Time  Nick       Message
03:28 MTDiscord  <s​rinivas> heh, how did that number ome up?
11:26 bebebeko   Hey folks! tldr for my project i need some features in program, and i'm ready to sponsored these through bountysource, for example, https://github.com/minetest/minetest/issues/10593 At this moment the description looks ogly in my opinion. Could anyone suggest a way how i can improve it, or it's fine?
11:29 bebebeko   this is my first experience with bountrysource, please don't judge me harshly ;-)
11:32 pmp-p      hi bebebeko, there's not much to judge, as description of what you need is terribly incomplete is it client/server side ,static or dynamic (moving entities) ?
12:29 bebebeko   pmp-p: thank! I updated the issue, could you please verify, maybe I miss smth? https://github.com/minetest/minetest/issues/10593
16:13 Krock      > Then when the mods have loaded and the server is stuck, interrupt with Control+C and enter thread apply all bt.
16:13 Krock      cool trick. I haven't used that in a while :D
16:20 Krock      when is lua_isnil() not the same as lua_isnoneornil ?
16:21 rubenwardy when it is none
16:21 rubenwardy none means that the index doesn't exist on the stack
16:21 rubenwardy nil means that the index exists and is nil
16:21 rubenwardy in     func(1, nil),   index 2 is nil but index 3 is none
16:22 rubenwardy lua_isnoneornil should pretty much always be used
16:22 rubenwardy over isnone or isnil
16:22 Krock      lua_isnil ensures that the arguments are provided, even if nil
16:22 Krock      so in terms of solid API that's a safer approach
16:23 Krock      perhaps
16:23 Krock      after all it might not matter because mostly the code would attempt converting this non-nil value to a number, which return 0
16:24 rubenwardy it makes the C++ API act differently to a Lua one, and you'd want to throw rather than treating none as a non-nil
16:24 Krock      that's what luaL_checknumber is for, no?
16:25 rubenwardy yes, which is better to use over lua_isnone or lua_isnoneornil
16:25 Krock      thanks for the explanation
16:25 rubenwardy my point is if an argument is nilable, none should be treated like nil. And if it's not nilable, you can use check*
16:26 Krock      so it's a question whether the arguments may be optional or not
16:26 Krock      and whether they may be nilable
16:54 celeron55  i think there's a semantic difference between an optional argument and there being an option to pass something as nil
16:54 sfan5      rubenwardy: (re 10368) how about ensuring that either the formspec version is bumped or the protocol version?
16:55 rubenwardy it's weird from a Lua point of view though
16:55 sfan5      what I'm thinking of is that it's not necessary to bump the protocol version if the formspec version has been bumped for this version
16:55 rubenwardy the only way to tell from Lua is to do   function(...)  print(#{...})  end
16:55 celeron55  of course you can define a style where one or the other is not allowed
16:55 celeron55  and that's your convention then
16:56 celeron55  is it?
16:57 celeron55  in that case maybe it's better to not distinguish between them in C++ either
16:57 Krock      set_string("test") is less obvious than set_string("test", nil)
16:58 Krock      sometimes you want to enforce an argument to be provided
16:58 celeron55  i don't care what's used as long as it's used consistently
17:23 nerzhul    is this me or celeron55 is now active as a coredev ? :D
17:33 rubenwardy I'll believe it when I see a PR for VAEs
17:33 rubenwardy out of the blue
17:33 rubenwardy tbh, that probably isn't the highest priority rn :D
17:42 celeron55  it would be so hacky you wouldn't accept it, trust me
17:43 rubenwardy :D
17:50 zughy[m]   so wait, should I address Krock's reviews or not? About #10564
17:50 ShadowBot  https://github.com/minetest/minetest/issues/10564 -- Fix ObjectRef crashes due to lua_isnil() by Zughy
18:02 Krock      zughy[m]: perhaps note what was the idea of doing these modifications? perhaps I missed an important part that's not mentioned
18:56 zughy[m]   Krock: ruben suggesting to change everything with isnoneornil supported by Desour https://github.com/minetest/minetest/pull/10564#issuecomment-717116073
19:23 Krock      apparently I'm blind. somehow I only read the first few posts. it's a design decision whether lack of arguments is counted as "nil" too
19:24 Krock      my opinion is towards lua_isnil, also because a large part of the code uses that check. if other developers decide to go with isnoneornil, then please adjust the documentation accordingly
19:32 zughy[m]   yes, but they prevented crashes declaring the variable and its value first (i.e. int x = 0) and then checking if the index in the stack was nil. If it wasn't, they override x
