Time  Nick           Message
11:53 Krock          Are the changed Mapblocks updated instantly or is there some delay?
12:01 nrzkt          server side or client side ?
12:15 Krock          client side
17:01 paramat        can hmmmm or anyone approve #3194 ? i'm fairly sure it's okay but a quick check would be good
17:01 ShadowBot      https://github.com/minetest/minetest/issues/3194 -- Mapnode: Add rotateAlongYAxisFull supporting 24 facedirs by paramat
17:01 Calinou        ShadowNinja, merged https://github.com/minetest/minetest.github.io/pull/32 and now top icon is broken :(
17:02 Calinou        I know why, will fix it
17:02 hmmmm          paramat, sure I'm looking at it
17:02 hmmmm          question though, about the index variable
17:02 hmmmm          why is it param2 * 4 + rot exactly?
17:03 Calinou        pushed fix
17:05 hmmmm          the array size is 24 * 4, but they're broken up into 6 chunks of 16
17:07 hmmmm          also it would make me feel better if you made that (param2 % 24) * 4 + rot instead
17:23 paramat        no it's 24 chunks of 4, i copied the [24 * 4] format from a lookup table elsewhere. each line is the initial facedir (first value) then the 3 resulting facedirs under rotations 90, 180 and 270 degrees
17:28 paramat        the 6 chunks visible are just to separate facedirs 0-3 from 4-7 etc, i copied the format from https://github.com/minetest/minetest/blob/master/src/mapblock_mesh.cpp#L706
17:30 paramat        okay (param2 % 24) in case higher bits are used for something else, will do
17:40 paramat        thanks
19:41 VanessaE       nrzkt: are you aware that the public server list does not work (no content) on the android build currently offered on minetest.net ?
19:41 VanessaE       (tick the checkbox, nothing happens - list is empty)
19:53 paramat        now merging 3194
19:56 paramat        complete
20:41 VanessaE       wtf?  connection timing out while trying to load a SINGLEPLAYER world?!
20:41 nrzkt          VanessaE: no i'm not aware est31 tested the build, i only distribute it.
20:41 VanessaE       ok.  well now you're aware :)
20:41 nrzkt          since the armv7 change i cannot run minetest anymore on myphone
20:43 VanessaE       in my case I was playing around on my S4 phone
20:43 VanessaE       the timeout issue I just mentioned is on my PC, though.
20:46 VanessaE       ok so apparently I can't start a singleplayer game anymore.
20:47 VanessaE       not even a "minimal" world works.
20:48 VanessaE       seems to work fine connecting to a remote server.
20:56 est31          VanessaE, cant reproduce
20:56 est31          engine 1adc7bf5c60830cc8dcbab3
20:56 est31          game: ?? dunno, something outdated
20:56 est31          can you reproduce with minimal game?
20:56 VanessaE       [09-22 16:47] <VanessaE> not even a "minimal" world works.
20:57 VanessaE       engine is a56aedb4e
20:57 VanessaE       it just sits there "connecting to server..." and eventually times out.
20:58 est31          fatal: ambiguous argument 'a56aedb4e': unknown revision or path not in the working tree.
20:58 VanessaE       it's there, I just re-cloned and recompiled to check.
20:58 est31          ah paramat's commit
20:58 est31          lemme see
20:59 VanessaE       only thing I can think of is an ipv6 issue
20:59 celeron55      that happens on my laptop
20:59 celeron55      i have a setup with multiple local networks and whatever
21:00 est31          recompiling
21:00 celeron55      before i made it, MT didn't have issues
21:00 waressearcher2 where exactly place in source code that is limits world's size ? or its all around source code and not just one variable ?
