Time Nick Message 00:28 paramat will merge #7257 in an hour (trivial) 00:28 ShadowBot https://github.com/minetest/minetest/issues/7257 -- Cave liquids: Use a more precise point for calculating biome by paramat 00:28 paramat ^ good screenshot 00:42 asourcefish Hi! 00:47 asourcefish Won't anyone mind two patches from OpenBSD's ports tree? 00:48 asourcefish https://github.com/openbsd/ports/blob/master/games/minetest/patches/patch-src_CMakeLists_txt 00:48 asourcefish https://github.com/openbsd/ports/blob/master/games/minetest/patches/patch-src_unittest_test_serialization_cpp 06:33 asourcefish I've sent 5 pull requests. 06:33 asourcefish It is for you to decide to accept them or not. 06:33 asourcefish See you at github. 10:50 opal master branch is 0.5-dev, correct? if so, i'll compile HEAD on backport-0.4 and see if i can reproduce the bug i'm having on that version 10:51 ANAND yes, 0.5-dev is the master branch 10:52 ANAND backport-0.4 is 0.4.17-dev though, your bug might've already been fixed by one of the many commits after 0.4.16 was released 10:52 opal yeah, that's what i'm gonna figure out 10:52 opal from the backtrace i have, the issue seems to be with path finding 10:53 Krock Will push trivial build fix for #7258 https://pastebin.com/raw/JSyhVD6f in 10 minutes 10:53 opal seemed to be triggered from the mineclone2 game i was playing since there was a call to a lua function just before that 10:53 ShadowBot https://github.com/minetest/minetest/issues/7258 -- Android build failure 10:53 opal if it helps at all, i can paste the backtrace on a site, or i can just compile this latest version and see if that takes care of it 10:54 Krock opal, is it this issue? https://github.com/minetest/minetest/issues/7132 10:55 opal no but it's very similar. recursion handling with musl then? 10:56 opal wait, it's actually the same issue but with different backtrace 10:57 opal "A segmentation fault happens in function Pathfinder::updateAllCosts." that describes my issue, i just read the backtrace before reading anything else in the issue lol 10:57 Krock hashtag metoo 10:57 opal lol 10:57 ANAND XD 10:58 Krock means, it's not only an issue on Alpine 10:58 opal yeah i'm hardcore and use alpine on my desktop lmao 10:58 opal ya all musl distros 10:59 opal i'm gonna take a trip to #musl to ask about the stack size limitation 11:04 opal ...would you call this a fault with musl then, and not with minetest? as in, would any fix on minetest's part just be a hack around musl's limitation? or are programs commonly built for glibc using false assumptions on how to write code? 11:05 Krock pushing build fix... 11:09 opal The default stack size for new threads on glibc is determined based on the resource limit governing the main thread’s stack (RLIMIT_STACK). It generally ends up being 2-10 MB. musl provides a default stack size of 80k. 11:09 opal yeah thats small 11:09 opal i really dont want to compile musl, i want my package manager to handle that lol. but i might end up having to repackage musl and replace the system build if this is a recurring issue 11:11 opal Krock: #musl's recommended fix is indeed to set pthread_attr_setstacksize() and stop relying on undefined behaviour 11:20 Krock now the question is how large the required stack is 12:03 sfan5 just set it to whetever glibc uses by default 12:35 rubenwardy Please can we not announce release dates any more 12:35 rubenwardy Like, on the forum 12:36 rubenwardy Instead announce release candidates on the start of feature freeze, and then the actual release 12:37 sfan5 0.4.17 will be released very soon (April 23rd IIRC) 12:37 sfan5 is this what you mean 12:42 rubenwardy Yes 12:42 rubenwardy Well, April 15 12:43 rubenwardy There's just no real reason to 12:43 Krock Wasn't it in December? 12:43 rubenwardy And it raises expectations too much 12:46 ANAND valid point 13:54 opal ok i changed src/threading/thread.cpp, gonna see if i have any issues now 13:54 opal should i release a patchfile for the changes i made? it's like 4 lines diff and pretty easy to implement 13:56 Krock -> new PR 13:59 opal 09:59.10 < opal:#minetest> if so, i'd like to report a bug (and possibly submit a patch) but i don't have a github account. is there any alternate method i can use to report the issue? 