Time  Nick         Message
00:04 paramat      hey hmmmm ive looked at your paste of generateBiomeColumn and there are some errors, also last night's discussion uncovered some errors in my original code, so because i wrote the original code and understand it please could i work on this? i would like to first correct my original code, then study and edit your generateBiomeColumn function to be logically identical
00:06 paramat      is (u32)-1 just a huge number?
00:11 paramat      essentially i feel the task of moving generatebiomes to biomemanager needs to be postponed, i need about a week to sort this out
00:23 hmmmm        what errors?
00:24 hmmmm        i went through it to make sure that it's logically correct, and i also corrected a couple places where you checked for non-air when in fact you meant to check for solid
00:24 hmmmm        thanks for looking at it though
00:30 paramat      okay i'll explain the errors..
00:33 paramat      need_biome_recalc is set to true so biome is always calculated at the top of every column. half of a world's chunks have air at column top so there is unnecessary calculation of biome 6400 times per chunk for a half of a world's chunks
00:35 paramat      my code only calculates biome if stone or water is detected
00:37 hmmmm        that's easily fixable
00:37 hmmmm        i was expecting some kind of deep, logic error that I missed out on
00:37 hmmmm        you scared me for a bit
00:37 paramat      heh
00:38 hmmmm        hmmm
00:38 paramat      line 13 checks for stone at the base of the chunk above, if the chunk above is already generated the overtop node could be anything, not just stone
00:38 hmmmm        for the next-gen biomes, I'm planning on making getBiome only consider heat and humidity so it'd be able to read from a precomputed table
00:39 hmmmm        this would make getBiome calls computationaly cheap
00:39 hmmmm        for what it's worth
00:39 hmmmm        hrm you raise a good point, I didn't think about that
00:39 hmmmm        no wonder why you made it check explicitly for water or air
00:40 hmmmm        anything else? so far
00:40 paramat      yeah..
00:41 paramat      i'm not sure about the logic for triggering recalc of biome, dependant on nplaced != 0
00:41 hmmmm        if nplaced is already 0, then that means it hasn't encountered stone yet
00:42 paramat      recalculation should only be triggered when there is stone or water below an air node, then that biome is in effect for all nods below until air is placed again
00:43 paramat      water below air needs to trigger recalc
00:44 paramat      ..then that biome tajes care of the seabed below
00:45 paramat      i admit im still trying to get my head around your logic
00:45 paramat      i like some of the tricks you've used to reduce redundancy
00:46 hmmmm        hrmmm
00:46 hmmmm        wait why does water below air need to retrigger
00:46 hmmmm        i didn't think of that tbh
00:46 hmmmm        i tried to make my logic super simple and understandable
00:46 paramat      becaus there's now biome secific water, whereas originally there was only dust-on-water
00:47 paramat      (specific)
00:48 paramat      so water counts as a new upper surface that triggers recalc
00:48 paramat      before it was the seabed that did that
00:49 paramat      the biome calculated at water surface then includes the seabed below, so after calculation at sea level no more calculation is needed
00:50 paramat      so, next-gen biomes won't consider altitude?
00:51 paramat      anyway, back on topic, it may be worth adding in checks for river water as this is now added as the 4th base terrain node
00:52 paramat      and will be used in v8 and my own new core mapgen i'm struggling with
00:55 paramat      but that could be added later or river mapgens could have their own genbiiomes function
00:55 paramat      (struggling with highland pools, very difficult)
01:00 paramat      anyway, lets not worry about river mapgens for now, their genbiomes function needs to be quite specialized so perhaps best as a separate function
01:01 paramat      in your code i think 'depth filler' should be called 'base filler' to not confuse it with biome->depth_filler which is different
01:03 paramat      finally, why nplaced = (u32)-1 is that a huge unsigned number?
01:16 paramat      i just realised i was wrong to say water should set have_air to false, that is needed to prepare for the seabed. my mistake is that have_air then triggers biome recalc for every water node
01:17 paramat      so don't base your function on my code, it has errors. i think we need a new 'air_above' flag
01:17 paramat      i need a few days to understand your code, and fix my own, and since we both seem a little confused =)
01:24 paramat      biome calculation should be triggered 'if (c != CONTENT_AIR && (y == node_max.Y || air_above)) .. i think
01:24 paramat      anyway i'll work on fixing my current code..
