Time  Nick        Message
00:28 hmmmm       Zeno`:  So have you or have you not been able to replicate it?
00:30 Zeno`       hmmmm, yes. But it's strange
00:30 Zeno`       I've been able to replicate it 75% let's put it that way. And the issue happens before your commits anyway
00:31 hmmmm       A++++++ accusations would raise again
00:31 hmmmm       s/raise/rage/
00:31 Zeno`       That would apparently be the case
00:33 jin_xi      hmmmm: im currently using a double vm setup to first find nodes placed by decorations and then replace them with generated structures. this produces holes in the map reliably
00:33 jin_xi      anyways, if decorations could directly call some lua it would be easier
00:33 Zeno`       hmmmm, as for the assertion failure (which I didn't report I don't think), it hasn't happened again so it must be some weird and very rare case. The server is still running under gdb anyway
00:34 hmmmm       jin_xi, that shouldn't happen.  can you show me how you're running it?
00:34 hmmmm       Zeno`, my guess is that it might occur in out-of-memory conditions.
00:34 hmmmm       a/in/under/
00:35 Zeno`       hmmmm, you might be right
00:35 jin_xi      sure, gimme a sec
00:35 hmmmm       that's literally the only condition in which get_data() could throw an exception
00:39 jin_xi      http://paste.ubuntu.com/8294886/ http://i.imgur.com/ESb1hYw.png
00:39 hmmmm       is that hole made of air or content_ignore?
00:40 jin_xi      how can i check?
00:40 hmmmm       try walking into it with noclip disabled
00:40 jin_xi      can walk
00:42 hmmmm       okay
00:42 hmmmm       anyway so find_dummies is run first, in on_generated
00:43 hmmmm       by the way jin_xi, that code is actually not required at all.  you're not using a rather obscure mapgen object that was made for this situation.
00:43 jin_xi      pls tell me more
00:43 hmmmm       decoration callback
00:43 jin_xi      its there?
00:44 jin_xi      oh
00:44 hmmmm       yes, it's been there for about a year now
00:44 jin_xi      i missed that, i asked for it long time ago
00:44 hmmmm       i used it to place chests inside of dungeons and desert temples
00:45 hmmmm       see "gennotify"?
00:46 jin_xi      um, yes but i fail to parse
00:46 hmmmm       actually sorry about that
00:46 hmmmm       shoot
00:46 hmmmm       it didn't have a schematic flag
00:47 jin_xi      Returns a table mapping requested generation notification types to arrays of positions at which the
00:47 jin_xi      corresponding generated structures are located at within the current chunk.
00:47 jin_xi      ???
00:47 hmmmm       i'll have to expand gennotify to include decorations
00:49 jin_xi      oh i see now. took a lot of repeated reading
00:50 hmmmm       ah uh oh
00:50 hmmmm       i don't think i have minetest.set_gen_notify() documented
00:51 jin_xi      it is kindof obscure as it is now, yes
00:51 hmmmm       yep, so minetest.set_gen_notify()
00:52 hmmmm       takes a flags table of events to notify on
00:52 hmmmm       current flags being dungeon, temple, cave_begin, cave_end, large_cave_begin, large_cave_end
00:52 hmmmm       i think the idea was to have one associated with a specific schematic ID
00:52 hmmmm       but that never got implemented because i forgot about it
00:53 hmmmm       then, in your on_generated callback
00:54 hmmmm       foobar = minetest.get_mapgen_object("gennotify") -- at this point, foobar = {cave_begin = {x = blah, y = blah, z = blah}, cave_end = {x=400, y=-200, z=200}}
00:54 hmmmm       so on
00:54 hmmmm       so it'd have a table containing tables
00:55 hmmmm       table containing an array of tables that are the position vectors
00:55 jin_xi      or a list of positions where decos were placed, nice
00:55 hmmmm       right
00:55 hmmmm       that was the idea
00:55 hmmmm       the list of the decorations for a certain node placed
00:55 hmmmm       so that should be trivial to implement
00:56 hmmmm       add an option to register_decoration where you provide a gennotify "feature ID" and that'll be the key where things are stored in the table
00:56 hmmmm       so it'd be like
00:56 hmmmm       minetest.register_decoration{ blah blah blah blah,
00:57 hmmmm       gennotify_feature_id = "SomeShittyDecoration23"
00:57 hmmmm       })
00:57 jin_xi      or just gennotify flag decorations, and get a list of all decorations for a chunk? should be small enough
00:58 hmmmm       so then in your on_generated, foobar = minetest.get_mapgen_object("gennotify"); foobar = {SomeShittyDecoration23 = {{x=400, y=-200, z=200}, {x=410, y=-220, z=220}}
00:58 hmmmm       or so on
00:58 hmmmm       no no no.
