Time Nick Message 01:29 paramat thanks for merging nerz =) 04:11 paramat merging game#2284 04:11 ShadowBot https://github.com/minetest/minetest_game/issues/2284 -- Fix small pine leafdecay: New schematics, new leafdecay radius by paramat 04:12 paramat done 06:54 nerzhul rubenwardy: 18sec ? what happen every 18 sec ? 06:54 nerzhul you talked about player attriute and said playermeta can mitigate it seems 08:15 nerzhul rubenwardy: https://www.sqlite.org/asyncvfs.html 08:15 nerzhul we should either ensure how we write, or use this vfs, which can be blazing fast :) 08:17 nerzhul oh i didn't see that it's not supported 08:17 nerzhul but maybe our wal mode is too heavy 08:48 oiaohm nerzhul: you do need both. sqlite wal is great on local file systems cannot be used at all on network file systems. asyncvfs is normally slower on localfile systems but can work on network ones. 08:49 nerzhul you should never store your sqlite db on a NFS 08:49 nerzhul asyncvfs is not supported anymore then i think it's not a good option 08:49 nerzhul i think the synchronous mode if enabled should be disabled 08:49 nerzhul let the filesystem do what he wants 08:50 nerzhul or we should make all our db calls asynchronous with our implementation 08:52 oiaohm nerzhul: sqlite on Linux nfsv4 can be safe. But you cannot use wal mod 08:53 oiaohm mod/mode 08:53 nerzhul i should verify how we save player, if we have a such penalty it's strange 08:53 oiaohm Really if you are needing to network it most likely it better to switch to the postgresql backend anyhow. 08:54 nerzhul i definitively recommend postgresql for large servers 08:54 oiaohm Its not like using network file system is without serous overhead. 08:54 nerzhul > 10 players connected everytime should must use postgresql 08:55 oiaohm nerzhul: so at that point even if you system is a beast of a local machine sqlite with wal set right is not keeping up? 08:56 nerzhul with SSD i don't think you will have problems on the db 08:56 nerzhul i'm reading the player save code 08:57 nerzhul we save them only if they are modifeid 08:57 nerzhul except on player disconnection 08:58 oiaohm Question what is higher postgresql in it current form is optimised with master-slave relication is optimsied for read heavy. 09:00 oiaohm If this is write heavy multi server postgresql at this stage could be a very bad choice. 09:00 oiaohm https://wiki.postgresql.org/wiki/Multimaster << Yes the performance graphs there are still about right. 09:03 nerzhul multi server postgresql for minetest is totally overkill 09:04 nerzhul minetest max db size is ~15GB only 09:04 nerzhul master slave is useful in MT only for backups 09:04 nerzhul and minetest is not very write heavy 09:05 nerzhul i have write heavy postgresql servers at work i manage, and they are as master slave and it just works well 09:05 nerzhul :) 09:05 oiaohm nerzhul: some cloud hosting you have to special ask to get single instance postgres 09:05 oiaohm nerzhul: otherwise you get multi master. 09:05 nerzhul if you use aws rds you have too many money 09:05 nerzhul :p 09:06 oiaohm I knew it most likely would be over kill. 09:06 nerzhul i push a trivial change to master (we create the players file backend directory in every backend on each saveLoadedPlayers 09:08 oiaohm I have played around with postgresql bpr that is the multi master it is better with write heavy work loads compared to normal master slave. 09:08 nerzhul wow rubenwardy i may have found the sqlit eissue 09:08 nerzhul and it's in pg too it seems 09:09 nerzhul yeah it's that 09:09 nerzhul player->setModified(false) is only called in file backend 09:09 nerzhul on sqlite & pg we forgot to set it to false 09:09 nerzhul then we write on each server loop 09:09 oiaohm Ouch. 09:10 nerzhul it's why sqlite is slow :p 09:10 nerzhul pg is more heavy but less synchronous because it's done by another process 09:11 oiaohm Even with pg a extra write are still not want you want. 09:11 oiaohm Its extra stuff for the pg wal logging to be sending to backup. 