Time Nick Message 09:54 Krock does anyone have a testing mod for #9378 by any chance? 09:54 ShadowBot https://github.com/minetest/minetest/issues/9378 -- Formspecs: Add state-selection to style elements by Df458 09:56 Krock I guess the minimal formspec is all for now... 10:09 kilbith nobody cares, Krock 10:09 Krock !next 10:09 ShadowBot Another satisfied customer. Next! 10:09 kilbith MT is becoming irrelevant 10:10 kilbith IRC is very quiet, less *good* modders, less things going in git tree 10:10 kilbith s/in/on 10:11 kilbith rubenwardy is too busy, sfan5 cares about its own PRs mostly, ... 10:12 kilbith I suggested several times Df458 & v-rob as core-devs, but none seems to care either 10:12 kilbith it's the usual inertia in this community 10:14 kilbith just compare the current situation with eg. 2015 10:14 kilbith MT is on the decline 10:15 kilbith even that: https://github.com/minetest/minetest/pull/9411 ... is too much asking to review 10:16 kilbith you suck balls guys 10:18 kilbith ah of course the commits fixing two-lines of documentation will certainly make the engine more exciting! 10:30 kilbith fortunately I don't hire some of you otherwise we'd go bankrupt 10:39 Krock interestingly nobody noticed or cared about the compatibility in 9411. as long it's merged before 5.2.0 that won't be much an issue, though. 13:02 Krock https://github.com/MultiCraft/deps needs to be forked to the minetest organization for #9066 so that Android builds for the Play Store are again possible 13:02 ShadowBot https://github.com/minetest/minetest/issues/9066 -- Android: add Android Studio support, completely redone all of the java part by MoNTE48 13:03 Krock celeron55_: would you please be so nice to do that? A more fitting name would be "minetest_android_dependencies" 13:03 sfan5 last time I tested that PR the resulting apk crashed on startup 13:03 sfan5 but the build at least worked 13:03 Krock I tested an older one - based on the latest stable IIRC 13:03 Krock worked, but I don't know about the current state 13:12 celeron55_ what a mess 13:15 celeron55_ so what i take from this is, the current android build is broken, and this is the only finished branch fixing it? 13:17 celeron55_ and does some other things (plus monte48 offends 50% of people, making everything he does stink, more news at 11) 13:19 celeron55_ and he's promising to fix remaining problems with this if it's merged? 13:22 celeron55_ the PR description is incredibly bad in explaining what it's for 13:23 sfan5 I'd say the chances are good he will contribute the required fixes for android support 13:23 celeron55_ i hate android so i'm not going to test this 13:25 celeron55_ i trust monte48 to make it work for himself personally, but if it breaks something that doesn't increase his profits i'm sure he doesn't care 13:26 celeron55_ Krock: what will happen after creating said fork? 13:27 celeron55_ i don't see discussion about that in the PR 13:27 celeron55_ oh it's in rubenwardy's review 13:29 Krock celeron55_: the android build is not that broken. it still builds, but is no longer compatible with the Play Store rules 13:30 celeron55_ i'll create the fork; please continue making the PR happen then 13:30 sfan5 while it would be possible to make the old code work for the new requirements, the build system is a terrible mess anyway and definitely needs replacing 13:30 Krock we'll test that stuff. No other actions needed from your side than the fork tbh :) 13:32 Krock IIRC, stu and srifqi worked on Android, but they're quite inactive nowadays 13:32 Krock there was also an attempt to use CMake.. but that's long time history 13:32 sfan5 there's a few other things however: https://github.com/MultiCraft/deps/blob/master/Android/Build/Makefile#L81-L88 13:33 sfan5 these currently point to his fork with some custom changes instead of downloading the official source and patching it as needed 13:33 celeron55_ https://github.com/minetest/minetest_android_dependencies 13:34 sfan5 I don't think that's critical currently, the dependency stuff can be re-done using the existing patches we used at some point 13:34 celeron55_ add an issue in minetest_android_dependencies about references to monte48's forks 13:34 sfan5 will do 13:34 sfan5 you might also want to change the repo description 13:35 sfan5 oh and enable issues... 13:35 celeron55_ oh, maybe that 13:36 celeron55_ hmm 13:36 celeron55_ i wonder if the repo should be created from scratch in order to not have the forked status 13:37 celeron55_ actually that's what's disabling the issues 13:38 sfan5 sounds like a good idea 13:40 Krock indeed 13:48 Krock celeron55_: the repository is still empty. shall I proceed pushing the history to it? 13:48 celeron55_ i'm doing that 13:48 Krock okay. thank you :) 13:49 celeron55_ just have a slow connection 13:49 celeron55_ considered squashing the history but maybe not 13:50 erlehmann w8 13:50 erlehmann why squash the history? 