Time Nick Message 09:24 celeron55 red-001 used to be an active contributor... well, based on the changelog, in 2016-2018 10:10 MTDiscord sfan5: Tbh ABM/LBM-like thing that would be ran for whole blocks and provide LVM... 13:07 Furi0us_mat Hi, I'd like to know how to export a minetest 3D model to b3d. 13:09 Krock Furi0us_mat: let's move to #minetest. This channel is mainly for the engine side of things, and less so for modding questions. 14:12 MTDiscord I'm trying to debug paroxysmal severe performance degradation that started after the 5.9 upgrade. After some number of hours of uptime, the server will just get shitty and lag will go all over the place. 14:13 MTDiscord I can run MT under perf and it looks like it will generate at least 2GB of data per day. Is there some way that I can crop these files down or something such that it covers only the critical portion? 14:13 sfan5 it sounds like the performance degradation is omnipresent 14:14 sfan5 right? 14:14 MTDiscord I've heard a lot of anecdotal reports of poor peformance on 5.9 but it's very hard to get much rigor on it. 14:15 MTDiscord In my case, on my server, I may be seeing SOME performance loss most of the time, I guess, but I'm mostly concerned with the parts where the server becomes suddenly completely unplayable. 14:15 MTDiscord I'm running a very lag-tolerant game, and players are also used to lag "weather" that's affected by things like background backup jobs and variable player traffic. 14:15 MTDiscord If you think that a trace during a non-meltdown period would be useful too, I suppose I could submit one. 14:16 MTDiscord There seems to be something going on though that's NOT in the main lua thread, as my attempt to dig something out with jitprofiler turned up nothing of interest, and that also tends to capture a lot of calls into the engine by mods. 14:16 MTDiscord It looks like whatever is going on is outside of env step 14:18 sfan5 what I was trying to say is: if the lag starts happening after 2 hours then you can just attach perf for 5 minutes at that point and likely get good data 14:20 MTDiscord I suppose I could try that. I'd have to do some funky manual shit then I guess 14:20 sfan5 https://github.com/minetest/minetest/blob/master/doc/developing/misc.md#remote-profiling 14:22 MTDiscord I'll give it a go. I'm not sure if the container has enough tools to do it, but I assume just having perf in there and a RelWithDebInfo MT server running should be enough ... 14:22 Furi0us_mat speaking of minetest 5.9, one thing has appeared in my game folders: a certain folder named mod_data has appeared, what's it for? 14:23 MTDiscord it's for the new mod data feature, for cross-world mod-specfic data 14:37 rubenwardy merging #14956 in 5 14:37 ShadowBot https://github.com/minetest/minetest/issues/14956 -- Use JSON file for credits by rubenwardy 15:22 rubenwardy How's glTF going 17:01 MTDiscord will have to look into the CI failure. there's also something about the way irrlicht does materials which we haven't properly taken into account yet. 17:02 MTDiscord we might also want to support something like float frames for gltf (since there you usually have timestamps in seconds rather than frames). 17:16 MTDiscord just checking, but no one has made an issue about mapblock rendering in 5.9 flickering? 17:20 sfan5 first time I hear of it 17:20 MTDiscord https://github.com/minetest/minetest/issues/14999 17:21 MTDiscord i see we have a 5.9.1 milestone, ill let someone else decide if it should be in it 17:22 Desour !title 17:22 ShadowBot minetest 5.9 mapblock flickering depending on camera view · Issue #14999 · minetest/minetest · GitHub 17:22 MTDiscord aha, ive found comments here that are similar, but no one made an issue yet that i can see 17:22 sfan5 > Steps to reproduce see videos above 17:22 sfan5 this is not a reproduction, just a proof 17:23 MTDiscord see the updated comment 17:23 Krock is it reproducible with the Mesa software renderer? 17:24 sfan5 if it happened in singleplayer that'd be helpful 17:24 MTDiscord working on seeing if i can make a minimal reproducable example. ideally with void+adv trains 17:24 Desour it's interesting how the water gets more opaque, as if some faces' tiles were moved 17:25 Desour don't forget to test with default settings 17:25 Krock transparency sorting issue? 