21:00 est31          waressearcher2, there are multiple limits
21:00 est31          first there is one for world coordinates
21:00 est31          it cant be changed because its inherent to the type
21:01 celeron55      i think what this network set-up causes is that you can't just listen for UDP packets at "any" and send UDP packets to 127.0.0.1; they don't end up in that listening socket
21:01 est31          then there is a second limit to stop the mapgen reach that first limit and do funny things
21:01 VanessaE       of course my box does not respond to "ping6 ::1" (but I dunno if it should)
21:01 celeron55      it will work if i modify MT's source to connect to an explicit local address, in this case something like 192.168.0.5 or whatever
21:01 est31          then there is a limit for objects to get spawned, its set to the same as that mapgen limit
21:02 VanessaE       come to think of it, I can't ping 127.0.0.1 either.  must be a config problem on my end.
21:02 est31          also, as player, you can't get into non-generated blocks, so the mapgen limit exists for you as well
21:03 celeron55      (i can ping 127.0.0.1 just fine)
21:03 est31          yeah, even on a56aedb4e I dont have problems
21:03 VanessaE       ok looks like I just botched something in my routing.
21:03 celeron55      (and run TCP programs that listen at 0.0.0.0 and connect to them as 127.0.0.1 or "localhost"
21:03 celeron55      )
21:05 est31          waressearcher2, there is the setting map_generation_limit, does that help you?
21:08 paramat        it's because we use s16 (signed 16 bit integers) for world co-ordinates
21:09 waressearcher2 paramat: why not 32 ?
21:10 waressearcher2 est31: yes
21:10 paramat        heh
21:10 est31          I guess if we have 32, people will demand for 64
21:10 paramat        the discussion to increase world size has been had many times, it's not going to happen =)
21:10 est31          and if we have 64, people will demand 128 etc
21:11 paramat        to much work
21:11 waressearcher2 so s16 was choosen on purpose ?
21:11 est31          waressearcher2, some number had to be chosen
21:11 waressearcher2 you could've use s32 but choose s16 on purpose ?
21:11 est31          I think the old world format was somewhat limited to 1000 nodes or so
21:11 est31          so compared to that it was quite an enlargement
21:12 paramat        it's just big enough, remember the vertical space for stacked realms
21:13 est31          yeah 30km vertical is really enough
21:13 est31          or even 62
21:13 est31          one idea would be to do it like TBC_x suggested, and to not have absolute coordinates anymore
21:14 waressearcher2 how TBC_x suggested ?
21:14 VanessaE       there, that's sorted.  stupid network manager :-/
21:14 paramat        s32 is too big, currently the edge is distant but reachable
21:15 est31          well, we just dont have absolute x,y,z coords anymore
21:15 VanessaE       paramat: I've had people tell me they've been way out past +/- 31K on minecrarft before, though I agree that 31/62K for vertical is more than enough
21:15 est31          but instead the position is given relative to a center point
21:15 VanessaE       -r
21:15 est31          and ofc there is not just one center point
21:16 est31          so, for example, you divide the map into VAEs of size 62kx62kx62k
21:16 est31          and the moment a player reaches an edge of a VAE, a new VAE is created next to it
21:17 est31          and they are made "attached" so that there are no physics calculations on them
21:17 est31          there would be some need for mapgen magic to blend over between the two
21:17 VanessaE       in any case, we should either expand the horizontal limit to some ridiculous amount, or stop using the phrase "near infinite"
21:17 est31          but I guess that would be possible
21:18 paramat        yes 'almost infinite' should go
21:18 waressearcher2 paramat: "to much work", consider how much work was done for previous 6 month or a year on minetest, could it be more work than that ? I mean minetest code grows anyway, so why not to implement that is more requested ?
21:18 est31          that idea would bring the mapblock approach to coordinates
21:19 paramat        there are much more important things to do, and we are understaffed
21:19 est31          yea understaffed indeed
21:19 paramat        and a larger world is not much needed
21:19 waressearcher2 understaffed but not underpaid
21:22 VanessaE       some minecrafters will argue that a larger world IS needed. there's other stuff that's needed more, though.
21:22 paramat        i consider the smallest dimension of a world more important, MC may be vast horizontally but i consider it much more limited than MT due to world height
21:23 paramat        vertical space brings much more interesting things..