13:50 celeron55_ it has old unused binaries in its history, being a binary repo 13:50 celeron55_ but it'll have more in the future anyway so whatever 13:51 erlehmann Krock, what are the play store rules that the android build breaks? 13:51 sfan5 lack of (working) arm64 build 13:51 erlehmann celeron55_, i think old binaries may be interesting for old phones? also, history editing is a cleaner way to remove the old binaries. 13:52 erlehmann than squash all commits 13:52 Krock I think they also increased the minimum ndk version 13:52 rubenwardy The repo should probably have "binaries" in the name, to differentiate between a future repo which contains build scripts 13:52 erlehmann what rubenwardy says 13:52 rubenwardy Not that it actually matter 13:52 celeron55_ well, it does matter 13:52 sfan5 rubenwardy: it contains the build scripts too 13:52 celeron55_ but yes, it contains both, which might be bad 13:53 rubenwardy In the future, I'd like the builds to be done using gitlab pipelines and exported as an artifact or docker image 13:53 rubenwardy Then the android build can be automated using pipelines as well 13:53 p_gimeno it has old unused binaries in its history, being a binary repo <-- you can use git filter-branch then 13:54 celeron55_ maybe minetest_android_deps_binaries and minetest_android_deps? ("deps" helps not having an insanely long naem) 13:54 celeron55_ name* 13:54 rubenwardy I've been feeling burnt out lately, which is why I've been less active 13:54 celeron55_ and i'd rename this now to minetest_android_deps_binaries 13:57 erlehmann rubenwardy, why docker image? 13:58 erlehmann rubenwardy, good reaction. continuing to work until burnout is bad makes recovery much worse. 13:59 rubenwardy erlehmann: it can then be used as a base in another gitlab pipeline 13:59 rubenwardy It's nice having all the dependencies in a docker image to make building faster 13:59 sfan5 sounds like a zip file would work just the same 14:00 rubenwardy This is something for another time though 14:00 erlehmann someone once explained docker to me as 14:00 erlehmann A: “but it works on my machine” B: “then we deploy your machine” 14:01 rubenwardy lol 14:01 erlehmann i mean, it is true 14:01 erlehmann in a way 14:01 rubenwardy I build my game, RVWP, using Gitlab Pipelines with this image: gitlab.com/rubenwardy/docker-sfml-rvwp 14:01 celeron55_ it's true, and that doesn't make it a bad idea 14:01 rubenwardy I guess I could just zip the dependencies 14:02 celeron55_ but docker images are kind of fuzzy compared to a zip with data, and a repo with code 14:03 rubenwardy plus a zip would be easier to integrate with the current gradle script 14:03 celeron55_ docker image is kind of like sharing a program on a floppy disk image, somehow weirdly it became popular again 14:03 celeron55_ s/on a/as a/ 14:04 rubenwardy I was getting ahead of myself, anyway 14:05 celeron55_ https://github.com/minetest/minetest_android_deps_binaries 14:05 celeron55_ i think that looks good now, and it has (0) issues 14:06 rubenwardy thanks 14:06 celeron55_ and minetest engine team access 14:07 celeron55_ i imagine someone will poke me when the non-binary counterpart is created, but that might be 2 years from now 14:08 rubenwardy I think any dev can create repos? 14:09 rubenwardy ah, no they can't 14:09 rubenwardy merging #9380 in 10 14:09 ShadowBot https://github.com/minetest/minetest/issues/9380 -- Add multiple element selection to `style` and `style_type` by v-rob 14:20 rubenwardy oh noes https://rwdy.uk/FOzSi.png 14:22 sfan5 start with this one https://github.com/minetest/minetest/pull/9369 14:29 rubenwardy ggrrrr, formspec.lua was CRLF 14:31 p_gimeno isn't there a .gitattributes entry for that? 14:31 rubenwardy potentially 14:31 rubenwardy all we have in .gitattributes is *.cpp diff=cpp \\ *.h diff=cpp 14:32 p_gimeno there's clearly room for improvement there :) 14:38 kilbith Krock: afaik I'm the only one to use animated_image[] in craftguide yet, so it doesn't matter 14:39 rubenwardy we also don't have to support things that are not yet in a release 14:39 rubenwardy except where it would cause world corruption 14:39 kilbith yeah that makes sens 14:39 kilbith +e 14:40 Krock rubenwardy: not like the assignments would change anything. I do the PRs that kinda jump "in front of my eye" and according to priority lists rather than assignments 14:41 rubenwardy ah right 14:41 Krock saves you a few clicks 14:41 rubenwardy