Time Nick Message 21:02 sfan5 merging #11963 #12020 in 10m 21:02 ShadowBot https://github.com/minetest/minetest/issues/11963 -- Make get_sky return a table by Zughy 21:02 ShadowBot https://github.com/minetest/minetest/issues/12020 -- Debug HUD flag by appgurueu 21:06 erlehmann i still think that zughy's transition plan for a new API makes very little sense 21:06 erlehmann because it is a two-stage thing: first you add an argument to get the table, then at a future point, you deprecate the same argument? 21:07 erlehmann or something like that 21:07 erlehmann i don't believe it has been thought out beyond “it should change behaviour, but lets preserve the old one for now” 21:08 erlehmann and this is not about future rug-pulls, i just think it's weird, because you could keep the old way probably at little cost (judging from the code), not deprecating anything. or you deprecate it and ALWAYS need the argument. 21:09 erlehmann (i.e. deprecating not needing it) 21:09 sfan5 thing is: we don't want to keep the old way 21:10 erlehmann yeah, but that means that first you deprecate not having the argument and then at some later point you deprecate the argument 21:10 erlehmann unless i am misunderstanding something major about the ”at some later point” 21:11 erlehmann i mean it seems fine if you always need the argument in the future 21:11 erlehmann because *then* you can definitely detect code that is deprecated 21:12 erlehmann (and automatically offer advice) 21:13 erlehmann (“fine” as in: among the rug-pulls it's the least bad ones) 21:14 MTDiscord Perhaps in game development, supporting old versions to no limit can lead to extreme bloat 21:14 sfan5 after the deprecation period we don't want to detect code that is deprecated 21:17 erlehmann i guess then to you my “deprecation could/should maybe done in a different way” is no different than my other “if engine devs put more work into their API design, game developers will have less changes” arguments, which makes it kinda useless to complain (as we wall know where we stand on this) 21:18 erlehmann jordan4ibanez just to be clear, i am not arguing for supporting old versions forever here. i am arguing that it's at least ”a bit weird” to change a function signature twice to basically achieve one function signature change. 21:18 erlehmann especially since the end state will be indistinguishable from the start state 21:19 erlehmann (from a function signature POV) 21:21 erlehmann regardless, i see no reason to rehash this if “deprecation should be done different” is structurally too similar to “nothing should be deprecated”. deprecation will probably proceed exactly as zughy outlined it and that's it. 21:21 MTDiscord Oh sorry 21:21 erlehmann no need to be sorry 21:23 erlehmann i just wanted to point out that this time i am only criticizing the way *how* it is done, not *that* it is done. 21:23 erlehmann (and i do not have a better proposal on hand that is acceptable to engine devs) 21:24 sfan5 the "better" proposal is to make a get_sky2() but I am sceptical 21:24 erlehmann it would definitely lead to a much easier deprecation detection 21:24 erlehmann i.e. contentdb could detect deprecated mods for example 21:25 erlehmann or you could just grep your whole codebase 21:26 erlehmann but as i said, we had these arguments already. i am interested in new objections to that kind of deprecation though!