21:23 VanessaE       indeed, but there's only just so much vertical space people can find a use for
21:23 nrzkt          atm using better than u16 is a little bit difficult in core, many concerns
21:24 VanessaE       after a couple thousand meters, you've about exhausted what one could do I think.
21:24 VanessaE       nrzkt: as I understood it, the biggest concern is how it affects some Lua-facing coordinate hash functions
21:24 est31          yeah
21:24 est31          that would be a problem
21:25 est31          those hashes would need to be stored in userdata
21:25 VanessaE       for the code I've written that uses them, it wouldn't affect anything (my code in biome_lib, for example, only cares that there is a unique value, but not how it's hashed)
21:26 est31          well, the problem is that it won't be unique anymore
21:26 VanessaE       it wouldn't be unless you extend the hash format.
21:26 est31          general problem of hashing :)
21:26 est31          yea
21:26 VanessaE       what is it now?  z*(4096^2) + y*4096 + x  or some such?
21:26 est31          yea I think so
21:27 est31          and there isnt much space left
21:27 est31          < 20 bits
21:27 VanessaE       wait, you're talking about some feature in the map itself?
21:28 VanessaE       (or in how it's sent to the client)
21:28 est31          no about the hash
21:28 est31          the hash function outputs a lua number
21:28 est31          and usually its stored as IEEE float
21:28 VanessaE       well, I'm talking about internal representation in the engine and how it's passed along to Lua for e.g. the coords-to-hash and related functions
21:29 est31          and floats of this size have a limit of 2 ^ fifty something if you use them as ints
21:29 est31          coords to hash is entirely lua I think
21:29 est31          that isn't the problem
21:30 est31          if we have a position table, its stored as three numbers
21:30 est31          x y z
21:30 est31          all three have each the space of 2^ fifty something
21:31 est31          and only 15 are used
21:31 est31          so lotsa space left there
21:31 est31          but yeah, if we considered 64 bit based positions, there would be a problem
21:32 T4im           technically its not a hash function, or the size would not be an issue :P
21:33 est31          yeah just at these small numbers, it is very well possible that a collision occurs
21:33 est31          so not that good of an idea to convert it into a "real" hash function
21:40 VanessaE       but what position table are you talking about here, that's my question
21:40 VanessaE       if it's something internal to the engine, who cares how it's stored?  change it to something better.  If it's part of the map...well...do the same thing.
21:41 VanessaE       i.e. for the map, why be afraid of adding a couple of bytes to something that's gonna compress-away anyhow?
21:41 T4im           hash_node_position is also being used as keys in tables, which get serialized
21:41 T4im           for example autocrafter
21:41 T4im           ;)
21:41 T4im           eh sorry I mean tp tubes
21:41 T4im           their locations are stored... so are probably nodes of other mods
21:42 est31          VanessaE, the coords of a mapblock aren't compressed
21:42 est31          only its content
21:42 VanessaE       est31: what about what the database backend does to it?
21:42 est31          well yeah, that one can indeed do such things
21:43 VanessaE       (of course sqlite does nothing in that regard, but leveldb does)
21:44 VanessaE       if mods are gonna break, they're gonna break and nothing can really be done to fix that in the engine, but if the map format were to change to accommodate larger coord ranges, what's the big deal?
21:45 VanessaE       I mean, it's changed before, what, three times I guess?
21:46 VanessaE       and server/client protocol (where it outright breaks) has changed even more
21:46 est31          the problem is that we cant provide mods with a "please do this instead" hint how to stop them breaking
21:46 est31          because there is no better native lua datatype
21:46 est31          none that can store that much data
21:47 VanessaE       then introduce a new hashing function that doesn't care about coordinate ranges, and don't worry if the mods break.
21:47 est31          I mean if we really had 32 bit coords, we would need 128 bit integers
21:49 VanessaE       I see the problem with the numeric range in Lua though
21:49 VanessaE       that just means the hash would have to be returned as a string then.
