Time Nick Message 00:17 MTDiscord Why can't I get more than 20 FPS on an Inspiron that's probably newer, I wonder. 00:22 erlehmann josiah_wi ofc i have tweaked settings 00:28 MTDiscord I'd be interested in which settings to tweak. 00:29 MTDiscord minetest-irc channel though 00:58 erlehmann there exist a bunch of unit tests that seem to run into errors and still claim to pass 00:59 erlehmann like take test_index_v3s16_all_pos 00:59 erlehmann ubsan says src/voxel.h:270:51: runtime error: signed integer overflow: -20442400 * -902 cannot be represented in type 'int' 01:00 erlehmann i don't think the test should pass under the circumstances 01:00 erlehmann even if the asserts are ok 01:00 erlehmann the network null pointer thing also shows up if you run the unit tests with ubsan 01:36 erlehmann Krock are my recent attempts to break devtest “actual game issues” for you? 02:21 erlehmann is the submitter of this seriously mistaken about on_drop handling or am i? https://github.com/minetest/minetest/issues/11818 10:30 erlehmann i am extremely frustrated with the abysmal build system structure and the fact that there are still no git submodules. anyone interested in collaborating on making a separate repository that contains all the same stuff but a) a different build system that is not totally broken (redo) b) dependencies as git submodules? 10:31 erlehmann i could continuously rebase that on the minetest master, but i doubt it makes much sense if i am the only one interested in it 10:32 erlehmann well, not sure if rebase would cut it, but i'd basically want to take all the features, just replace the build system and how the source is handled. 10:32 erlehmann then try to keep it up to date 10:43 MTDiscord how is the current build system bad? 10:50 erlehmann Sublayer plank i opened an issue some time ago when i noticed that *some* incremental builds were not producing correct binaries, it contains a long explanation: https://github.com/minetest/minetest/issues/11749 TL;DR: the only way to get a somewhat reliable build right now is to rebuild absolutely everything from scratch, every time. 10:52 erlehmann also sfan5 explained some time back one reason why appgurueu und me continue running into bisection problems, the current build system literally *can't* handle going back to a previous revision … unless you rebuild everything from scratch, which is why i haven't opened a separate issue for it. 10:55 erlehmann the thing is, i have written a build system myself and know how to solve this in probably only a few hundred lines of code. but my solution involves applying a different paradigm to the build: instead of figuring out before what you want to build, topologically sort it, then hope for the best – i collect information during or after the build. i will trace the compiler syscalls if necessary, but usually there exist bette 10:55 erlehmann r (read: faster) methods. 10:56 erlehmann i.e. whatever solves this is not only not compatible with cmake in theory, i have seen someone who knew much more than me about cmake try to make cmake work like it and fail 10:56 erlehmann (i barely know anything about cmake beyond “it does not work”, you have to ask josiah_wi for details) 10:58 erlehmann sublayer plank, if you are interested in the technical side, check out the game liberation circuit and build it. 10:58 erlehmann the build uses some clever tricks that minetest could also have (but not if cmake is used) 10:59 erlehmann disclaimer: ofc i am biased bc a bunch of those tricks (but not all of them) were implemented by me. 11:17 sfan5 we're not changing our build system for this 11:19 erlehmann sfan5, yeah i totally expected that, which is why i asked who is interested as doing this separately 11:20 erlehmann seriously, it is hard to imagine a build system that's worse than the current one and still mostly works 11:20 erlehmann i guess it's like the particle situation “we couldn't make this worse if we tried” 11:34 MTDiscord Commercial build systems can are are far worse than ours 11:35 MTDiscord The fact that I can do it in under ten minutes on windows is impressive. You have no idea how convoluted build systems from even 5 years ago are. 13:09 erlehmann exe_virus ok i take it back, it could be worse. it could download dependencies during the build. 13:09 erlehmann we're not changing our build system for this 13:09 erlehmann why 13:09 erlehmann like, i'm not saying do it 13:10 erlehmann but i'd like to know what you love so much about the thing that you want to keep it while usually stuff that does not make sense is kicked out of minetest even if it worked before if it has better alternatives and doesn't break too much 13:10 erlehmann “it works 13:11 erlehmann ” can't be it, because it doesn't, reliably 13:11 erlehmann also you told me once in a ticket that you have no idea how to actually track dependencies that aren't tracked, so it can't be domain expertise with cmake 13:12 sfan5 the cost-benefit ratio is off 13:12 erlehmann (which would be a good reason btw, keeping a system whose pitfalls you know can be better than having someone where you have no idea) 13:13 erlehmann what cost? adding an additional build system costs *you* nothing. it only needs to be implemented once. 13:15 erlehmann the benefit would be: faster build times, actually working incremental builds, working bisect, ppl stop having these ridiculous wrong binaries 13:15 erlehmann oh and i'd ofc keep it simple enough so that adding deps is not black magicks 13:16 sfan5 no implementation cost? no maintenance cost? no support costs? 13:16 erlehmann also i can plot the dep tree in dot and stuff 13:16 sfan5 you should aim for a nobel prize with things like this, not improve a random block game 13:17 erlehmann what if i offer to do the task 13:18 sfan5 use your brain to answer that question considering all available information 13:18 erlehmann i actually *have* written a build system that is faster, smaller and more reliable than make, in 2014. that task also involved a lot of reading about previous work and using other build systems, before i even started. 13:20 erlehmann well my brain tells me “sfan5 is being dismissive and sarcastic about the topic of build systems to a domain expert who offered to do work for free (me)” 13:20 erlehmann look, i get it, i'm not getting an answer out of ppl who are “think about it, you'll get to it yourself” if i ask them about their opinion 13:21 erlehmann i can't read minds 13:21 erlehmann and “think about it, you'll get to it yourself” works only for stuff that is objective. like if i tell you there is a backdoor in minetest 5.3 that was accidentally fixed, you can probably find it if you go looking for it (would be useless to look for it though). 13:22 erlehmann well, backdoor. a vulnerability. 13:22 erlehmann wrong choice of words. 13:22 erlehmann whatever 13:22 erlehmann this is not helping 13:22 sfan5 stop leaving before I can answer, fucker 13:23 sfan5 I highlighted three problems in my answer, you answered "what if I solve one of them" 13:24 sfan5 I don't know if you are dumb but if you are smart enough to write a build system I bet you are smart enough to understand a simple argument 13:24 sfan5 regardless, please stop wasting the time of me and everyone who is reading this chat 16:39 MTDiscord it appears the github ci is stuck still 17:09 Fleckenstein Do lbms with run_at_every_load = true run after a new block was generated? 17:20 sfan5 I'd say not but should be easy to test 21:17 v-rob sfan5: #11821 - better a day late than never 21:17 ShadowBot https://github.com/minetest/minetest/issues/11821 -- Add padding[] element to formspecs by v-rob 21:20 MTDiscord thanks vrob 21:45 sfan5 thanks