02:13 hmmmm        oh right
02:13 hmmmm        alright, three things to adjust
02:14 hmmmm        the last one makes the code uglier :/
02:34 hmmmm        paramat, if you read this: (u32)-1 is because we need an nplaced that is guaranteed higher than any possible max filler depth so that it'd go right to placing stone and not top/filler nodes
02:35 hmmmm        i'd set it equal to the max filler depth, but in that scenario we don't know what that would be since a biome has not been calculated yet
02:52 paramat      thanks, this is what i have so far https://github.com/paramat/minetest/blob/feca5e6c0c048f2fbb9c72b9260d8597c7ebabd4/src/mapgen_v7.cpp#L592
02:53 paramat      still needs testing though, and i'll add that (u32)-1
02:53 paramat      bbl
03:03 hmmmm        I think I might be able to do this without any added booleans still
03:05 hmmmm        is 128 the maximum base_filler?
03:06 hmmmm        also i wanted to ask you about stone_type
03:07 hmmmm        i realize that taking the X+, Z+ corner biome stone type seems to work sufficiently, but what is the absolute correct way of calculating this?
04:05 * paramat    reads
04:06 paramat      base filler of 128 needs changing to (u32)-1, i just chose a big number
04:09 paramat      currently stone type is set to stone by default (for deep underground), then during any calculation of biome if any biome containing desert stone or sandstone is chosen at any point stone type gets changed
04:09 paramat      also the default of stone means empty chunks get cobble dungeons
04:11 paramat      so sonce it's per-chunk often there will be temples or sandstone dungeons up to 80 nodes away from deserts or sandstone biomes
04:11 paramat      perhaps ideally it would be calculated for chunk centre
04:13 paramat      so my head cleared. i realised the problem with to your code was the logic controlling the triggering of biome recalculation. it must be independant of nplaced because consider a series of underwater ledges: nplaced will go through several cycles but biome should not be recalculated
04:14 paramat      so the air_above bool seems essential
04:15 paramat      studying your code and working on mine i know what you mean by 'messy and with redundancies' studying your code has been very helpful and i integrated your improvements into my latest code
04:16 paramat      air needs chacking because a chunk does not only contain air, water or stone. there will be bits of overgenerated tree and dungeon too
04:17 paramat      so the main 'if elseif' block needs an 'else' at the end to deal with these solid tree and dungeon nodes
04:22 paramat      for stone type, the problem is that due to biomes having ymin/ymax if we choose a single point to choose stone type it may miss the biome containing desert stone or sandstone. so i suggest leaving it as it is
04:23 paramat      as it is there's no chance of missing the triggering stone
04:34 paramat      actually i'm not sure base_filler needs to be set to a big number at the top, perhaps not
04:40 paramat      oops my nplaced should be a u32..
05:25 paramat      errors fixed and quickly tested https://github.com/paramat/minetest/blob/874d0e0475e3b3915a88565121135350bcd7035f/src/mapgen_v7.cpp#L592
15:38 kilbith      have a dumb question... do only the mapblocks exposed faces or out of camera's view are only rendered ?
16:00 kilbith      s/out of/inside
16:01 Calinou      we do use view frustrum culling, as well as occlusion culling
16:01 Calinou      occlusion culling is never perfect, but we believe it's aggressive enough
16:02 kilbith      ok thanks
16:11 RealBadAngel Calinou, you may want to try it: https://github.com/RealBadAngel/minetest/tree/minimap3
16:15 Calinou      do you need to use a setting to enable minimap?
16:16 RealBadAngel no i dont need that setting
16:16 RealBadAngel i think i will remove it
16:17 RealBadAngel but on the other hand later on i plan to allow control minimap from lua
16:17 Calinou      is there a key binding? I don't see minimap
16:17 RealBadAngel F9
16:17 Calinou      F9 already bound to cinematic here…
16:17 Calinou      <_<
16:17 RealBadAngel https://github.com/RealBadAngel/minetest/commit/a3dd74534ac56c7598c72cc397b1a621ba2d74f7#diff-3474ba74d8eb9c72a47c6542f51eb4fcR52
16:17 Calinou      oh it's changeable in GUI
16:17 Calinou      thanks
16:18 Calinou      oh no it isn't
16:18 Calinou      only cinematic is changeable, not minimap
16:18 Calinou      it should be changeable in GUI
16:18 Calinou      maybe I want X for that
16:18 rubenwardy   F8 is cinematic
16:18 RealBadAngel thats good idea to let user define the key
16:19 Calinou      you should make Sneak + Minimap toggle go the other way around
16:19 Calinou      ie. zoom out
16:19 Calinou      also why is the map not in the top right corner? it's almost at the center of the screen
16:19 RealBadAngel btw, with threaded scanning we can afford 256x256 as zoom x1
16:20 RealBadAngel calinou, hillPlaneMesh has flaws
16:20 RealBadAngel i will use another method for it
16:20 RealBadAngel but it definitely have to be a mesh
16:21 Calinou      your fading function is quite bad, when I'm on mountains I barely see stuff on minimap
16:21 RealBadAngel i have already found code which does such thing
16:21 RealBadAngel Calinou, thats the point
16:21 RealBadAngel youre too far away
16:21 Calinou      make it optional :/
16:21 Calinou      Rei's minimap had such an option
16:22 RealBadAngel no i cant, otherwise you wont be able to scan for surface on other levels
16:22 RealBadAngel structures in the air, etc
16:22 RealBadAngel theres no fixed surface level
16:22 RealBadAngel and btw, do you know how many nodes it scans to get 256x256 minimap?
