Time  Nick       Message
07:03 nerzhul    sfan5: fix here: https://github.com/minetest/minetest/pull/11245/files
07:03 nerzhul    exe_virus, if you are affected try my pr
07:03 nerzhul    it should do the trick
07:04 nerzhul    i moved the function outside of the client, i forgot client is big but not enough to cover the main menu :)
07:05 nerzhul    i had to tidy a bit the lua stack in order to store the filesys pointer properly, but it's fine
09:03 sfan5      "leaking" the IFileSystem into the scripting doesn't look like a good solution
09:03 sfan5      if the gui engine has access to the irrlicht stuff, can't a wrapper method be added there?
09:40 nerzhul    i can put the IFileSystem inside it if you prefer
10:41 sfan5      GUIEngine has a m_rendering_engine, so it can wrap extractZip for use from the scripting
10:43 sfan5      ah that's not even needed, ModApiMainMenu can access private members of GUIEngine
10:44 sfan5      so you can do this: lua_pushboolean(L, fs::extractZipFile(getGuiEngine(L)->m_rendering_engine->get_filesystem(), zipfile, destination));
12:47 Guest76367 What is v3s16 and similar types?
13:06 MTDiscord  <josiah_wi> They are 3d vectors with specifically sized fields. For example v3s16 is a vector of 16 bit integers I think.
13:07 MTDiscord  <josiah_wi> Does that answer your question?
13:28 nerzhul    sfan5, it's exact, a bit ugly but exact, i can use that :)
13:34 nerzhul    sfan5, PR is fixed, and far more tiny now :)
13:45 Guest76367 Thanks, josian_wi.
14:01 nerzhul    who are you Guest76367 ? :p
14:01 nerzhul    merging #11245, and mybad about the mistake :)
14:01 ShadowBot  https://github.com/minetest/minetest/issues/11245 -- fix: extractZipFile is not part of Client but more generic. by nerzhul
17:09 Krock      Will merge #11158 and #11039 in 15 minutes
17:09 ShadowBot  https://github.com/minetest/minetest/issues/11158 -- Improve liquid documentation by Wuzzy2
17:09 ShadowBot  https://github.com/minetest/minetest/issues/11039 -- Add metatables to lua vectors by Desour
17:10 sfan5      you can merge #11246 too
17:10 ShadowBot  https://github.com/minetest/minetest/issues/11246 -- Fix build for newer versions of GCC by lhofhansl
17:10 Krock      +1
17:11 pgimeno    Krock: can you hold 11039 so I can take a look?
17:11 Krock      no problem
17:12 pgimeno    thanks, looking
17:12 Krock      another PR which would be rather easy to review: #11031 (to anyone)
17:12 ShadowBot  https://github.com/minetest/minetest/issues/11031 -- Fix misleading builtin command syntax of /shutdown by Wuzzy2
17:18 pgimeno    +	local v = setmetatable({x = x, y = y, z = z}, metatable)
17:18 pgimeno    +	return v
17:19 pgimeno    why not return setmetatable, which would be a tail call?
17:20 sfan5      if you have a diff I can push it into the PR before we merge it
17:23 Krock      merging (2)
17:24 Krock      done
17:25 pgimeno    http://www.formauri.es/personal/pgimeno/pastes/fast_new-tail-call.patch
17:29 pgimeno    +	-- scalar/vector makes no sense
17:30 pgimeno    and vector/vector?
17:30 MTDiscord  <appguru> The local was apparently faster IIRC
17:30 MTDiscord  <appguru> Kinda makes sense as the immediate return would have to take multiple return values of setmetatable into account
17:31 pgimeno    no it doesn't make sense, tail calls are transformed into jumps
17:31 pgimeno    the latest function in the chain would return the value directly to the first caller
17:33 pgimeno    that, or I'm not following your logic
17:36 pgimeno    vector.divide can divide two vectors, but metatable.__div can't - isn't that inconsistent?
17:36 rubenwardy setmetatable only returns one thing, so multiple returns is irrelevant
17:37 pgimeno    not that there's a lot of use for dividing two vectors, but if vector.divide supports them, so should __div IMO
17:38 MTDiscord  <appguru> rubenwardy: multiple returns are relevant, as Lua(JIT) can't know that, unless setmetatable is localized
17:38 pgimeno    actually, same applies to __mul
17:38 rubenwardy it's not relevant as it has the same result but is faster
17:38 rubenwardy even PUC Lua supports tail calls
17:38 MTDiscord  <appguru> are you actually sure that it is faster?
17:39 rubenwardy tail calls are well documented to be faster
17:39 MTDiscord  <appguru> but multiple returns are slower
17:39 rubenwardy it may only be nonneglibly so when doing actual recursion
17:39 rubenwardy it's not multiple returns, it's a single return
17:39 MTDiscord  <appguru> multiple return values, to be precise
17:39 MTDiscord  <appguru> evil me could be doing _G["set" .. "metatable"] = function(...) return ... end
17:40 MTDiscord  <appguru> Lua has to account for that case
17:40 pgimeno    hm, maybe appguru has a point
17:40 pgimeno    http://wiki.luajit.org/NYI
17:40 pgimeno    CALLT 	partial 	Tailcall. Some tailcalls to frames lower than the starting frame of the trace are not compiled.
17:40 MTDiscord  <appguru> https://github.com/minetest/minetest/pull/11039#discussion_r592596882
17:41 MTDiscord  <appguru> Should probably be tested
17:41 pgimeno    if the tail call is going to break traces, that makes it inconvenient no doubt
17:41 pgimeno    I am not sure what it means when it says "frames lower than the starting frame" though
17:57 pgimeno    no other objections from me