Time Nick Message 09:34 Baytuch Good evening 09:55 erlehmann does anyone of you who was there in 2011 remember why some commit messages occur twice in the git history? 13:34 proller https://github.com/minetest/minetest/pull/11843 13:34 proller https://github.com/minetest/minetest/pull/11910 13:35 proller 2 days afrer release.. 13:37 rubenwardy proller: those PRs are on the meeting agenda for the next core dev meeting 13:43 erlehmann proller so how are you getting around floatland problems? 13:43 erlehmann proller i mean having floating point inaccuracies far from zero zero 13:43 erlehmann just double? 13:44 MTDiscord has minetest 5.5.0 been pushed to google play yet? 13:45 proller erlehmann, first simplest solution is double, yes. 13:45 proller in future need to split all calculations to base int32+ float offset 13:46 erlehmann proller does your approach finally introduce proper overflow checks instead of the MAX_MAP_GENERATION_LIMIT historical baggage or are the overflows just, uh, larger? 13:46 proller rubenwardy, when it will be? 13:46 rubenwardy no idea 13:46 erlehmann proller that future sounds nice! 13:46 erlehmann proller i wish you great success 13:47 proller erlehmann, i cant do something before this prs will be merged 13:49 proller 2 months already spent for nothing.. 13:49 rubenwardy you should have opened or commented in an issue first 13:50 rubenwardy or on IRC 13:53 proller to spend another half year? 14:15 MTDiscord "in future need to split all calculations to base int32+ float offset" - doesn't sound reasonable to me 14:16 MTDiscord doubles are literally this: extending the mantissa by 32 bits 14:19 proller maybe doubles will be ok, but with second pr still something wrong with movement on huge coords 14:48 MTDiscord sfan5, I implemented my preferred solution to #12026. I'd like support of the solution so I can PR it, or if you don't support it then I'd like to know what your preference is. 14:48 ShadowBot https://github.com/minetest/minetest/issues/12026 -- Throw fatal error if enabled database backends are missing 18:41 Pexin luatic: I interpreted proller's statement more along the concept, int32 to get the mapblock (or something), then float offset from that reference 18:42 erlehmann does anyone have an opinion on this? https://github.com/minetest/minetest/pull/11978 i need test nodes for various texture formats (as in: “how does this look in-game” + a proven combinaton of feature flags) often enough that if it is not part of devtest, i'd make a separate mod probably. or does something like devtest already exist on cdb? 18:43 erlehmann i mean anyone who has not yet shared their opinion 18:43 erlehmann also where do people *usually* put the stuff that is not accepted into devtest? 18:50 Pexin as I think about it, the opposite (float for crude coord, int offset from there) makes more sense, IF you can ensure you are using float numbers whose precision you can calculate. eventually, far enough away from origin, ints (any width) will become insufficient to offset within the imprecision range, but it would be enormous.