09:11 nerzhul also in file backend sao->getMeta().setModified(false); is not called 09:12 nerzhul then we write meta each time 09:12 nerzhul current sao->getMeta().setModified(false); is called before commiting to sqlit e& pg, ti's bad because database hasn't ack the write :) 09:13 nerzhul why do we call sao->getMeta().setModified(false); on RemotePlayer::serializeExtraAttributes 09:13 nerzhul it's not logicall 09:13 nerzhul if i want to print that i cheat the save 09:14 oiaohm Sometime with how lacking logic code can be you wonder how it works so well for so long without causing major problems. 09:16 nerzhul i fix all of that 09:16 nerzhul each backend doesn't commit the engine in the same way 09:16 nerzhul ti's crazy :D 09:16 nerzhul and yes i agree with rubenwardy that th e current sqlite backend is slower than files because we do heavy writes 09:16 nerzhul sqlite3 journal should be growing fast 09:20 nerzhul pushed to master 09:20 nerzhul this is not a very heavy pr it's just code move 09:22 nerzhul see #c1d7dbfc38e8f48b42d957afad2696a8e56b740b 09:22 nerzhul !c1d7dbfc38e8f48b42d957afad2696a8e56b740b 09:22 nerzhul no it seems we cannot link it with bot 09:22 nerzhul https://github.com/minetest/minetest/commit/c1d7dbfc38e8f48b42d957afad2696a8e56b740b 09:24 oiaohm For repeating mistake might pay to check the map saves as well if you have not already. 09:24 nerzhul map save is okay 09:38 nerzhul interesting to find a huge save bug while trying to merge #8043 :) 09:39 ShadowBot https://github.com/minetest/minetest/issues/8043 -- Make sqlite3 default auth & player backends for new worlds by nerzhul 09:39 nerzhul at least a very anoying bug fixed for 5.0.0 10:49 Krock reviewed #8043 10:49 ShadowBot https://github.com/minetest/minetest/issues/8043 -- Make sqlite3 default auth & player backends for new worlds by nerzhul 10:53 nerzhul krock yeah ! in the previous version of the pr is was not the problem but here yes :) fixing 10:54 nerzhul ty 10:54 Krock nice 10:55 nerzhul pushed 10:59 nerzhul don't hesitate to tell me if it's okay or not :) 11:26 Krock nerzhul: Travis failed due to an unrelated build setup failure. Merge when you want 11:54 nerzhul nice, ty Krock 12:52 nerzhul note i merged it :) 13:11 nerzhul nice the beta program works as intenteded 13:11 nerzhul 10 users has installed it 13:11 nerzhul and we got Crash java.lang.UnsatisfiedLinkError: dlopen failed: cannot locate symbol "stderr" referenced by "libgmp.so"... 13:11 nerzhul i tried to promote the apk on 5% of the users in production and the store warns me about this problem 13:11 nerzhul we should fix it 13:12 nerzhul it's related yes to devices before 6 13:12 nerzhul !tell stujones11 nice diag on the libgmp, the beta users proove you are right 13:14 nerzhul wow google records screen video 13:17 Krock nerzhul: tell didn't work. space. 13:17 nerzhul i bump android version code in stable-0.4 and master because of the beta program 13:17 nerzhul Krock yeah my client handle it i tried with a space, not good :p 14:13 nerzhul sfan5 does multicraft uses minigmp ? 14:13 nerzhul i have linking issues using it 14:14 sfan5 https://github.com/MultiCraftProject/MultiCraft/blob/master/build/android/Makefile 14:14 sfan5 I can't find gmp here, so I'd guess yes 14:15 nerzhul yes it's the case 14:15 nerzhul https://github.com/MultiCraftProject/MultiCraft/blob/master/build/android/jni/Android.mk 14:15 nerzhul oh 14:15 nerzhul typo fix in the variable name on which append the source 14:15 nerzhul it should be that :) 14:31 nerzhul erf it crash with mini gmp on my device 14:33 nerzhul and logcat doesn't show me any mt log, interesting 14:38 nerzhul oh i forgot to fix java part 14:38 nerzhul it always load libgmp.so it seems :) 14:41 nerzhul yeah, recompiling it should be good 15:08 nerzhul in debug it seems or not in release, i clean all an retry 15:29 nerzhul Krock are you there ? 