Time Nick Message 08:14 MTDiscord My phone CPU got 32-bits and the device is from 2018 so not that old 08:15 MTDiscord There is a super big difference between the specs of a low-end budget mobile device and a high-end gaming pc and I hope MT will continue to support both for the foreseeable future 08:22 MTDiscord it's better to take everyone with you rather than drop others for the stronger ones. 08:34 MTDiscord > most 08:35 MTDiscord but i get your point 08:35 MTDiscord I like MT approach of supporting older hardware: Keep default values generally low and make it possible to increase them to the absolute max for the gaming rigs (this is also my last message here cause it's off-topic now) 11:58 pgimeno on yesterday's discussion about extending the content_id space, I haven't seen a suggestion to use 24 bit node ids. I think that's already reasonably large, and leaves room for one more byte of info, making it possible to use e.g. rgb888 for lighting, or a new param3 12:01 sfan5 that's potentially much slower since you have to do bit operations for every access 12:02 pgimeno a masking on read is cheap, and you don't write node ids that often, plus if lighting is fixed, you don't write light info that often either 12:04 pgimeno bit fields are supported by C++ directly 13:18 MTDiscord I still find the best solution is instead of us using defaults sizes for node ID's like 16 bit, 24 bit, etc, we should just stop with 16 bit as the default, and make it extendable via lua and client side mods when we get those. 14:02 MTDiscord Isn't 16 bit the default? 14:02 MTDiscord Or atleast the hard coded limit 14:12 MTDiscord I noticed that the CMake build doesn't handle ENABLESYSTEM options uniformly. GMP switches to the bundled lib if the search fails, but Jsoncpp will throw a fatal error, and only uses the bundled version if the option is set to OFF. Which way should be preferred? 14:33 sfan5 I think that the options should not throw fatal errors and that their default should be ON 14:33 sfan5 question would be: how does that compare to the behaviour of ENABLE_LUAJIT 14:53 nerzhul pgimeno, 24 bits is hard for a CPU to use, if you ask for a change, just use a 32bit one, it's at least correctly handled by CPU buses 14:53 MTDiscord I've looked at it and it is yet a third design. It has 2 options, ENABLE_LUAJIT behaves like ENABLE_SYSTEM_GMT unless REQUIRE_LUAJIT is set and then it behaves like ENABLE_SYSTEM_JSONCPP. 14:56 MTDiscord Also it's not as simple as setting all of them to ON by default, because the System JSON option is OFF by default because of potential segfaults etc., according to the documentation on the setting. 14:59 sfan5 yeah that documentation is outdated 14:59 MTDiscord Should be fixed then... 19:26 Krock will merge #10952 in 15 minutes 19:26 ShadowBot https://github.com/minetest/minetest/issues/10952 -- Server-side authority for attached players by SmallJoker 19:40 Krock merging 20:43 sfan5 second release candidate now? 20:48 sfan5 merging game#2824 in 5m 20:48 ShadowBot https://github.com/minetest/minetest_game/issues/2824 -- Update Spanish translation by NathanielFreeman