00:58 hmmmm       no.
00:58 hmmmm       you can't control other mods that way
00:58 hmmmm       do it this way
07:39 Zeno`       are you awake ShadowNinja?
07:39 Zeno`       Thanks for the detailed explanation
07:40 Zeno`       I still don't like the idea of sandboxing  (not in a 0.4.x release anyway) but... *shrug*
08:59 kahrl       could someone who speaks French take a look at #1202?
09:03 kahrl       also I'm adding a Translation subsystem (with no subsystem maintainer at the moment)
09:04 kahrl       just so we can label these kind of PRs as "@ Translation"
09:04 kahrl       if anyone is against that subsystem please say so
09:17 kahrl       I'd like to resolve as many of the issues/PRs at https://github.com/minetest/minetest/labels/@%20Translation as possible before weblate goes back up, so comments on those are welcome
14:58 jin_xi      hmmmm: ok, so i tried adding the feature, but im not exactly sure what im doing.
15:00 hmmmm       well no no
15:00 hmmmm       you don't need to!  i will!
15:00 hmmmm       just not right now, I'm on work
15:01 hmmmm       i was just raising awareness that there is something already existing that you tried to do in your mod.  really I'd like to solve your voxelmanipulator issues
15:02 hmmmm       now yesterday you said that there were two voxelmanipulators but in your posting i only saw one where you scanned for nodes and then called remove_node() on it
15:04 jin_xi      see line 45 http://paste.ubuntu.com/8294886/ (same paste)
15:04 jin_xi      this calls the turtle system, which does a dry run to determine the size of the structure and then calls vm with that area
15:08 jin_xi      here is what i tried so far... http://paste.ubuntu.com/8300598/
15:43 hmmmm       jin_xi:  what does p[2].spawn(p[1]) do?  where is spawn()?
15:44 hmmmm       what you've done so far on implementing decoration stuff so far is mostly correct
15:47 jin_xi      http://paste.ubuntu.com/8300881/ http://paste.ubuntu.com/8300591/ its a cascade of functions named spawn... so the call p[2].spawn(p[1]) is basically tetris2.spawn(pos)
15:47 jin_xi      tetris2.spawn calls turtle.spawn which does the vm stuff
15:48 hmmmm       ahhh okay... so you build up a buffer table of things to write
15:48 hmmmm       you realize that voxelmanip is a buffer to the map
15:48 hmmmm       so you're writing to a buffer so that you can write to a buffer
15:49 hmmmm       you can change the table.insert things to vm_data[vm_area:index(rpos)] = materials[mat];
15:50 hmmmm       hrmm, the code looks solid enough.  you don't write to the mapgen object VM
15:50 hmmmm       however if another mod runs after your code uses a voxelmanip
15:50 hmmmm       that'd overwrite your changes
15:55 hmmmm       assuming turtle.spawn is called once, this code should work.
15:55 hmmmm       and assuming there are no mods that use voxelmanip that might be overwriting your changes
15:56 hmmmm       could i have a copy of your mod to debug this?
15:58 jin_xi      sure
16:36 hmmmm       you know, I don't usually recommend hacks, but here's a really simple way to prevent cavegen from shitting over your stone/sand/dirt structures:
16:36 hmmmm       build the structures with nodes identical to the original node you wish to use, except with is_ground_content = false
16:36 hmmmm       and then on dig, have them drop the regular, is_ground_content = true variety
16:37 VanessaE    I already suggested that to Sokomine weeks ago :P
16:37 hmmmm       as long as people don't go overboard it should be okay
16:37 hmmmm       this is good enough until we get hardware lighting in minetest and we're able to free up an extra 8 bits in a mapnode
16:38 hmmmm       VanessaE:  I think it may be worthwhile to add some sort of "inheritance" mechanism to nodedefs.
16:38 VanessaE    how would it work?
16:38 hmmmm       add a new api?
16:39 hmmmm       minetest.register_node_inherit("default:stone", {is_ground_content = false}) ?
16:39 VanessaE    I meant...
16:39 hmmmm       it would copy the nodedef from default:stone, and then override the settings with the custom ones you specify
16:39 hmmmm       the best part about what I'm suggesting is that in order to implement this feature, there are no additions to the core required.
16:39 VanessaE    the equivalent of doing this:  https://github.com/VanessaE/angled_walls/blob/master/init.lua#L108 --->  https://github.com/VanessaE/angled_walls/blob/master/init.lua#L145
16:40 VanessaE    but not in that sort of hacky way
16:40 VanessaE    ok so yeah, the same sorta thing.