19:33 zughy[m]   imho, using is_noneornil makes instantly explicit to whoever wants to contribute that none and nil are not the same thing in lua
19:36 zughy[m]   also, an exception are tables: when getting the field, lua_isnil works perfectly to check whether it exists
19:37 zughy[m]   that's why I didn't changed the whole set_sky/moon/etc. checks
19:37 zughy[m]   *change
19:40 sfan5      functions that have optional arguments should use lua_gettop or lua_isnoneornil
19:41 sfan5      functions that don't have that shouldn't be using lua_isnil, yet it seems many do
19:41 sfan5      since this is kind of a mess I suggest doing the changes needed to restore compatibility and leaving the rest alone
19:45 sfan5      in fact it almost seems like a good idea to revert the cleanup since who knows how much time has been wasted discussing and doing things that worked before and didn't bother anyone
19:45 sfan5      (don't get me wrong I'm not actually in favor of a revert but I am annoyed at all the time and discussion that is now spent on this)
19:49 zughy[m]   (sad thing: the bridge is slow, I'm looking at your replies from the IRC log)
19:49 zughy[m]   I don't agree with the "it works, don't touch it". That code was a mess, like, a huge one. You may understand it as you've been cruising through it for years, but as a newcomer it's just a headache
19:49 zughy[m]   also: according to the docs, set_animation_framespeed should be 15.0 by default, but from l_object.cpp that's now what it happens. It pushes the value and a true value when there is a number and a false one where is none. Is that... normal?
19:50 sfan5      lots of methods have an undocument boolean return value that indicates success
19:51 sfan5      and regarding the default: the docs are lying and there is no such default
19:52 sfan5      ..perhaps that's supposed to be the default if not set, but the method argument is definitely mandatory
19:53 rubenwardy yeah, that's the default if you don't call `set_animation_framespeed`
20:02 pgimeno    <rubenwardy> the only way to tell from Lua is to do   function(...)  print(#{...})  end   <-- nooo! # is not guaranteed to work with embedded nil values. The way to do with lua is using select('#', ...)
20:09 zughy[m]   ok, I'll fix it later
20:12 pgimeno    I think the right thing to do in general, is to use isnoneornil when there's no default value for the parameter but nil is a valid value, and isnil when there's a default and it is something other than nil.
20:14 rubenwardy the right thing to do is not care about nil vs none in the C++ api
20:15 rubenwardy because Lua itself rarely cares about nil vs none
20:17 pgimeno    sorry, my experience is with the Löve API, e.g. https://love2d.org/wiki/love.graphics.draw - values in parentheses are defaults
20:22 pgimeno    it is rare that nil values are accepted as distinct parameters, actually
20:23 pgimeno    here's a somewhat delicate case: VoxelManip:write_to_map(bool), defaults to true - what should nil do?
20:25 pgimeno    IMO VoxelManip:write_to_map(nil) should be treated as false, distinctly from VoxelManip:write_to_map()
20:47 zughy[m]   PR is ready if you want Krock
20:48 red-001    btw if you want me to I can change the function to pascalCase
20:48 red-001    camelCase*
20:48 red-001    for #10589
20:48 ShadowBot  https://github.com/minetest/minetest/issues/10589 -- Block attempts to connect to the client. by red-001
20:49 Krock      will merge #10589 in 10 minutes
20:49 ShadowBot  https://github.com/minetest/minetest/issues/10589 -- Block attempts to connect to the client. by red-001
20:50 Krock      zughy[m]: looks good, didn't test yet
20:51 Krock      strictly spoken the lua_api.txt would now need an update to properly state optional arguments, though.
20:52 Krock      will test tomorrow
21:20 Krock      merging...
22:07 nore       hmmm, I can't find the place where we said why LuaJIT was dangerous to use in (SS)CSM?
22:08 nore       also, does anyone want to review #10451? (at least reviewing the code, I'll work on testing a bit more)
22:08 ShadowBot  https://github.com/minetest/minetest/issues/10451 -- Client-side translations: gettext support by Ekdohibs
22:10 nore       btw, another question about this PR: it does require a translation update in all languages if we want to get plurals working, maybe weblate changes should be merged? or the PR merged, weblate updated, and then getting all translations from weblate after I update them all with the translation of the plurals?
22:28 red-001    I think it's just a generic concern about JIT implementations generating code and being more complex -> more edge cases that code result in exploitable bugs
22:29 red-001    s/code/could