13:59 opal 09:59.41 < opal:#minetest> and before someone says "make a github account", no, i dont agree to their terms of service 14:00 opal most i will do is git request-pull and send in / paste the resultant patch 14:07 opal well this is fun, i cant even test my change because compiling with pthreads fails. compiling with C++11 threads works fine though 14:08 opal and the error isnt a result of my code (i reverted my changes to be sure, same compile error) 14:09 opal so i'll try running 0.4.16 with C++11 threads for a while and see where that takes me 14:13 opal then again i know hardly anything about C++11 and maybe the underlying implementation is still pthreads 14:14 opal or even more underlying than that, it still doesnt set the stack size 14:14 opal i'll find out 14:23 sfan5 backports branch updated! 14:24 rdococ Is the bug I encountered fixed now? 14:26 sfan5 I can't read your mind 14:29 rdococ an error in content_mapblock.cpp, I can't remember the details but it was something to do with "const"/"constexpr" 14:35 sfan5 was fixed yesterday 14:50 rubenwardy Just post a link to a patch here 14:51 rdococ looks like latest backport-0.4 compiled correctly 15:21 opal https://stackoverflow.com/questions/13871763/how-to-set-the-stacksize-with-c11-stdthread 15:21 opal the hell am i looking at 15:21 opal @ZamfirKerlukson: The language designers did think about this. You get the native handle from the Thread object. Then manipulate it with appropriate native function calls for your platform. But the reason they made it hard is that doing it is usually a mistake and when actually required is very/very rare. – Martin York Dec 14 '12 at 4:20 15:22 opal so uh, c++11 doesnt require a stack for std::thread but it still exposes an underlying thread implementation? what? 15:22 sfan5 no 15:22 sfan5 the 'doesnt require a stack' remark is about C++ being able to run on platforms without a stack 15:22 sfan5 thus std::thread does not include functions related to stack sizes 15:24 opal one of the many reasons i prefer C 15:25 sfan5 C11 threads do not provide this feature either 15:25 opal didnt know C11 had threads 15:25 opal lol 15:26 opal i only do C89/99 dev when im dealing with C 15:34 opal /home/wowaname/src/minetest/src/threads.h:99:33: error: cannot convert 'threadid_t {aka std::thread::id}' to 'pthread_t {aka long unsigned int}' for argument '1' to 'int pthread_equal(pthread_t, pthread_t)' 15:34 opal thr_compare_thread_id() takes two threadid_ts 15:35 opal wouldnt the appropriate thing to do be to change the entire function depending on whether someone is compiling with posix thread support? 15:46 rubenwardy As we're using c++11, we shouldn't need to change depending on os 15:46 rubenwardy Not checked that area though 15:48 opal 0.4.x still shows support for pthreads. idk what youre doing for 0.5.x, maybe you removed that "dead code" now but as it stands, 0.4.x is buggy in that regard 15:48 opal unfortunately i am barely familiar with the codebase and i'm not sure if you guys already tested pthreads or if it's really just dead untested code at this point 15:49 opal i'd be willing to fix it if the latter is true 15:49 opal otherwise i dont want to poke my nose where it doesnt belong, maybe the issue isnt what i mentioned and maybe its something else 19:01 paramat wll merge trivial #7257 in 15 mins 19:01 ShadowBot https://github.com/minetest/minetest/issues/7257 -- Cave liquids: Use a more precise point for calculating biome by paramat 19:02 paramat will merge #7244 in 2 hours if no objections 19:02 ShadowBot https://github.com/minetest/minetest/issues/7244 -- Mgvalleys: Use shared tunnel / cavern code instead of internal by paramat 19:03 nerzhul paramat where is the approval ? 