21:49 VanessaE       with a BIG warning somewhere that it has changed.
21:50 T4im           it's a fixed index atm, 1:1 based on the size of the map.. there's certainly some way to encode the coordinates a bit "denser"
21:50 VanessaE       or new Lua function calls for the purpose, with warnings that the old calls don't return unique values outside the +/-31k range
21:53 VanessaE       all I'm saying is that it's not an unsolvable problem.
21:55 est31          agreed
21:57 est31          if we change the format though, we should also consider what to do with VAEs
21:57 est31          eg how to manage those
21:57 VanessaE       what is there to even do with them now?  they're vaporware :P
21:58 est31          VanessaE, some people say we should hack around the problem and store them into the "free" space we have thanks to the fact that we stop at 31k and dont go where the data types would allow us to go
21:58 VanessaE       fuck no
21:58 VanessaE       VAE's ought to be stored as their own entries in the map database.
21:59 est31          Thats my opinion at this as well
21:59 VanessaE       trying to use the "free" space at the edges of the map is gonna lead to headaches down the road
21:59 paramat        agreed
21:59 VanessaE       especially if this same "make the map bigger" argument comes up again
21:59 est31          it will break mods of course
21:59 est31          that has to be kept in mind
22:00 VanessaE       fair enough, but you can't break what hasn't been implemented yet
22:00 VanessaE       mods don't care where or how data is stored in the map anyway
22:00 est31          ah yeah
22:00 est31          in fact you are right
22:00 est31          it will break mods either way
22:01 est31          just people argue that if you store the blocks in that free space, you already have coords to give them so that you can use existing callbacks for modding on those positions
22:02 est31          that would be an advantage of that approach i admit
22:03 est31          but the more proper solution IMO is to give mods a fourth position argument, or a separate "realm" value.
22:04 paramat        wooo 4D
22:05 T4im           that would allow for very easy interaction between realms too :D place something in one realm, let it have some fringe effect on another in the same position
22:06 est31          lol
22:07 T4im           protection checking etc would be easier for a 4d position then as well
22:08 est31          heh lol these things are all bugs, not positive features
22:10 est31          you would have kinda of a discrete topology on the realm coord
22:12 est31          well, its topology is a bit more complicated
22:13 T4im           well, not talking about fading in and out between realms, though that would of course be a very near effect too :P
22:13 T4im           neat*
22:17 est31          but yeah talking with topology terms is quite stupid
22:21 VanessaE       another thought: extend the map, damn the torpedoes.  on startup, if the map is the extended variety, then disable all existing lua-facing functions that deal with coord hashes entirely, making only the newer functions available.
22:21 VanessaE       i.e. make mods that depend on them fail hard instead of silently fucking up.
22:23 VanessaE       all existing maps would be treated as they are now (with those functions available), anything created after that point is "extended" and the API only offers the newer functions
22:23 VanessaE       that's the sanest way I can think of, anyways.
22:24 est31          or just make it a setting
22:24 est31          it can work with both extended and non extended maps
22:24 T4im           expose some sort of map version to builtin, then adapt lua behavior accordingly?
22:25 VanessaE       T4im: that was my thought, yeah
22:25 est31          and then if you flip the switch the mods will fail
22:25 est31          but I think I dont want to extend it to 32 bit as well
22:25 est31          because I think its better done with VAEs
22:25 VanessaE       then extend it to 64 bit ;)
22:25 VanessaE       um, VAEs are supposed to be for big stuff that moves around I thought?