16:23 Calinou      256³?
16:23 RealBadAngel 8388608
16:23 RealBadAngel 256x256x128
16:23 Calinou      whoa, your minimap goes behind blocks when you're in third person
16:24 RealBadAngel for surface mode, scan height is 128 nodes, for radar 32
16:24 Calinou      https://lut.im/TCZnoYid/WAHXhhP6
16:24 RealBadAngel heheh, funny
16:24 RealBadAngel position and size is correct only for 4:3
16:25 RealBadAngel pity that method doesnt work for ohter ratios
16:25 Calinou      make position changeable by user :P
16:25 Calinou      also, what about making minimap transparent instead of black when it scans nothing?
16:26 RealBadAngel doesnt look good
16:26 RealBadAngel at least imho
16:26 RealBadAngel makes the minimap look like crap
16:27 RealBadAngel ofc when not complete yet
16:27 RealBadAngel about position, texures etc, all will be changeable from lua
16:28 RealBadAngel so games could make minimap a tool, aviable via crafting for example
16:29 RealBadAngel you will be able even to use minimap texture in formspecs
16:30 RealBadAngel 512x512 doesnt take too long to scan
16:31 Calinou      scan only once per second or so
16:31 Calinou      only when player moves
16:31 Calinou      there should be max rate, in case you move very fast
16:31 Calinou      10 scans per second sounds reasonable
16:31 Calinou      20 maybe
16:31 RealBadAngel scannin is done in separate thread
16:31 Calinou      definitely not more
16:32 RealBadAngel if you will trigger refresh on player move you will not get loading blocks around you
16:32 RealBadAngel until you move again
16:32 RealBadAngel anyway, scanner itself doesnt slow down fps at all
16:33 RealBadAngel only the code that shows the texture
16:33 Calinou      how much times per second do you scan?
16:33 Calinou      should be limited in any case
16:33 Calinou      especially when user has menu opened
16:33 RealBadAngel what for?
16:33 Calinou      to limit CPU usage
16:33 RealBadAngel its in another thread
16:33 Calinou      doesn't mean you shouldn't limit it
16:33 RealBadAngel when not scanning it will run in empty loop
16:34 RealBadAngel no point in limiting anything
16:35 RealBadAngel it runs in background, reports to main thread if texture is ready. once grabbed it starts new scan
16:36 Krock        ShadowNinja, there's an ugly calculation error somewhere in s_security.cpp. 'num_read' is '1088' when 'size' is '1126'
16:36 Krock        Lua fails to open builtin\init.lua
16:36 RealBadAngel Calinou, anyway, just compare FPS with and without minimapper
16:38 RealBadAngel but on the other hand the code can be optimized for sure, like everythin else
16:40 RealBadAngel also, about fading. if i remove it the area out of scanner reach will just disappear
16:41 RealBadAngel and you wont see difference on different planes full of same nodes
16:41 RealBadAngel all will become just flat
17:00 est31        pushing in 5 minutes https://github.com/est31/minetest/commit/36b58964b2b4be37ea667ab263d719f818cc093d
17:01 est31        err
17:01 est31        doesnt build
17:01 est31        will fix that
17:01 est31        then pushing the fixed version
17:02 est31        https://github.com/est31/minetest/commit/9facb40738e7c51c838ad2ccffd96028862160cb
17:06 Krock        est31, could you also push a commit to change the "r" to "rb" in  https://github.com/minetest/minetest/blob/master/src/script/cpp_api/s_security.cpp#L267 please?
17:07 Krock        (fixes windows builds)
17:08 est31        eh, is this \r\n stuff?
17:11 est31        hm, this basically boils down to the question whether we should have \r\n endings on windows, or \n.
17:12 Krock        est31, no, it's about windows special features. It needs a 'b' to read every character as how it is
17:12 rubenwardy   Isn't rb just read binary
17:12 est31        yes
17:12 Krock        rubenwardy, aren't text files just binary files?
17:12 rubenwardy   Yeah...
