Time Nick Message 00:19 HuguesRoss erle: As suggested by savilli, the number 4096 comes from division by the mapblock size 00:22 HuguesRoss I'm not sure if it's correct to clamp the values, since this could result in some weirdness elsewhere, but we definitely ought to do *something* for this call. From a quick check of the lua API code I think there are several other calls that can repro similarly problematic ranges 00:24 HuguesRoss Ignoring calls like your mod does could be an option, though it would require us to check everywhere (and not forget to do so in the future). Maybe we can make a call specifically for parsing a pair/region of int vectors with a safety check? 10:44 MTDiscord hey btw i finished the work on #11016 ... its not like super urgent just wanted to make sure you know 10:44 ShadowBot https://github.com/minetest/minetest/issues/11016 -- Dual Wielding by EliasFleckenstein03 10:59 MTDiscord will 5.5.1 include android support? 11:15 sfan5 exactly as much as 5.5.0 11:36 paradust I'm getting closer to finding the Android bug. It's definitely something going wrong in LuaJIT. 11:37 paradust With built-in lua it doesn't crash, just painfully slow. 11:37 paradust And the log from crashing always shows it is trying to jump to an invalid address, which ends up being near the luajit generated code area 11:41 MTDiscord by the android bug do you mean the increased crashes on arm64 builds? 11:41 paradust yea 11:46 erle paradust any idea why it did not happen before? 11:47 paradust i'm not sure what changed. Was it stable before on arm64? 11:48 erle i can not possibly comment, maybe rubenwardy can share some of the crashlog data from the dashboard or how its called 11:53 MTDiscord arm64 has always been considered somewhat unstable compared to armeabi-v7a builds, but it increased significantly after 5.5.0 I'd assume 11:56 erle ROllerozxa, is arm64 the same as aarch64 11:57 erle i am confused by all these names 11:57 MTDiscord yeah it's what they call android native libs for aarch64 11:57 sfan5 paradust: you might want to check which luajit version we used before then 11:57 sfan5 and plain lua shouldn't be "painfully slow" 11:58 sfan5 (depending on what you're doing) 11:58 erle i have used lua5.1 and luajit on my thinkpad t60 and minetest is totally playable with both 12:00 erle paradust if you find a crash i can try on reform2 to reproduce 12:00 sfan5 current is 2.1.0-beta3, previous was apparently whatever was at https://github.com/LuaJIT/LuaJIT/commits/v2.1 at the time 12:01 sfan5 probably this revision https://github.com/LuaJIT/LuaJIT/commit/d85d6b3c1b2c3d1ecf89ccff6ad7519e767fcebe 12:02 sfan5 now luajit does not do stable releases since quite a while so we could just upgrade to the v2.1 branch 12:02 sfan5 but you never know if you're pulling in a regression like that 12:02 erle even a single reproducible crash test could possibly allow to bisect it though 12:03 sfan5 why bisect something that's probbly already fixed? 12:03 sfan5 +a 12:06 paradust LuaJIT-2.1.0-beta3 is dated 2017-05-01 12:07 paradust it's worth testing it against latest to see if it stops crashing, i'll do that 12:11 paradust sfan5: re: painfully slow, I'm building debug, so that probably has turned off all optimization 12:12 paradust it's noticably slower on mineclone, nodecore 12:15 MTDiscord paradust: TYSM for looking into the android problem. I cannot play minetest on my newish phone because of it, or at least, im stuck on 5.4, which sucks 12:16 sfan5 did you try the 32-bit version? 12:20 MTDiscord no, I was only aware of the google play version 12:21 sfan5 google play gives you the one that best matches but you can download any from https://github.com/minetest/minetest/releases/tag/5.5.0 12:21 MTDiscord oh great! Ill try it 12:21 sfan5 (armeabi-v7a is 32-bit) 12:23 erle paradust mineclone games do A LOT of unnecessary lua stuff though 12:23 erle no idea what nodecore does 12:24 erle sfan5 it is impossible to upload an aarch32 build to google play and let it run on aarch64 computers, right? 12:25 MTDiscord google play requires 64-bit builds being uploaded nowadays 12:25 MTDiscord just like how they required developers to move to scoped app storage... 12:26 erle what about fdroid, is it the same constraint? 12:26 erle uh, fdroid has 5.2 :( 12:26 MTDiscord only constraint in f-droid is that it can be fully compiled from available source code 12:26 erle https://f-droid.org/en/packages/net.minetest.minetest/ 12:26 MTDiscord which is why it's been outdated... 5.5.0 would technically be compileable fully from source on android again but they haven't gotten around to updating the buildscripts 12:27 erle wait how did 5.4 get to android 12:27 erle if it is not fully compileable from source 12:27 erle like what's the thing that was missing for fdroid 12:28 MTDiscord prebuilt dependencies 12:28 erle something something irrlicht? 