Time Nick Message 10:57 MTDiscord done (ps. i'll probably postpone a re-review of 14343 till 5.9 ships, since we've already entered feature freeze) 13:21 sfan5 merging #14744, #14753 soon-ish 13:21 ShadowBot https://github.com/minetest/minetest/issues/14744 -- Move malloc_trim invocations to background thread by sfan5 13:21 ShadowBot https://github.com/minetest/minetest/issues/14753 -- Disallow formspec debug if the player does not have the debug privilege by l-koehler 13:32 grorp regarding #14545: I'd like to comment and explain our use case on https://github.com/libsdl-org/SDL/issues/3559: https://gist.github.com/grorp/c67b67f10d1a61280ff8be10676cfb05. Do you have anything to add or correct? cc @y5nw 13:32 ShadowBot https://github.com/minetest/minetest/issues/14545 -- SDL: Some keybinds broken due to missing character lookup 13:33 MTDiscord let's try this again, Discord. @y5nw 14:01 MTDiscord grorp: Text input events are simply not appropriate for getting individual keypresses since the text sent by SDL_TextInputEvent is the result of a series of (i.e. potentially multiple) keypresses processed by the IME (if one is present), in which case the resulting text may have little correspondence with the keys pressed by the user. (That said, I don't think the SDL devs really need this explained to them, considering that text input 14:01 MTDiscord handling is partly what they work on.) 15:27 MTDiscord Ehh... very pity. It is almost (or already entirely) ready and very close to merging in 5.9 15:31 MTDiscord Could it be moved to 5.10 milestone then? 15:36 MTDiscord If it is postponed for the next release cycle, then in this period I think it is worth to me to separate the new vertex format adding from this PR since it is not concerned to that directly and was rather a "semi 15:36 celeron55_ grorp: that post proposal looks good 15:37 celeron55_ i have the same issue also on my finnish keyboard. the chat should launch when typing / which is typed as shift+7, just like ona german keyboard 15:37 celeron55_ on a* 15:38 MTDiscord If it is postponed for the next release cycle, then in this period I think it is worth to me to separate the new vertex format adding from this PR since it is not concerned to that directly and was rather a "half-measure" and not very good workaround. I think it needs in the new full-fledged format specifically for two colors and also it should work for the vertex lighting when the shaders are disabled 15:39 celeron55_ there's another way to word the requiremnet from SDL: it should be possible to get text and key events simultaneously, in such a way that for each text event it is known which key events caused the text event so that those key events can be ignored. but in practice you have to "compress" them just like irrlicht does, as you can only ignore the last key event anyway as you didn't know the first 15:39 celeron55_ ones were about to generate the text you were looking for 15:39 celeron55_ requirement* (i can't type today) 15:41 celeron55_ it's possible this will have to be worked around though. it's not unconceivable to generally implement mapping of key combinations and then you just manually map shift+7 15:41 celeron55_ and the SDL people might require that from us 15:41 celeron55_ the problem is, it's impossible to ship the default config to all possible keyboard layouts 15:41 celeron55_ compared to irrlicht, where it is trivial to ship it 17:18 grorp celeron55: "compared to irrlicht, where it is trivial to ship it": huh, why? 17:19 grorp anyway, thanks, submitting the thing now. 17:37 nekobit rubenwardy: i hope i can chat here for my contribution 17:37 nekobit where should i rewrite the pkgmgr stuff at? src/network? 17:37 nekobit and is there a CURL multihandler somewhere I can use? 17:38 rubenwardy src/content 17:38 rubenwardy And I'd want to use this as an opportunity to unify the implementations and unit test 17:40 nekobit wdym, want me to write a unit test? 17:40 nekobit is there a file that specifies coding conventions and all that 17:41 nekobit Ah, its on the wiki 17:41 nekobit brb, got a job interview, then ill work on this 17:44 nekobit >avoid operator overloading like the plague 17:44 nekobit THANK you. 17:47 rubenwardy It's very useful for things like Vector3 and math classes. But most of the time should be avoided 17:48 nekobit Ive seen codebases that use it in awful ways 19:15 MTDiscord @andrey2470t sure, I've added it to the 5.10 milestone. 