15:30 Krock no, I'm in #-hub 15:30 nerzhul xD 15:30 nerzhul https://github.com/minetest/minetest/releases/download/0.4.17.1/Minetest-0.4.17.3.apk 15:30 nerzhul new version with mini gmp 15:31 nerzhul if it's okay i push the new version to beta program and merge #8047 in either stable-0.4 and master 15:31 ShadowBot https://github.com/minetest/minetest/issues/8047 -- Drop libgmp on Android and use mini-gmp by nerzhul 15:35 nerzhul tell me if now it starts :) 15:36 Krock yes it does 15:36 Krock yay mainmenu loaded 15:36 nerzhul nice 15:37 nerzhul you can try more but it was the goal 15:37 Krock wrong mintest_game version 15:37 nerzhul hmm ? 15:37 Krock get_decoration_id is nil 15:37 nerzhul it's 0.4.17.1 15:37 Krock or maybe it didn't delete the games properly? 15:37 nerzhul it's totally possible 15:38 nerzhul if i remember we are not very clean with the sdcard 15:38 nerzhul you can unzip the apk to ensure it's right if you have doubt 15:38 nerzhul apk are just zip 15:39 nerzhul i diffed my build workspace and the games/minetest_game folder in 0.4.17.1 and diff -ru said they are same 15:40 nerzhul i merge the GMP PR 15:40 nerzhul it's not related to the release problem you mention 15:40 Krock works 15:40 Krock I deleted the games dir and reinstalled itr 15:41 Krock it's running surprising smooth on these 700 MiB RAM 15:41 Krock and weak 4-core ARM 15:41 nerzhul yeah and it's not the 5.0.0 :D 15:41 nerzhul irrlicht is not very heavy 15:42 nerzhul maybe also irrlicht in release mode is smooth haha 15:42 Krock getting 29 FPS, but it doesn't look too beautiful 15:44 Krock actually tested on CyanogenMod but I don't think that makes much of a difference 15:46 nerzhul no because it's a low level system problem 15:46 nerzhul 29 fps is quite nice 15:46 nerzhul i cherry pick dda844be1628976cf1f41180e05e37afb374ef91 to master 15:48 nerzhul version has been published to beta on the play store 15:49 nerzhul we will see on sunday if the crash will disappear and if yes on all devices i promote to probudction mode 15:58 nerzhul thanks Krock for your time 15:59 nerzhul one less point for the 5.0.0 release :p 15:59 Krock :) 16:06 nerzhul i'm working on building android debug apk in the gitlab CI 16:07 nerzhul if it works as intended and not too long i will push it this weekend 16:07 nerzhul in a pr 16:24 p_gimeno wow nerzhul you're on fire :) well done 16:26 nerzhul p_gimeno: we should remove that shitty problems on android, yeah i know it remains gui scaling, but if the app starts it's better at a point :D 16:26 nerzhul and i want this 5.0.0 to be released 16:26 nerzhul other coredev please start this freeze 16:26 nerzhul and we embed the CDB problem in the freeze 16:27 p_gimeno me too, but I would hope that #7914 goes in. It's very minimum and will allow a long standing bug fix to be merged in 5.x. 16:27 ShadowBot https://github.com/minetest/minetest/issues/7914 -- Future-proof 5.0.0 for fixing #4183 16:30 nerzhul i'm not the best to review that part 16:30 nerzhul but rubewardy, Krock can you take a look ? 16:31 p_gimeno Krock already did and I believe he approved it 16:32 p_gimeno nerzhul: it's very easy to test, e.g. install mobs classic and mobs_redo and try to enable one of them, you can't 16:32 p_gimeno you don't even need a modpack to reproduce 16:32 Krock Approval: https://github.com/minetest/minetest/issues/7914#issuecomment-443827039 16:32 nerzhul yep but i talk about the fix in the code :) 16:32 nerzhul do we have a pr to set the review in the correct place ? :( 16:33 p_gimeno no :( 16:33 nerzhul why don't you push as you participate on discussion ? 16:34 p_gimeno I've archived the repo and I'm going to delete it and my user. I'm not using repos in github. I'm holding deleting my user until 5.0.0 is out. 16:34 p_gimeno It's in gitlab though 16:35 p_gimeno you can review there if you feel like that, https://gitlab.com/pgimeno/minetest/merge_requests/2 16:36 nerzhul p_gimeno how did you talked with us after ? :p 16:36 p_gimeno um, here? 16:36 p_gimeno at least until you move, if you do 16:36 Krock let's move to this lit new service called discord 16:36 nerzhul you should have forked gitlab.com/minetest/minetest instead :p 16:36 p_gimeno there's also #7901 16:36 ShadowBot https://github.com/minetest/minetest/issues/7901 -- Accepting PRs/patches from non-Github users 16:37 p_gimeno I have forked that 16:37 p_gimeno ^@nerzhul 16:38 Krock nerzhul: mind checking #8013 as well? It's not high-priority but a nice bugfix to have in 5.0.0 16:38 ShadowBot https://github.com/minetest/minetest/issues/8013 -- Fix various bugs (Anticheat, Lua helpers) by SmallJoker 16:39 p_gimeno but I can't submit PRs to gitlab.com/minetest/minetest, they are disabled 16:39 p_gimeno so I have to submit them to my own repo 16:39 p_gimeno hence why 7901 exists :) 16:51 rubenwardy nerzhul: good work recently :) 16:51 rubenwardy and nice fix, glad to have the cause found and fixed 16:54 Krock how about these GL_INVALID_ENUM errors? are they gone? 16:54 nerzhul Krock: do you see it ? 