Time Nick Message 00:32 paradust Did it gradually become this way, or all of a sudden? 00:33 paradust When looking up or down, the FPS goes back up to full. Decreasing the viewing range (using - key) also increases the FPS 00:35 AliasStillTaken We do not have exact measurements. On the testserver, we tried to remove stuff to see if it gets better, but didn't come to a single conclusion 00:36 paradust If your spawn area been getting more complex/expanding, this could be a general performance problem. Minetest's occlusion culling may not be working well enough for your case 00:37 paradust I'll keep poking around and see if I can find anything specific, but it is looking more like there is just too much being rendered simultaneously. 00:38 AliasStillTaken Teh interesting part is taht other cities seem to not have the same problem 00:38 AliasStillTaken They're not less complex though 00:45 AliasStillTaken The testingdl account is yours? 00:45 paradust yea 00:46 paradust it might be a specific feature... I'll see if I can break down the rendering time per block 00:47 AliasStillTaken Is there any way I can help you? On the testserver we can remove mapblocks, copy them or whatever you deem necessary 00:50 paradust Have to figure out what to remove first. I'm adding logging to the code, recompiling, and rejoining the server. 00:50 paradust I might be able to narrow it down that way 00:55 Sokomine that would be great. walking along the main road leads to the client freezing for a second or so - at least for me. things got worse over time but neither gradually nor was there any moment i could pinpoint when it went haywire 00:59 AliasStillTaken Sokomine I need go go for a bit, if paradust joins the testserver, you can give him all privs necessry and show him around. You have all privs there, too 01:18 fluxionary oh, the 1-2 second lag while running down the main road is due to the bazillion entities in the area loading all at once 01:19 fluxionary and by bazillion, i mean 1485-ish 01:21 fluxionary i'm not even sure the issue is even trying to render them in 3d; it's just trying to *initialize* them all 01:26 MTDiscord do some of them have terrible on_activate functions? 01:26 fluxionary it's mostly signs and smartshops 01:27 fluxionary i think signs have to re-initialize the sign text image every time they load? 01:27 fluxionary minetest could really do w/ some actual font support.. but i'm not volunteering to implement that 01:34 paradust Here's the time spent rendering each block: 01:34 paradust https://minetest.dustlabs.io/blocktimes.txt 01:35 paradust unfortunately the tall glass building seems to be taking up most of the time 01:36 paradust two buildings down there's a house with maps. I'm not sure why, but that also seems to be taking a lot of time 01:43 fluxionary the pet shop, with a lot of signs with *tons* of text, seems to be #1? 01:43 ShadowBot https://github.com/minetest/minetest/issues/1 -- GlowStone code by anonymousAwesome 01:44 fluxionary also half the map building 01:44 Pexin tall glass building you say? #11696 01:44 ShadowBot https://github.com/minetest/minetest/issues/11696 -- Depth sorting for node faces by x2048 01:45 fluxionary tall glass building involving a lot of nodebox drawtypes 01:47 Pexin what I'd be tempted to do is temporarily change all the glass blocks to opaque, reboot server, and see how it performs. but that's me. 01:48 fluxionary unfortunately only Alias has the required access for something like that 01:48 AliasStillTaken Can be done on the testserver. Or we could remove and see how we fare FPS wise 01:51 AliasStillTaken use_texture_alpha = true -> use_texture_alpha = "opaque" ? 01:53 paradust I've noticed poor performance on other servers in places with lots of glass. I'll have to do more digging to see what's going on. (I have a suspicion) 01:55 fluxionary Alias: probably the easiest way would be to dirty "builtin/game/register.lua" and change the definitions in `core.register_node` 01:56 AliasStillTaken We can do any of that. Which shall it be? 01:57 fluxionary turning off transparency entirely seems like an interesting experiment. what's the other option? 01:58 Sokomine oh. sorry. i will join the test server and then we can do the changes 01:58 paradust could delete those blocks on the test server 01:58 AliasStillTaken deleting should have an immediate effect. Plus we can paste in again to see if it starts lagging 01:59 fluxionary deleting isn't scientific 01:59 fluxionary we should change just 1 variable at a time. who knows what else goes if we delete the nodes. 01:59 AliasStillTaken 1. remove glass building, see if Sokomine lags less 2. replace use_texture_alpha = true -> use_texture_alpha = "opaque" 3. change builtin/game/register.lua to ??? 