17:25 MTDiscord desour: the client was literally just extracted with mtg downloaded for those videos 17:27 Krock cannot reproduce locally 17:30 sfan5 merging #14749, #14842, #14977, #14998, #14905 in 12m 17:30 ShadowBot https://github.com/minetest/minetest/issues/14749 -- [no squash] Split touch controls from UI by okias 17:30 ShadowBot https://github.com/minetest/minetest/issues/14842 -- Translate access denied strings by Emojigit 17:30 ShadowBot https://github.com/minetest/minetest/issues/14977 -- Irrlicht cleanups (mostly getting rid of `core::array`) by sfan5 17:30 ShadowBot https://github.com/minetest/minetest/issues/14998 -- Remove SAO::onAttach() and SAO::onDetach() by lhofhansl 17:30 ShadowBot https://github.com/minetest/minetest/issues/14905 -- Add shared mods path to get_modpaths by j-r 17:38 MTDiscord just to be sure, i compiled the server on stable-5 branch (5.9 releases) and the issue still occurs 17:39 MTDiscord the water only seems to glitch when in a mapblock that also has the advtrains mesh node(s) 17:45 MTDiscord interesting, short of using the same map locally, build some sections of track in single player, same game and advtrains mod and i cant replicate 17:59 MTDiscord disabling shaders appears to solve the issue, at least i cant replicate with them off 18:07 MTDiscord i can provide a trace log level report from debug.txt if anyone wants it, but not seeing any errors in it 18:09 sfan5 btw you can use local map saving to grab a copy of the server map to check locally 18:10 MTDiscord probably faster than me downloading the whole map from the server 18:15 MTDiscord ok, yeah, i can replicate it locally with locally map saving a part of it 18:19 MTDiscord added that to the issue if anyone wants it 18:36 rubenwardy What about turning off post processing? 18:39 MTDiscord it flickered a few times (locally) with it disabled on start up, and then i havent been able to trigger it since 18:40 MTDiscord on the server with post processing off i cant seem to trigger it 18:40 MTDiscord the flickering looks like occlusion culling failure, why would postproc have anything to do with it? hmm 18:40 MTDiscord anyway 18:40 MTDiscord what's the reason why macos 10.14 is required by MTE? 18:41 MTDiscord supposedly devices with intel CPUs (so before M1) can't update to it without some deep patching 18:41 MTDiscord (I'm writing it after reports from some community members on the VL discord) 18:42 sfan5 we don't have a mac maintainer so this question has no answer 18:43 MTDiscord ok, so it looks like post processing off doesnt have an effect, restart the client and can replicate the glitches 18:43 MTDiscord it seems the only "fix" that works so far is to turn shaders off 18:43 sfan5 a cursory google search shows that apple still supports intel macs even with as new as macos 14 18:44 sfan5 10.14 is already long out of support and from 2018. did you confuse it with something else? 18:45 cx384 I want to merge #14983 into master in 15 min. Are there any any objections? 18:45 ShadowBot https://github.com/minetest/minetest/issues/14983 -- [no squash] CI whitespace checks by cx384 19:03 Krock lib/gmp/mini-gmp.c is likely to break again on the next update but whatever ¯\_(ツ)_/¯ 19:03 Krock that is: its formatting. 19:14 MTDiscord it's failing because of trailing whitespace on "doc/lua_api.md:8955:Used by minetest.register_entity. " 🙃 19:16 cx384 I will fix it. 19:18 cx384 I will merge #15001 in 5 min. 19:18 ShadowBot https://github.com/minetest/minetest/issues/15001 -- Fix trailing whitespace from #14945 by cx384 19:18 MTDiscord 👍 feel free to push the trivial fix straight to master. the trailing whitespace was introduced recently in ea96f6e, you probably started your PR before that, hence leading to no CI failure on your PR. 19:18 MTDiscord i'd recommend rebasing on master before merging PRs that are related to CI. 19:19 cx384 I just want to check the CI again with my PR. 19:20 MTDiscord cx384: makes sense 19:20 cx384 Yes sorry about this I didn't think anyone would create more whitespace problems in the meantime ... 