16:54 nerzhul i don't have it 16:55 nerzhul p_gimeno: right i disabled it to prevent errors :p 16:55 nerzhul and also gitlab sync from github 16:55 nerzhul rubenwardy: on sqlite or on android ? :p 16:55 Krock no I don't, that's why I'm asking 16:56 Krock because it could be that the OpenGL capabilities are reported correctly on this device 16:56 nerzhul Krock: we will see if users are complaining because we update and it's always here, but the root cause was found and i think i had a non cleaned-up workdir at a point 16:56 nerzhul exact :( 16:56 nerzhul #7901 16:56 ShadowBot https://github.com/minetest/minetest/issues/7901 -- Accepting PRs/patches from non-Github users 16:57 nerzhul Krock: the lua increment is indeed mandatory, api was broken in those serializations 16:57 rubenwardy both for first, and sqlite for second 16:58 nerzhul did you tested it ? 16:58 nerzhul i'm sure it fix your problem :p 16:59 nerzhul very agressive player saving, we ensure it ^^ 16:59 nerzhul Krock: can you remove the part sfan5 was not happy 16:59 nerzhul other parts are okay for me 17:02 p_gimeno I get the GL_INVALID_ENUM on my phone, I can try the new APK, where is it again? 17:02 nerzhul same place 17:02 nerzhul GRADLE_OPTS: "-Dorg.gradle.daemon=false" 17:02 nerzhul oops 17:02 nerzhul https://github.com/minetest/minetest/releases/download/0.4.17.1/Minetest-0.4.17.3.apk 17:03 nerzhul go android build on gitlab, go : https://gitlab.com/nerzhul/minetest/-/jobs/141611538 17:04 nerzhul wtf libiconv C compiler problem 17:04 nerzhul i'm using r17c 17:04 nerzhul (like on my machine) 17:05 rubenwardy I've deleted my player files since, so no, I can't really test it 17:05 rubenwardy well, I could C+P a shit load of files then import them 17:05 p_gimeno which one should I test exactly? the github or the gitlab one? 17:06 rubenwardy p_gimeno: github 17:06 nerzhul gitlab won't work :D 17:06 nerzhul take the built apk 17:06 p_gimeno ok 17:06 nerzhul it's the current beta release on the play store 17:06 nerzhul also my gitlab test is on master 17:12 nerzhul rubenwardy, Krock, paramat sfan5: do you agree to start the free on sunday (including remaining CDB PR in https://github.com/minetest/minetest/milestone/9) ? 17:12 nerzhul the freeze* 17:13 p_gimeno hm, the app has disappeared from the menu... maybe my OS is too old 17:13 Fixer nerzhul: before 5.0, please update all those translation templates so we can translate some new shit before release 17:13 nerzhul if i remember Krock has done it 17:13 nerzhul but maybe weblate has problem 17:13 nerzhul we can include that in the freeze 17:13 Krock I've seen a merge failure there but don't know who can fix it 17:14 Krock nerzhul: translations take some time, though 17:14 nerzhul rubenwardy, Krock, paramat sfan5: if you are okay we can define the release date to the 1st or 2nd of february 17:14 nerzhul i should look at it my key is on weblate, maybe ruben too 17:14 nerzhul else i should ensure to push rights to everybody who wants 17:14 Krock 1st is Friday FYI 17:14 nerzhul oops 17:15 nerzhul 2nd or 3rd :) 17:15 nerzhul 4 weeks of feature freeze, permitting to polish, test, update translations 17:16 p_gimeno ok, the GL_INVALID_ENUM seems fixed \o/ 17:16 Krock sounds good 17:40 sfan5 sounds good 17:50 nerzhul rubenwardy, paramat ? 17:52 Krock paramat will know due to !tell (did that) 18:16 nerzhul ty Krock 18:16 nerzhul seems we can go ahead :) 23:47 nerzhul !tell paramat did you see my request ? 23:47 ShadowBot nerzhul: O.K.