I do the opposite - mark things I want to review as review requested, and then look through them a while later 14:41 rubenwardy I have this page bookmarked: https://github.com/pulls/review-requested 14:42 kilbith "I do the PRs that kinda jump in front of my eye" <- that may be the reason why the old PRs neglected 14:44 kilbith thank you rubenwardy for saving the day btw 14:44 rubenwardy huh? 16:26 DS-minetest now that I've added a subfolder in minimal in #9065, a unittest that checks whether the count of all media paths is a multiple of 5 fails 16:26 ShadowBot https://github.com/minetest/minetest/issues/9065 -- Load media from subfolders by Desour 16:26 DS-minetest how should I change it? 16:29 Krock fix the unittest 16:38 DS-minetest is there anything more to do when running the unittests than calling minetest with --run-unittests? 16:38 DS-minetest is there anything more to do when running the unittests than calling minetest with --run-unittests ? 16:38 DS-minetest oops 16:44 DS-minetest hmm, that unittest doesn't make any sense anymore. the paths aren't added anymore if the folders don't exist 16:46 Krock no, they just need to pass with --run-unittests 16:48 DS-minetest ok 16:49 DS-minetest I wondered because testStreamRead and testBufReader are failing 16:49 DS-minetest I've simplified the failing unittest now 16:54 p_gimeno I seem to recall there's an issue for that 16:55 p_gimeno if you mean it fails in release mode only 16:57 DS-minetest I've compiled with the default CMAKE_BUILD_TYPE 16:57 DS-minetest so Release, yeah 16:58 p_gimeno yeahhttps://github.com/minetest/minetest/issues/3943 16:58 p_gimeno *yeah https://github.com/minetest/minetest/issues/3943 16:58 p_gimeno #3943 16:58 ShadowBot https://github.com/minetest/minetest/issues/3943 -- testStreamRead and testBufReader failures 17:02 p_gimeno #7988 should have fixed that 17:02 ShadowBot https://github.com/minetest/minetest/issues/7988 -- IEEE floats: Fix unit test in release mode by paramat 17:04 p_gimeno (assuming it's the same failure) 17:05 p_gimeno anyway, -ffast-math is looking for trouble... is the difference really so significant? 17:05 DS-minetest yes, exactly the same failure as in 3943 17:05 p_gimeno DS-minetest: clang? 17:05 DS-minetest gcc 17:06 DS-minetest gcc (Ubuntu 7.4.0-1ubuntu1~18.04.1) 7.4.0 17:06 erlehmann who here knows about undefined behaviour? 17:07 DS-minetest is it even possible to know about something that is undefined? ;) 17:08 p_gimeno https://github.com/minetest/minetest/blob/master/src/unittest/test_serialization.cpp#L668 is supposed to not run that test if IEC 559 semantics are not supported. I guess gcc is lying again. 17:09 erlehmann https://stackoverflow.com/questions/7263404/mingw32-stdisnan-with-ffast-math 17:09 erlehmann > On linux, the gcc flag -ffast-math breaks isnan(), isinf() and isfinite() - there may be other related functions that are also broken that I have not tested. 17:09 erlehmann > Since -ffast-math instructs GCC not to handle NaNs, it is expected that isnan() has an undefined behaviour. 17:10 p_gimeno erlehmann: it's not about NaN, it's about about reciprocal math enabled by -ffast-math 17:10 p_gimeno basically, gcc believes that it's ok to change a / constant to a * (1.0/constant) 17:17 DS-minetest p_gimeno: testFloatFormat doesn't fail. the fails are at https://github.com/minetest/minetest/blob/7dffd08c1a068e5e0748a046fd8a1e884947b597/src/unittest/test_serialization.cpp#L308 (testStreamRead) and https://github.com/minetest/minetest/blob/7dffd08c1a068e5e0748a046fd8a1e884947b597/src/unittest/test_serialization.cpp#L471 (testBufReader) 17:21 erlehmann https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84949 17:21 p_gimeno oh right, I only fixed testFloatFormat 17:22 erlehmann > -ffast-math implies -ffinite-math-only which causes GCC to assume no inputs to isnan/fpclassify are NaN. As a programmer using -ffast-math you give this guarantee to the compiler, so what you report is a pilot error? 17:23 p_gimeno erlehmann: the test fails because -ffast-math implies -freciprocal-math 17:23 p_gimeno see the patch at the last post of #3943 17:23 ShadowBot https://github.com/minetest/minetest/issues/3943 -- testStreamRead and testBufReader failures 17:25 p_gimeno DS-minetest: apologies, I've forgotten quite some details by now. The patch I pointed out only addressed a failure that appeared later, not 3943. 17:25 DS-minetest so when reading a f1000 it does *0.001 instead of /1000 17:25 p_gimeno that's right 17:26 DS-minetest i see 21:30 kilbith can someone delete this please? https://github.com/minetest/minetest/pull/9430#issuecomment-593144764 21:30 kilbith this adds nothing to the discussion