16:40 hmmmm       yeah it seems you have the idea mostly
16:41 hmmmm       that's what I'd want to do, except all in one call and it'd be officially part of the API
16:41 VanessaE    right
16:41 hmmmm       does it make sense?
16:41 hmmmm       I realize that modders had probably thought of all these tricks a long time ago
16:41 VanessaE    it DOES make sense but I wonder if it's the right way to actually do it
16:41 hmmmm       but I'm just saying, we should actually do something about it rather than just let people do their own thing
16:41 VanessaE    *nod* I agree
16:42 hmmmm       VanessaE, there is no other way to do it
16:42 VanessaE    this clone method originally came from pilzadam, though what I do here required a recursive copy to avoid some hidden pass-by-reference fuckups I was running into.
16:42 VanessaE    but yeah, if there's no other way, then so be it
16:42 hmmmm       I mean you're probably thinking of not maintaining a copy of the node internally
16:43 hmmmm       but that won't make a difference because the vector containing the nodedefs isn't a list of pointers, it's a list of actual NodeDef structures
16:43 hmmmm       so you wouldn't be saving any space
16:44 VanessaE    right
16:44 VanessaE    well I got that ^^^^ code from the Roblox wiki but it's just a generic copy so feel free to use it :)
16:44 hmmmm       it'd also make access non-trivial.  you'd need to use getters/setters for all of the nodedef members
16:44 VanessaE    (line 11 in the file)
16:45 hmmmm       so, yeah, copying a node definition of something already registered then applying modifications, and registering the new one is the way to go.
16:45 hmmmm       i don't think it's hacky really.
16:49 VanessaE    I'll back it, since I've had to use this same construct more times than I care to remember ;-)  but you can expect others will bitch
17:58 diemartin   re: subclassing, it's trivial to do it with `setmetatable()` and `__index`
17:58 diemartin   (assuming register_xxx works by using `lua_getfield` instead of `lua_rawget`)
18:00 diemartin   just point `__index` to the "superclass", and explicitly set the overrides
18:22 ShadowNinja How should a Lua API for mkdir be implemented?  Add os.mkdir from ScriptApiUtil?  minetest.mkdir?  minetest.{fs,path}.mkdir?  Should we include something like LuaFileSystem?
18:22 VanessaE    minetest_fs.mkdir
18:22 VanessaE    strictly filtered to never overwrite and only to create within certain areas e.g. the world dir or mod dir
18:23 * VanessaE  looks at Luafs
18:23 kaeza       minetest.fs.mkdir doesn't follow the naming conventions; we don't see minetest.game.register_node or such
18:23 Krock       ShadowNinja, minetest.create_folder("foobar") ?
18:23 VanessaE    folder?
18:23 Calinou     it needs to work easily on all operating systems
18:23 VanessaE    no fucking way
18:24 Krock       VanessaE, Y no?
18:24 VanessaE    it's a G*d damned directory
18:24 VanessaE    that's why no.
18:24 kaeza       also, `listdir()` pls
18:24 ShadowNinja VanessaE: Nah, no new table.  Then we'd need core_fs too.
18:24 VanessaE    ShadowNinja: hrm.  well ok
18:24 VanessaE    core.mkdir or minetest.mkdir then
18:24 ShadowNinja kaeza: for x in require("lfs").dir(...) do...
18:25 VanessaE    either one is about the same
18:25 ShadowNinja VanessaE: (core == minetest) == true
18:25 VanessaE    looking at lfs, there's a lot of stuff here that some OS's might not allow
18:25 VanessaE    I wonder if lfs might be overkill.
18:26 VanessaE    ShadowNinja: case in point.  it's just a matter of which name looks better in the API doc.
18:27 VanessaE    I guess you could add in lfs and just disable the functions that we don't need.
18:27 VanessaE    e.g. symlinks, file locking a
18:27 VanessaE    -a
18:32 ShadowNinja Symlinks work on Windows, and file locking might be usefull.
18:34 VanessaE    mmmh
19:15 Columbus    Excuse me, a bit off topic question, about Graphics Processing
19:15 Columbus    A CPU uses Pipelining, so that multiple instructions at once are processed.
19:16 Columbus    Would a programm be able, to emulate such function,
19:16 Columbus    that a program pipelines the draw (and so on) process
19:16 Columbus    Could that work?
19:18 Columbus    When someone could, please give that idea to someone of OpenGL, because maybe, that it works, and it could create new technology
19:18 Columbus    Excuse me for my bad english, it isn't my mother-language
19:20 luizrpgluiz guys because minetest runs faster on Linux?
19:25 luizrpgluiz ?