20:27 Sokomine help! i'm always getting connection timed out when connecting to a server now. the password check still worked. others are logged in. other servers (with older versions) work 20:42 MTDiscord hurm. So they are on 5.9 and you are on 5.9 or? 20:42 Sokomine the switch to 5.9 was planned but iirc stopped due to connection problems 20:43 Sokomine i now compiled latest git. no chance. not even with an older mt binary from july 20:43 MTDiscord so you are on 5.10 and they are on 5.? 20:43 Sokomine got Irrlicht: SDL_PollEvent took too long: 15ms in verbose 20:44 Sokomine admittedly the server is under high load. lots of monsters, people fighting them right now. an important event going on. so more load than usual. but others report no connection problems 20:45 Sokomine servers' on 5.8.0 and i just compiled latest git. with the version from yesterday i get the timeouts, and even with one from july 20:50 MTDiscord could be something specific to client side changes then 20:51 Sokomine an older binary alone does not help 20:52 MTDiscord wait, you can't connect from like 5.7? 20:54 MTDiscord Sokomine: see #14765 20:54 ShadowBot https://github.com/minetest/minetest/issues/14765 -- New 5.9.0-dev networking is not allowing legitimate players to connect under some rare circumstances 20:55 Sokomine i don't remember when 5.7 was current. i'm copying the entire folder of a year old version now and hope to have luck with that! 21:00 sfan5 as far as I'm aware the problem is with 5.9 on servers, client version not relevant 21:05 Sokomine i've heard that as well. i had trouble logging into that server on 5.8.0 in the pasth weeks/months - often took 3-5 attempts to log in 21:06 Sokomine my connection is still very low bandwidth if that info helps (1 mbit)) 21:06 Sokomine now i get invalid password messages. either i can't type anymore or i mistyped too hoften in previous failed attempts to connect 21:13 sfan5 seems likely to me that the network code does not appropriately respond to congestion and this was not noticed because it ignored packet loss previously 21:17 Sokomine that could be 21:19 Sokomine packet loss is unavoidable in some places. the server spams poor clients with packages due to mods changing the world too frequently in some places. or other not yet discovered bugs. there has to be a reason why server spawns become imberable after a while. i can go to view range 400 almost everywhere - but at server spawn on yl - 20 view range isn't enough for the packages to squeeze through an 21:19 Sokomine 1mbit line. and that's with mesecons forbidden in that area 21:21 Sokomine even back in the old times vanessaes' survival server suffered from that and i could no longer go there. back then it was probably my computer too slow. but that no longer is an issue 21:23 Sokomine in the current case, there's a big fight going on on the server. many of those ~33 players reportedly online will have joined. there are very likely several hundred of mobs_redo stone monsters and the like trying to kill those players 21:24 Sokomine last time i checked movement code meant one package per entity changing direction? quite a lot of packages 21:27 MTDiscord yeah it's a huge amount of packets and operations when you have collisions and big fights, literally working on optimizing that right now for 5.10 fingers crossed it gets in and works flawlessly 21:28 Sokomine aah! good 21:28 Sokomine if you want to see such a thing in action and see why it is important (helps to raise motivation - at least for me - if i see people using my code. perhaps you feel similar :-)): there's a huge fight going on on the your land server. hundreds of entities attacking players 21:34 MTDiscord right, that is what I'm testing, don't worry I just register half my mob monster as NPC and half as Monster and they battle each other. I am working with ~1000 with essentially max health, trying to get that better is my 5.10/5.11 goal 21:35 MTDiscord YL will profit a lot from this snippet and we'll gladly test 😉 21:35 Sokomine good :-) 21:36 Sokomine but by all means feel free to take a look :-) it's probably very chaotic. but there on yl you can see it live now how it's now - with actual players adding to the chaos :-) 21:37 MTDiscord well aware, I'll be pining when this rework is done, red-001 had some good points to improve it.