17:12 rubenwardy   I wasn't saying they weren't
17:12 Krock        however, without that additional 'b' character, it results in a wrong character count on reading
17:13 est31        so, then that should be fixed instead
17:14 est31        yea, quite the error happened forecasted here: http://c-faq.com/stdio/textvsbinary.html
17:15 TeTpaAka     Thank you, est31.
17:16 est31        thanks for pointing out :)
17:18 TeTpaAka     Is there any Information available for client side scripting? Or is there currently no design for it?
17:19 est31        hmmmm has one, but hasn't yet talked about it.
17:19 rubenwardy   It's on the core dev's todo list, but there's been no real design discussion yet
17:19 rubenwardy   I'd be interested in a GH issue, wiki page or maillist
17:19 est31        but generally I wouldn't rely on any features while its still in development
17:20 est31        or before implementation has even started
17:20 TeTpaAka     Yes, I know. But he pointed out, that features like nametag colors should be done client side.
17:20 est31        yes
17:21 TeTpaAka     And so I wondered how it would be done. How does the server tell the client, which function to run?
17:21 TeTpaAka     Does the server provide the code or does the client has to download it?
17:21 est31        the server provides the code
17:21 TeTpaAka     Ah, ok.
17:22 est31        we won't make it like minecraft, where people have to download stuff only to join servers
17:22 est31        of course, this also opens a security risk
17:22 TeTpaAka     Sound reasonable.
17:23 TeTpaAka     But would it be possible to install client side scripts that only alter the appearance and don't have any impact on the gameplay?
17:23 est31        how do you mean that?
17:23 TeTpaAka     Something like shaders...
17:23 rubenwardy   The user should be notified that the server is sending client side scripts, though, and have the option to cancel #2324
17:23 ShadowBot    https://github.com/minetest/minetest/issues/2324 -- Dialogs whilst game loading
17:23 rubenwardy   It should also display licenses etc
17:23 rubenwardy   ie: get out of the JS trap
17:24 est31        why is js a trap?
17:24 est31        this discussion again
17:24 est31        I hate it
17:24 est31        only because it runs on your box, why does it have to be open source
17:24 est31        as long as its inside a sandbox who cares
17:26 est31        already now we have no license for the map
17:26 rubenwardy   I don't care, others do
17:27 est31        Calinou mostly
17:28 est31        but yes the sandbox has to be very strong
17:31 Calinou      running untrusted client code has implications
17:31 Calinou      it's not because browsers did a wrong thing we should do it again
17:32 est31        which implications Calinou ?
17:32 Calinou      ethics, security
17:33 est31        is there ethical and unethical javascript?
17:33 RealBadAngel hmm, whats more dangeroes, client side code which user can see and inspect or connecting to server with its mods (able to act client side) which user cannot see at al?
17:33 Calinou      the ethical one is free, the unethical one isn't
17:33 Calinou      and the unethical one is sometimes even distributed in obfuscated form
17:33 Calinou      RealBadAngel, the server-side code can't do harm to your PC
17:34 est31        yes, security is a big point
17:34 rubenwardy   That's your version of ethics, though. For the majority of the world non-free is still ethical
17:34 Calinou      some server admins might get angry at their userbase and start distributing dangerous code
17:34 Calinou      and some other admins will obfuscate it to prevent “theft”
17:34 est31        so what
17:34 RealBadAngel calinou, i bet i could do a littlle mess on client side with servers mods
17:35 est31        the server code is protected from "theft" too
17:35 RealBadAngel like filling up whole clients disk for example ;)
17:35 est31        Calinou, do you want to put minetest into AGPL?
17:36 Calinou      no
17:36 RealBadAngel but whatever, we do need client side mods
17:36 est31        I think we should call it client side lua
17:36 RealBadAngel without them we wont get any good mobs, carts etc
17:37 est31        yea carts are shit
17:37 est31        currently
17:37 Calinou      add a way to vizualize code from clien
17:37 Calinou      like Web browsers do
17:37 Calinou      at least we can shame server owners who obfuscate their code :)
17:38 est31        perhaps
17:38 est31        we need some way to debug yes
18:01 jin_xi       i like helicopter mod, but it breaks badly when you ride it across camera offset. no matter how buggy the mod, such internals should not interfere with users
18:02 jin_xi       anyway i hope that the possibility of future features does not stop people from trying to make current ones better
23:54 paramat      thoughts on game#519 ? i realised it would not break existing worlds to enable by default
23:54 ShadowBot    https://github.com/minetest/minetest_game/issues/519 -- Should mgv6 snowbiomes be enabled by default?
23:57 est31        why breaking?
23:57 est31        "why does it break"