12:29 MTDiscord something something throwing binary dependencies into a github repo, something something archived now in favour of a different repo with only buildscripts that can build the same binaries from source 12:44 erle :( 15:41 rubenwardy #12030 15:41 ShadowBot https://github.com/minetest/minetest/issues/12030 -- Deprecate game.conf name, use title instead by rubenwardy 15:41 rubenwardy #12284 15:41 ShadowBot https://github.com/minetest/minetest/issues/12284 -- [No Squash] Show dep errors in Select Mods modal by rubenwardy 16:43 sfan5 #12323 is ready, I will prepare the windows builds now 16:43 ShadowBot https://github.com/minetest/minetest/issues/12323 -- Backport 5.5.1 by sfan5 16:44 rubenwardy Have you gone through all fixes on master? 16:44 sfan5 hmm I guess I should fix the docker build too 16:44 rubenwardy I'll test content stuff 16:45 sfan5 yes this is all of them 16:45 sfan5 you may want to double check content-related fixes but it should have everything that is backportable 16:55 rubenwardy it's possible to enable two mods with the same name in backport-5 16:56 rubenwardy (obviously, this doesn't keep on save - but it shows in the UI) 17:01 sfan5 which commit would have fixed it? 17:03 rubenwardy for some reason `disable_all_by_name` is completely gone 17:04 rubenwardy oh wait, that's a virtual path thing 17:05 rubenwardy idk whether this is a problem on master 17:06 rubenwardy the fix is https://github.com/minetest/minetest/commit/128f6359e936bcdc5e26409ddd73438bce9c6dd6 though, which is pretty heavy 17:06 rubenwardy otoh, many of the future content patches rely on that 17:06 rubenwardy s/on master/on stable-5/ 17:08 sfan5 it's not like there are high impact bugs related to this so it's not worth spending time on IMO 17:10 rubenwardy ok this is a bug on stable-5 too, so it's fine 17:13 nrz_ it's release day today ? 17:13 sfan5 because I almost forgot it a reminder: when building don't forget to reset MTG to `backport-5` too 17:15 erle so how much time from release candidate to release 17:15 erle i can help test 17:15 sfan5 zero 17:15 erle well then i hope you got everything right 17:16 sfan5 you can test the backport-5 branch if you want but there will be no public rc build 17:16 sfan5 rubenwardy: you may want to wait with android until paradust can report whether a newer luajit version fixes issues 17:16 rubenwardy makes sense 17:16 sfan5 nrz_: tomorrow, just preparing things 17:16 rubenwardy also, tested content changes, all ok 17:17 rubenwardy other than virtual path related issues, but that can wait to 5.6 17:18 rubenwardy which is hopefully sooner than later - yet another user was confused about multiple mod installs today 17:19 rubenwardy wait that's weird, it's not enabling dependencies 17:19 rubenwardy ok there's a bug where it won't enable a dependency until you enable it at least once 17:20 rubenwardy this seems familiar 17:20 rubenwardy yeah 17:20 rubenwardy https://github.com/minetest/minetest/commit/7db751df3bc4e3b3aff80cdacd2883f3a7a0940b 17:20 rubenwardy !title 17:20 ShadowBot Fix broken dependency enabling due to missing `enabled` field (#12093) · minetest/minetest@7db751d · GitHub 17:20 rubenwardy that has partial dependencies on the virtual path fix 17:21 rubenwardy oh and the bug was a regression from that PR 17:21 sfan5 oh man I can reproduce the "media loading gets stuck in SP" on the win64 build (running in wine) 17:22 sfan5 with mineclone of course, MTG is not so heavy 17:22 erle sfan5 does it get stuck at 100%? 17:22 sfan5 no 17:23 sfan5 'stuck' not literally, it just slows down in the middle and eventually completes 17:23 rubenwardy can I push a fix to backport-5? Amended to be before the release commit 17:23 rubenwardy - if mod_to_enable.enabled == false then 17:23 rubenwardy + if not mod_to_enable.enabled then 17:23 erle oh, then it's different from the “media stays at 100%” thing i have still to figure out 17:23 sfan5 rubenwardy: sure 17:24 erle no 100% CPU, no RAM usage, minetest just loading media and then chilling the fuck out. once i figured out how to reliably trigger it, i'll you (on the issue tracker). 17:25 rubenwardy done 18:20 MTDiscord the 32 bit version works! 18:21 MTDiscord can we but the 32bit version of 5.5.1 on F-Droid please? 18:21 MTDiscord Sfan5 18:21 MTDiscord sure, if you submit the PR for it :p 18:22 rubenwardy yeah, f-droid has no problems with 32-bit only 18:26 MTDiscord strange... the 64bit version from github works too 18:27 MTDiscord Ima check the fdroid version again 18:27 MTDiscord wait no... 18:28 MTDiscord was it the google play version? Which one is on 5.5? 18:28 rubenwardy play is 5.4.1 as 5.5.0's rollout was stopped 18:29 MTDiscord oh... well, both github version work on my device 18:29 MTDiscord *s 18:34 MTDiscord well, at one point, one of the versions did not work (either from F-droid or from play). But I have now tested all of them (both githubs', fdroid's current, and Play's current), and they all work. so IDK