22:25 est31          and then its only limited by the VAE id counter
22:27 est31          I'd like to have a second non physically interacting VAE type that's position,rotation etc can either freely modified by lua without a physics engine asking it do do things, or be "attached" to other VAEs, or the "main root" map
22:27 VanessaE       mmh
22:27 est31          so if you walk into one direction for a long time, the server creates VAEs with new blocks
22:28 VanessaE       that just seems super-hacky
22:28 est31          and it allows the map to be far more larger
22:28 est31          because
22:28 est31          because you aren't limited in xyz per se
22:28 est31          you are only limited by the VAE id
22:28 T4im           why not solve the "Infinite world size" problem by letting people pop up on the other side of the map? :P
22:28 est31          which can be e.g. 32 bit
22:28 VanessaE       T4im: that would help to a small degree
22:29 est31          T4im, mods can already do this
22:29 T4im           by teleporting you mean?
22:29 est31          yea
22:29 T4im           nah, I mean tiling it seamlessly
22:29 VanessaE       but it doesn't do much to address the larger concern:  as much as it pains me to say it, minecraft beats the shit out of minetest on this front.
22:29 T4im           as if minetest would provide a sphere as map
22:29 T4im           (at least horizontally)
22:30 est31          T4im, you mean a cube?
22:30 est31          there are no spheres in minetest
22:30 est31          :)
22:30 VanessaE       technically it would be a torus wouldn't it? :)
22:31 est31          no
22:31 VanessaE       sure it would - north wraps around to south
22:31 T4im           but west to east too..
22:31 VanessaE       with a globe topology, you'd have to "bounce" off of the "pole" and end up somewhere further east or west
22:32 VanessaE       if only X wraps, it's a cylinder.  if X and Z both wrap, it has to be a torus.
22:33 est31          VanessaE, the difference between torus and sphere (or cube xD) is that on spheres you can contract every circle into a single point
22:33 est31          on a torus this isnt possible
22:33 VanessaE       fair enough
22:34 T4im           a hollow sphere
22:35 waressearcher2 speaking of spheres toruses and cubes as map http://www.youtube.com/watch?v=bJr4QlDxEME#t=60s
22:35 T4im           unless digging through the bottom should end weirdly :>
22:37 est31          I like how the renderer shows the landscape if the voxels aren't loaded
22:38 T4im           or better yet, est31 will rewrite the physics entirely and we let the player end up on a kleinbottle
22:38 est31          lol
22:38 T4im           i.e. if he walks far enough, he'll be upside down to the other players :p
22:39 VanessaE       hah
22:40 est31          http://media.indiedb.com/images/downloads/1/69/68101/thumb0.1.6.jpg
22:41 est31          this is way better than fog
22:41 T4im           nice
22:42 est31          (not for the fire, but about how it renders only part of the map as voxels)
22:42 T4im           certainly hart to mod/game change though
22:42 T4im           hard*
22:42 cg             ok guys, in .13 and .12 dev builds i cant save into the (worldpath)/player dir, is that now locked out or?????
22:42 T4im           from a mod you mean?
22:42 cg             yes
22:42 T4im           do you have security enabled? if so, then yes
22:43 T4im           probably that is
22:43 waressearcher2 est31: you said there is a mod that teleports player when he reaches edge, don't you think that mod could've achieved the same by just turning player 90 degree left or right ?
22:43 cg             it seems to have broken my mod again lol
22:43 cg             ive got it disabled
22:46 cg             http://pastebin.com/iqKKXq6T
22:46 cg             that code works untill 0.4.12dev
22:48 T4im           you show me read code there, not save-into-directory code
22:49 T4im           btw, you might want to look at some of the comments in https://github.com/minetest/minetest_game/pull/669/files
22:49 T4im           it applies to your code too ;)
22:50 T4im           oh, wait you're talking about the erase?
22:51 cg             it wont read or write inth the dir
22:51 cg             into*
23:02 T4im           cg: lets move that somewhere outside of -dev, I don't think it's an engine issue, so lets not bother this fine people; and perhaps send more context, it might help (is that part of cg72_game?)
23:04 T4im           (at least I can't reproduce any issue like that locally in a dev build)
23:05 VanessaE       sounds like the result of that mod security patch, if not a simple permissions issue.
23:06 paramat        http://i.imgur.com/fRA61E9.png mwahaha, a new option for mgv7