01:59 Sokomine change them on the test server to opaque blocks 01:59 fluxionary Alias: do you want to apply 2. to *every* node for *every* mod? 02:00 fluxionary 3. is a very quick way of doing 2., and i don't think 1. tells us anything interesting 02:00 paradust for the sake of core devs, perhaps let's move this conversation to #minetest 02:00 AliasStillTaken +1 02:00 fluxionary +1 02:00 Sokomine i can imagine that glass leads to problems. i wouldn't expect such catastrophic lags from it 06:14 MTDiscord are there any chances of this issue getting a PR 06:14 MTDiscord https://github.com/minetest/minetest/issues/12294 06:19 Pexin #10705 um? also maybe release_bone_position is a better name.. 06:19 ShadowBot https://github.com/minetest/minetest/issues/10705 -- unset_bone_position by theviper121 06:20 MTDiscord This PR is dead though 06:20 MTDiscord its TWO YEARS OLD 06:21 Pexin ¯\_(ツ)_/¯ 06:22 MTDiscord "um aktually 10705" well, maybe if you like- oh I don't know, looked past the name you'd realize that this PR will literally never happen because it's 2 years old and is for an outdated version of MT and therefor would more than likely have to be entirely restarted in the first place 06:23 MTDiscord well why won't you try rebasing it yourself 06:23 MTDiscord because I'm not a game dev 06:23 MTDiscord yes but can you be an engine dev? 06:24 MTDiscord I swear to god if you don't shut up... 06:25 Pexin maybe someone needs a sandwich and a nap! :D 06:26 MTDiscord me!’ 06:26 MTDiscord says the man who linked a 2 year old PR... 06:26 MTDiscord i love sandwishes 06:27 MTDiscord fatalerror was friendly before now hes a madness of a man 06:27 MTDiscord life is grating 06:28 MTDiscord I need more copium 06:28 MTDiscord FatalError: Install Linux, learn how to compile, look at the diffs on the old PR and apply it to the master branch source code. 06:28 MTDiscord copium? 06:28 MTDiscord otherwise you probably should get out of minetest-dev 06:28 Pexin true facts: git is harder than coding 06:29 MTDiscord either way it pisses me off that this has been an issue for YEARS and hasn't been resolved- it's such a simple thing to implement and it's almost useless without its counterpart 06:29 MTDiscord well 06:29 MTDiscord maybe not simple to implement 06:29 MTDiscord does fatal use windows? 06:29 MTDiscord LinoTastic: yeah 06:29 MTDiscord because I don't have the time to learn how to use a whole new OS 06:29 MTDiscord at least it's not mac. 06:30 MTDiscord https://tenor.com/view/linux-you-proprietary-bitch-proprietary-open-source-foss-gif-21858113 06:30 MTDiscord I haven't had any freedom lost 06:30 MTDiscord in fact, with all of the things linux cant have 06:30 MTDiscord I've gained some things 06:30 MTDiscord "gained" 06:30 MTDiscord using linux was painful 06:31 MTDiscord "painful" 06:31 MTDiscord Its a skill, and I don't like learning new ones. 06:31 MTDiscord linux is kernel 06:31 MTDiscord linux is not a os, a gruel in taxes are a os 06:31 MTDiscord wow, congratulations, here's this badge: 06:31 MTDiscord "oh I don't like cooking, I'd rather eat unhealthy precooked meals until I die" 06:32 MTDiscord it says "professional smartass" 06:32 MTDiscord (btw you discord kids really need to understand IRC, neither replies or edits show up on the other side) 06:32 MTDiscord how is this even similar to this situation 06:32 MTDiscord minetest-dev-irc 06:32 MTDiscord are you blind?? 06:33 MTDiscord I like how every time I talk axzorell turns it into a "REEEE YOU DON'T USE LINUX AND YOU HAVE YOUR OWN LIFE THAT ISN'T LIVED JUST LIKE MY OWN REEEEE YOU'RE ENTIRE LIFE IS NULL AND INVALID REEEEEE" 06:33 celeron55 this seems off topic 06:33 MTDiscord yes. 06:33 MTDiscord exactly 06:33 celeron55 so, go elsewhere, now 06:34 MTDiscord okay 06:34 MTDiscord back to my original topic, what are the chances that anything will be done about this issue 06:34 MTDiscord are there any plans for this 06:35 MTDiscord or is it just entirely dead and never going to be fixed 06:35 MTDiscord because this is an outstanding problem with the engine 06:38 celeron55 it seems the implementation doesn't take much code 06:38 celeron55 so doing something about it should be possible 06:42 MTDiscord why the fuck did #10704 get closed 06:42 ShadowBot https://github.com/minetest/minetest/issues/10704 -- Extend bone override capabilities by appgurueu 06:42 MTDiscord oh 06:42 MTDiscord no progress... 06:42 MTDiscord oh right it's app, makes sense. 