19:31 MTDiscord Ok, thanks 20:25 sfan5 huh? are we rewriting the pkgmgr into c++? 20:29 sfan5 celeron55_: getting text events and key events at the same time disagrees with the principle of IMEs 20:30 sfan5 but getting the emitted character along with the scancode is not an unreasonable ask 20:31 rubenwardy nekobit: #12295 20:31 ShadowBot https://github.com/minetest/minetest/issues/12295 -- Add content/pkgmgr unit tests 20:32 nekobit sfan5: I was told to do so before adding cli support for downloading 20:33 rubenwardy technically you could load up a Lua environment for the cli 20:33 nekobit Ofc, but would it be nice to get that out of way? 20:33 sfan5 not only technically. this would be the most straightforward solution 20:33 rubenwardy there's currently two different content implementations, one in src/content and the other in builtin/mainmenu/content/pkgmgr.lua. One of my longer goals was to merge these and unit test it 20:34 nekobit sfan5: well, i was told to rewrite it in C++ 20:34 sfan5 rubenwardy: ah I see 20:34 nekobit but i agree just loading in Lua would be easier (and a bit safer) 20:35 nekobit ask ruben 20:35 rubenwardy I said that it's in Lua and tied to the mainmenu, the later is more the issue than the former 20:36 nekobit was there any advantage in rewriting it in C++ though? besides easier access? 20:37 nekobit I'll do it, but was just curious 20:41 nekobit Anyway, is there a curl multi handler anyway i can reuse, or must i define a new one? 21:02 nekobit rubenwardy: 21:02 rubenwardy I don't even know what that is 21:02 rubenwardy I suggest searching for curl mentions in the code base 21:02 nekobit you dont know what it is, yet you want me to write it in C++, right? 21:03 rubenwardy there's an async job manager for curl requests 21:03 nekobit Ah 21:03 nekobit that probably uses curl multihandler under the hood then 21:03 nekobit also 21:03 rubenwardy httpfetch.h 21:04 nekobit > was there any advantage in rewriting it in C++ though? besides easier access? 21:04 nekobit Ta 21:05 rubenwardy the idea was to merge the implementations and not require a lua env for the cli, but I never really looked into it propery as it's still far off on my to do list 21:07 nekobit Sure, i just dont think setting up a lua environment for the cli would be difficult 21:07 nekobit but id still need to merge those 2 files like you said anyway 21:07 nekobit might as well make it c++ 21:14 nekobit btw if you dont already use a curl multi handler, thats probably how youll accomplish async the best 21:16 [MTMatrix] pretty sure Minetest does 21:16 [MTMatrix] wow wow wow: https://github.com/libsdl-org/SDL/issues/3559#issuecomment-2174419503 21:19 nekobit grorp, you are correct, it already does 21:20 nekobit im curious as to what protocol this game uses for networking 21:20 nekobit hopefully not json 21:22 [MTMatrix] s/game/engine/ 21:22 [MTMatrix] something custom on top of UDP 21:22 nekobit Nice 21:24 [MTMatrix] re SDL again: I just hope they don't break something else this way. but they're competent people after all. 21:29 nekobit rubenwardy: Why is HTTPFetch not a class? 21:29 nekobit httpfetch_init(...), httpfetch_caller(free)... 21:29 nekobit is this intentional? 21:30 nekobit *httpfetch_caller_free(...) 21:44 [MTMatrix] Please avoid flooding, this is a pretty low traffic room. Less and longer messages would be great 21:45 nekobit I apologize. 21:49 nekobit Why is fs used instead of std::filesystem? 21:50 ROllerozxa minetest hasn't used C++17 since rather recently 21:50 rubenwardy yeah, Minetest predates C++17 21:50 ROllerozxa I assume refactoring to use std:: filesystem would be welcome 21:50 nekobit I'll just use FS for now unless anyone objects. 21:51 nekobit Refactor could be in another PR 22:03 nekobit Not sure why pkgmgr.get_folder_type is just returning path. This doesn't make much sense. Should I still do this? https://github.com/minetest/minetest/blob/master/builtin/mainmenu/content/pkgmgr.lua#L221 22:28 nekobit hmm... how do I map a C++ function to a lua one? The Lua docs site is down.