good though
17:48 sweetbomber   the kernel tips are nice
17:48 sweetbomber   ty
17:48 sweetbomber   im a bit rusty on code...dont program anything serious for a long time
17:49 sweetbomber   despite the fact that i like to code in my spare time....but thats another reason why i want to contribute to this project
17:54 hmmmm         are you an applied math guy?
17:57 sweetbomber   lol, nope :p
17:57 sweetbomber   electronics *.*'
17:58 hmmmm         ah, found it kind of interesting that you know so much about statistical/stochiastic stuff and used matlab for things
17:58 hmmmm         celeron55 does electronics
17:58 sweetbomber   heard about ml-class standford e-learning?
17:58 sweetbomber   quite interesting course
17:58 sweetbomber   did it last year
17:58 hmmmm         ah machine learning
17:58 hmmmm         i've read a little on the topic but i'm no expert
18:00 Exio          just bugging around, would anyone check https://github.com/EXio4/minetest/commit/b18e5e4939a3865251d8089ff65ccbc82d0084a7 ?
18:01 Calinou       just bugging around, would anyone check https://github.com/minetest/minetest/pull/409 ?
18:01 Calinou       :-)
18:02 VanessaE      Calinou: fuck no!
18:02 Calinou       well, make "climbable" accept a speed value
18:02 Calinou       create a node named default:glitch_ladder
18:02 Calinou       set climbable to 20 blocks/second :P
18:07 Jordach       Calinou, these are shift ladders right
18:07 Jordach       people still use them
18:08 * PilzAdam    votes for fixing the shift ladders
18:08 Jordach       many public minetest builds use them
18:08 Jordach       and its been here longer than you have PilzAdam
18:08 sweetbomber   hmmmm: sorry, lost internets
18:09 sweetbomber   my forever-alone log:
18:09 sweetbomber   "me neither, but it helped a lot to know what exists and where it is used"
18:09 sweetbomber   i recommend it
18:09 sweetbomber   there are a lot of problems that can be solved effortlessly with ml
18:09 PilzAdam      Jordach, following this, we will never have hardware lighting, since people use the current lighting and its longer here than I am
18:09 Jordach       PilzAdam, the people want hardware accelerated stuff, not gameplay elements
18:10 Jordach       or you wonder why Call of Duty sells with each release
18:10 Jordach       theyre drooling at the poly count
18:11 * sweetbomber imagines a game where poly count was the only metric used :D
18:11 Jordach       sweetbomber, thats how most modern games are unfortunately
18:12 sweetbomber   i was just taking it to the absurd...imagine no gameplay at all...just a load of random polygons :)
18:12 Zeg9          Dont fix shift ladders until someone makes a lift mod at least
18:13 PilzAdam      Zeg9, a mod like this is already done
18:13 VanessaE      don't fix sneak elevators at all.
18:13 PilzAdam      as laser or something like this
18:13 VanessaE      they're too entrenched.
18:13 NakedFury     fix them
18:13 NakedFury     they are a glitch or bug
18:14 Zeg9          PilzAdam: this is just a modified water iirc
18:15 Jordach       PilzAdam, if your term of bug is "Minetest" stop fixing what people use
18:15 Exio          https://xkcd.com/1172/
18:16 sfan5         since when does xkcd use https?
18:16 NakedFury     we really need to stop god damn glitches as gameplay or features. stop depending on them
18:16 PilzAdam      ^
18:18 sweetbomber   i'm ghosting myself more than usual...guess i'll have to fill a complaint to my fon neighbour
18:19 VanessaE      bbl
19:25 sweetbomber   hmmmm: in calcBiomes is the height right?
19:26 sweetbomber   i tp'd myself to 0 1000 0 and the height prints it still give around -10~10
19:26 sweetbomber   just to get it straight: isn't the height supposed to be the height of the block being generated?
19:27 sweetbomber   or is it directly from the height map?
19:27 sweetbomber   otherwise, how could sky-high biomes be generated?
20:22 sweetbomber   hmmm: got some profiling -> http://pastebin.com/uU553ec0
20:52 hmmmm         the height in there is the elevation of the land
20:52 hmmmm         hmm... so in general it's a 2-3x improvement.... i was expecting much more
20:53 hmmmm         and this is @ 300 biomes, right?
21:55 sweetbomber   yup, that is @ 255 biomes, but time is not O(n) now
21:55 sweetbomber   ill try with less
21:59 sweetbomber   i was expecting more too.
