Time  Nick       Message
08:42 Krock      Oh cool! 5.0.0 is out
08:42 Krock      thanks guys, finally we did it :D
08:44 nerzhul    lol Krock :D
09:08 rubenwardy Krock: 5.0.0 out, 5.0.1 soon:tm:
10:21 paramat    2 big bugs crawled out straight after release >_>
10:25 Krock      working on game#2329
10:25 ShadowBot  https://github.com/minetest/minetest_game/issues/2329 -- Node formspec screws up if node got destroyed
10:26 Krock      paramat: not our fault when they don't test it during this long period
10:29 paramat    lol
10:31 Krock      oh shit. The u16 limit is far too low #8325
10:31 ShadowBot  https://github.com/minetest/minetest/issues/8325 -- Large inventory lists are not sent correctly
10:40 paramat    that was the 1st big bug
10:40 paramat    2nd is #8300
10:40 ShadowBot  https://github.com/minetest/minetest/issues/8300 -- Segfault in noise.cpp
10:45 paramat    there's no quick and safe solution for 8300, so for 5.0.1 it looks like we may have to say 'use >1 emerge thread at your own risk'
11:53 paramat    see https://github.com/minetest/minetest/issues/8300#issuecomment-471170530 , please can a core dev merge those 2 commits to 5.0.1? then i'll update docs with a warning. i think 5.0.1 should be released ASAP and this will attend to 8300
12:27 nerzhul    Krock yeah
12:27 nerzhul    i pushed a PR to unify our string serialization to ensure we always have u32 limits (with a real limit at 64MB like longString) but it should be evaluated for 5.1
13:49 Krock      nerzhul: and how do you want to keep the compatibility in #8330?
13:49 ShadowBot  https://github.com/minetest/minetest/issues/8330 -- Fix definitively string serialization coding problems by nerzhul
13:50 Krock      I mean 5.x compat
13:50 Krock      your changes would require another hard protocol bump
14:45 Krock      seriously
14:50 * Krock    tries to figure out why there's no deprecation entry for get_perlin_map
14:54 Krock      apparently the function name is not read correctly
15:05 Krock      what the heck is wrong with the Lua debug interface? on stack level 0 it returns the local function name
15:06 Krock      instead of the C++ executed function
15:07 nerzhul    Krock yes it requires it, and it's why i wanted to do it just after release to prevent anoying problems in a 5.1
15:07 nerzhul    i proposed that solution day after release. to prevent fragmentation. Now it's too late, see you in 2 years :)
15:16 Krock      :/
18:23 nerzhul    i can add a compat mode if needed
18:23 nerzhul    but it will less proper than my pr, but possible
18:29 Krock      simple fix: #8246
18:29 ShadowBot  https://github.com/minetest/minetest/issues/8246 -- Correct the checkbox selection box position by SmallJoker
18:30 p_gimeno   \o/
18:49 kaeza      Krock: >[...] level 0 is the current function (getinfo itself); level 1 is the function that called getinfo; and so on.
18:49 kaeza      https://www.lua.org/manual/5.1/manual.html#5.9
18:51 Krock      kaeza: get2dMap_flat gets called from here: https://github.com/minetest-mods/mg/blob/master/init.lua#L380-L390 so what would the function name then be at level 0?
18:53 kaeza      At level 0 it's always getinfo, level 1
18:54 kaeza      ...
18:54 kaeza      Where's the getinfo call?
18:55 p_gimeno   I think it's because that's a tail call
18:55 Krock      kaeza: C++. https://github.com/minetest/minetest/blob/master/src/script/lua_api/l_base.cpp#L101-L103
18:55 Krock      this must get the currently called function name
18:55 Krock      which is in this case get2dMap_flat, but Lua decided to return get_perlin_map (local function) instead
18:56 p_gimeno   yes, because of the tail call
18:56 p_gimeno   if you change get_perlin_map and do: local result = pm:get2dMap_flat({x = minp.x, y = minp.z, z = 0}); return result
18:56 p_gimeno   then it will report the correct function
18:57 Krock      eeks
18:57 p_gimeno   tail call optimization causes quite some headaches
18:57 Krock      yeah, you're right. any hope to work around that?
18:58 Krock      I doubt that a call history exists
18:58 p_gimeno   I thought it actually replaced the function name by the called function, are you on LuaJIT?
19:01 p_gimeno   https://paste.scratchbook.ch/view/f9617905
19:02 Krock      LuaJIT here, yes.
19:02 p_gimeno   I was wrong, it doesn't replace the caller
19:06 p_gimeno   interesting, Lua 5.3 does report the right function
19:12 p_gimeno   http://lua-users.org/lists/lua-l/2014-02/msg00475.html also worked for me
19:13 p_gimeno   the notail() thing, I mean
19:15 p_gimeno   both lua and luajit report f1 and f2 in the stack trace, and skip notail
19:16 p_gimeno   https://paste.scratchbook.ch/view/b1f291fd
19:45 Krock      p_gimeno: thank you for your research, so it seems to be a 5.1 issue
19:46 p_gimeno   a <5.3 issue; 5.2 has the same problem
20:01 nerzhul    it's the problem with the lua state with a old 5.1 version, some issues exist and will never be fixed :(
20:01 sofar      wireshark went to lua-5.2
20:02 nerzhul    our problem is we support jit
20:11 Krock      well, because it's faaast
23:19 kaeza      Just wondering, why was the current translation file format chosen? There are many mods out there using intllib that migrated to gettext .po file format, and minetest could take advantage of that.
23:30 kaeza      [20:29]  <kaeza> The issue is that client-side translations use a format like the old .txt files, so it's trivial to convert, but gettext .po files have better tooling, so I don't know what to do.
23:31 p_gimeno   there's been some work towards implementing a translation toolchain
23:31 p_gimeno   nore and I were working on that
23:35 p_gimeno   there's also #7290
23:35 ShadowBot  https://github.com/minetest/minetest/issues/7290 -- Switch to PO file format for client-side translations
23:37 p_gimeno   #6325 is the one we were working on
23:37 ShadowBot  https://github.com/minetest/minetest/issues/6325 -- Add scripts to find and update translations. by Ekdohibs
23:43 kaeza      p_gimeno: I see. Thanks for the links.