19:11 paramat it's fairly trivial and discussed with an owner of a mgvalleys server, i'll label it 19:13 nerzhul i see at least 500 lines changes 19:16 paramat yes a deleted function, it calls the shared code instead 19:32 paramat merging #7257 19:32 ShadowBot https://github.com/minetest/minetest/issues/7257 -- Cave liquids: Use a more precise point for calculating biome by paramat 19:33 nerzhul merging #7259 19:33 ShadowBot https://github.com/minetest/minetest/issues/7259 -- Fix i386 build at OpenBSD by mazocomp 19:36 paramat done 19:58 opal .... 19:58 opal did USE_CPP11_THREADS and USE_POSIX_THREADS seriously both simultaneously get defined when i tried to build 19:58 opal ugh 20:04 opal if one of those are manually defined from cmake then it shouldn't automatically try to assign... guess my fix is gonna be a little less simple 21:06 paramat will merge #7244 in 30 mins 21:06 ShadowBot https://github.com/minetest/minetest/issues/7244 -- Mgvalleys: Use shared tunnel / cavern code instead of internal by paramat 21:23 rubenwardy Just because you're an expert on an area doesn't mean you won't make midtAkes 21:24 rubenwardy For example, I'm probably an expert on formspecs and the Lua API yet will still need my stuff reviewed 21:24 rubenwardy It's fine if it's mostly moving code thouvh 21:24 rubenwardy Argh, mobile 21:25 fwhcat Sometimes, even if you review your own code, you'll find mistakes (especially after some time) =) 21:26 rubenwardy Oh yeah, definitely 21:26 rubenwardy And also clean per of eyes help 21:27 nerzhul rubenwardy, +1 21:28 rubenwardy Trivial means there's 0 chance of causing bugs 21:28 rubenwardy Well, very low chance 21:28 fwhcat nerzhul, are you integrating a postgresl connector directly into core ? 21:29 fwhcat or already did 21:29 nerzhul it's already the case lol 21:29 nerzhul since 0.4.15 or before ? 21:29 nerzhul it's my favourite :) 21:30 fwhcat Yeah it's the best, HOT backups is a neat feature compared to mysql 21:30 rubenwardy Postresql as been there for a while 21:31 fwhcat Sometimes i'm like waking up from a coma, sorry 21:32 fwhcat Btw, don't it bother you guys that lua_api.txt is actually written in markdown but not rendered on github? 21:33 rubenwardy It's because you can't link to lines in github markdown 21:34 paramat rubenwardy, i agree (about 7244) :) 21:36 fwhcat hum okay 21:37 fwhcat but you could links to any title ex: https://github.com/minetest/minetest#default-controls 21:38 fwhcat so if we put titles then we could still point to a precise part of the api 21:38 Krock fwhcat, how would you directly reference to an API function which is 200 lines below the title? 21:39 Krock "More titles" sometimes it's that easy as the functions can be used in various ways 21:39 Krock s/it's/it's not/ 21:39 fwhcat well, function would have a h3 or h4 depending on how you organize that code. but you really need to point to a specific line of code in a function? 21:41 Krock specific line of the documentation? Yes. 21:41 paramat merging 7244 21:42 Krock paramat, the variable near_cavern looks quite useless but the code is fine otherwise 21:44 paramat hmm .. 21:45 Krock if you didn't merge it already - also consider adjusting the noise params to the float notation 21:45 paramat actually 'near cavren' is needed in that form 21:45 paramat *cavern 21:46 Krock I mean, it's an useless variable. You could call that function inside the if 21:46 paramat yes about noise parameters, i have investigated whether they need to be specified as flots, they don't because they are converted to floa when read 21:46 paramat *floats 21:46 Krock or initialize `near_cavern` directly with the return value :) 21:47 paramat oh i see what you mean about 'near cavern' 21:48 paramat still all mapgens are like that so i'll fix that later 21:49 paramat i'm going to follow this up with a big cleanup of mgvalleys anyway 21:49 paramat merging .. 21:54 paramat done 21:54 paramat yes the code is still a mess but i deliberately left out code cleanup as i'll do that next 22:00 paramat ahh at last, i've been itching to continue the cleanup