21:59 sweetbomber   do you think its worth finishing it?
22:09 sweetbomber   hmmmm:  50 biomes: http://pastebin.com/jub9Q9UA
22:11 hmmmm         eek :/
22:11 hmmmm         maybe not
22:11 sweetbomber   on the other hand it is much more scalable
22:11 hmmmm         with even less biomes i'm assuming the situation is somewhat worse
22:11 hmmmm         the original would come out ahead with the normal amount
22:12 sweetbomber   with 5 biomes its about the same
22:12 hmmmm         i don't intend to have more than 20 biomes in practice
22:13 hmmmm         (of course, you can imagine that modders would go overboard with biomes like they do all things)
22:14 hmmmm         the 255 biome limit is there basically so that biome maps don't need to be more than 8 bits, and to encourage people to not go insane with it
22:14 hmmmm         i feel a lot of modders don't really put any imagination into what they make and end up adding content rather than something well thought out and fun
22:16 hmmmm         I know how minecraft deals with this problem... they simply don't take elevation into account for biomes
22:18 kaeza         but then again, MC's worlds are not very high
22:28 sweetbomber   (sry, phone)
22:29 sweetbomber   there was one thing that i did not understand about biomes
22:29 sweetbomber   i tought that the mechanism to create e.g. aether was with biomes
22:30 sweetbomber   so that there were biomes with height of 500 ~ 4000
22:30 sweetbomber   specifically to generate land in the skies
22:30 sweetbomber   but now it seems like it isnt the case
22:32 sweetbomber   in that case, biomes generated in aether would have as "seed" the temperature, heat and height of the ground-zero world, right?
22:38 hmmmm         see, i considered this, and it would be covered by the "biome type" case
22:38 hmmmm         you remember that createBiome factory function?
22:38 sweetbomber   yup
22:39 hmmmm         well those were for biomes with fundamentally different terrain generation methods
22:39 hmmmm         while that specific way of doing things turned out to not work well, i can still use the same type of idea for aether/nether
22:40 hmmmm         i can have a height range where a new set of biomes occur, and this 'realm' i guess you can call it has its own generateTerrain()
22:40 hmmmm         need to work out the details, but fundamentally, that's how it *must* be accomplished
22:41 hmmmm         right, so really i guess we can say that a realm occurs in a certain range of height and has a set of possible biomes
22:41 hmmmm         each realm has its own generateTerrain method
22:41 hmmmm         and right now, the only supported realm is the "normal" realm
22:43 sweetbomber   hmm
22:43 hmmmm         i can do layers of crust this way too, come to think of it
22:45 hmmmm         so i guess, in between the normal realm and the nether realm, there can be the "mantle" realm, which has a generateTerrain() that simply sets the block to whatever the filler node of the biome is
22:45 hmmmm         which can be all sorts of things like iron blocks and whatever, and then lava, something really interesting
22:45 hmmmm         and perhaps there are no caves, smaller caves, or at least less caves in this "realm" too
22:46 sweetbomber   one thing, is your perlin noise really 3d? dont get me wrong, my only question is if it returns a point in (x, y,z)?
22:46 hmmmm         errm, we use both 2d and 3d
22:46 sweetbomber   yes, but particularly speaking of 3d
22:46 hmmmm         well *i* use 3d for mapgen v7
22:47 hmmmm         and of course, 3d perlin noise by definition is w = f(x, y, z)
22:47 sweetbomber   can you give me an idea of how much time does it take to generate a block?
22:48 sweetbomber   (generate a block of 3d perlin noise)
22:48 sweetbomber   (no rendering)
22:48 hmmmm         well
22:48 sweetbomber   rought estimate, oc
22:48 hmmmm         i have a xeon e3-1230v2, and at -O1 with gcc 4.2.2, it can generate an 80x80x80 buffer of 3d noise with 4 octaves in 22ms
22:49 hmmmm         so, 22ms / 25 = your answer
22:49 hmmmm         all types of perlin noise are insanely fast after i optimized it
22:50 hmmmm         for almost all purposes, you can consider it "free"
22:50 VanessaE_     s/25/125/
22:50 hmmmm         erm, i mean 25
22:50 hmmmm         oh
22:50 hmmmm         i mean 125
22:50 VanessaE_     math FAIL ;)
22:51 hmmmm         that's a lot more blocks than i thought
22:51 hmmmm         holy crap i've been making that mistake for a long time
22:51 VanessaE_     "He is intelligent, but uneducated.  His pattern suggests two-dimensional thinking."
22:51 VanessaE_     (jk)
22:52 hmmmm         my estimate on generating blocks vs reading from disk was off too
22:52 hmmmm         no more than 100ms for *125* blocks, vs. 30ms for *1* block
22:53 hmmmm         we need leveldb FAST
22:53 hmmmm         we also need to optimize mesh making somehow
22:53 VanessaE_     G*d yes.
22:53 VanessaE_     or wait, ignore that.
22:53 hmmmm         i haven't payed much attention to mesh making because it wasn't my area
22:53 VanessaE_     (I'm thinking of something else)
22:54 hmmmm         on singleplayer it's the single biggest bottleneck, don't forget
22:54 hmmmm         i am also looking forward to kahrl's IP stack bypass for singleplayer
22:55 VanessaE_     one thing I've noticed, and I don't know if this is the mesh builder or the extrude routine, is running around on my server and having my client stall from time to time in very short bursts (say 1/2 second).
22:55 hmmmm         vanessae, what do you think about the realm idea?
22:56 VanessaE_     lemme re-read
22:56 hmmmm         would you go for that?  an area of map above the nether that's made up completely of lava, and above that is solid iron blocks, and so on
22:56 sweetbomber   the idea is nice
22:56 hmmmm         this is more of a gameplay idea, but the realm concept would be motivated by this
22:57 VanessaE_     hmmmm: yeah, I think I could get used to that, provided the default realms are sufficiently separated in height (none of this -5000 crap for nether, for example - it belongs at -30000 or lower)
22:57 Anchakor      so if you build a big tower you hit the bottom of nether? :P
22:57 VanessaE_     let's really USE that 62K height range
22:57 hmmmm         there's never going to be any Y-axis wraparound, no idea who came up with that
22:58 hmmmm         vanessae, i agree too, i think the nether should be a lot lower than it is
22:58 VanessaE_     Y-wraparound is dumb, but X/Z wraparound would be smart
22:58 VanessaE_     (if it's doable)
22:58 hmmmm         many think that you ought to be able to dig to it
22:58 Anchakor      could you do flying islands biome with that?
22:58 sweetbomber   Y-wraparound would allow one fall forever :D lol
22:58 VanessaE_     dig to the nether?  um.
22:58 sweetbomber   *to fall
22:58 hmmmm         vanessae, celeron said -800
22:58 hmmmm         ;/
22:58 VanessaE_     it's supposed to be another realm, so for all practical purposes it should be impossible to dig to it at all.
22:59 VanessaE_     -800?  for nether?
22:59 hmmmm         yep..
22:59 hmmmm         crazy huh?
22:59 hmmmm         no way am i doing that
22:59 VanessaE_     ehm, wut?
22:59 VanessaE_     (to coin a phrase)
22:59 Anchakor      different realms should really be different maps
22:59 VanessaE_     we have a 62 km high map, let's fricken USE that height.  Look at what paramat is doing, for example
22:59 hmmmm         i agree with -30000, i'm also going to have to work on making the map generation cutoff more predictable though
22:59 VanessaE_     Mt. Meru or whatever it is, is 1km high
23:00 hmmmm         i saw that and it's awesome
23:00 hmmmm         no idea why paramat doesn't contribute to the indev mapgen
23:00 hmmmm         he has the skills
23:02 VanessaE_     good question
23:02 VanessaE_     I can only guess it's the same issue prestidigitator ran into maybe
23:02 VanessaE_     (which on the surface may not make sense but I think you get what I mean)
23:02 hmmmm         prestidigitator wanted to make huge, sweeping changes with almost no benefit though
23:02 VanessaE_     true
23:05 VanessaE_     now, about the mesh building code
23:05 VanessaE_     where would I see the effect of slow-building, in practice?
23:05 VanessaE_     just general map loading, or those hiccups I mentioned? or somewhere else
23:25 PilzAdam      https://github.com/PilzAdam/economy/commit/6bd53c97564906c61e750416f09f7c03d62061e8
23:26 PilzAdam      ooops
23:26 PilzAdam      wrong channel
23:26 * VanessaE_   pokes hmmmm
23:32 proller       VanessaE_, look at "sky" server with spawn on ~1km mountain
23:59 VanessaE_     can someone please PLEASE fix the horrible CPU usage of the liquid flowing code?