06:43 celeron55 well, it has a "to do" that's not done 06:44 celeron55 anyway i think 10705 is short and simple. it probably does need a discussion about api design and compatibility 06:45 MTDiscord and if that's done later it may take another 2 years is the problem 06:45 MTDiscord at this rate anyway. 06:46 celeron55 well, it just needs a person who can code C++ and needs this for their game 06:47 MTDiscord I can't imagine somebody going through all the trouble of making a PR, waiting years and hoping for a merge, when they can just use broken hacky solutions 06:47 MTDiscord if they have the patience to wait a year, they could just make their own engine/game 06:48 Pexin oh funny joke time! 06:48 MTDiscord what 06:49 MTDiscord you shouldn't be so rude to the developers, this system isn't a joke, it's just... challenged 06:49 MTDiscord kek 06:49 Pexin making a full game in a year 06:49 celeron55 there are better and worse times in PR handling 06:50 MTDiscord Pexin, if they're making a minetest game, they probably aren't trying to make anything super high quality. So I can see that happening quite easily 06:55 celeron55 i guess you could currently argue that unset_bone_position falls within the current medium term roadmap https://github.com/minetest/minetest/blob/master/doc/direction.md in which case it should go through fairly easily if a PR is opened now 06:55 celeron55 as we have "2.4 Object and entity improvements" in there 06:56 celeron55 but also someone could say it doesn't 06:56 MTDiscord This is why we need somebody actually in charge of the project 06:57 MTDiscord or like a board of select people in charge 06:57 MTDiscord that way decisions like this can actually be made 06:57 celeron55 eh 06:58 celeron55 any coredev can decide to review and merge the PR 06:58 celeron55 if it was even fewer people, then even less would happen 06:59 celeron55 and if you're arguing the board of select people wouldn't be C++ programmers then no 07:00 celeron55 anyway if you want something to happen, the most straightforward way to do it is to do it yourself 07:03 celeron55 anyway, 10705 should probably be added to the agenda of the next dev meeting 07:04 celeron55 the only valid argument against it in my opinion is if someone has greater goals regarding to bones and doesn't want to have a partway solution includeed 07:04 celeron55 -e 07:04 celeron55 (and that could be what happened previously) 07:06 MTDiscord I mean, it beats nothing 07:06 MTDiscord I dont think there's any issues with it, but id especially like to see scaling in set_bone_pos personally 07:07 MTDiscord Thats a seperate issue though 07:08 celeron55 if its design doesn't impede future improvements, again, i don't see why it's left hanging in there 07:09 MTDiscord Indeed 07:13 sfan5 tall glass building you say? #11696 07:13 ShadowBot https://github.com/minetest/minetest/issues/11696 -- Depth sorting for node faces by x2048 07:13 sfan5 you could disable that and see if it's better 07:14 sfan5 what I'd be tempted to do is temporarily change all the glass blocks to opaque, reboot server, and see how it performs. but that's me. 07:16 sfan5 you can do that by modifying the client code too, but if you don't do it properly you risk having glass that is opaque but still has the performance problem 07:16 sfan5 so I guess changing the nodedef is a better test 07:17 MTDiscord Wait did the bot break 07:17 MTDiscord Its saying you're sfan lol 07:21 celeron55 sfan5 is quoting Pexin, you silly discord kids 07:23 MTDiscord Uhm 07:23 MTDiscord Uh 07:23 MTDiscord Im not stupid... just speshal 07:45 Zughy[m] 10705 will have been closed by tonight with an "Adoption needed" label 13:51 MTDiscord celeron55: you can add #12388 too 13:51 ShadowBot https://github.com/minetest/minetest/issues/12388 -- Extend bone override capabilities by appgurueu 13:51 MTDiscord @FatalError app is back in town 19:14 MTDiscord Halo monk chanting 19:52 x2048 Sokomine: did you manage to solve the problem with the spawn and the tall glass building on your-land? 19:54 x2048 For transparency sorting there is a setting `transparency_sorting_distance`, changing it to 0 disables any sorting during rendering and impact on the FPS 20:14 Sokomine x2048: sadly not yet. after a lot was changed - glass no longer transparent (didn't change much), entities removed and metadata erased, it got better and at least the 1-2 seconds freeze went away. fps still were pretty low 20:16 Sokomine i still suspect that it might be similar to the technic gold chests with the petz on tunneler's abyss: something sending the same blocks again and again without any real change and my machine struggling to decompress and update. but i am not sure 20:16 Sokomine others with weaker devices have similar problems 21:23 erle sfan5, you are right about https://github.com/minetest/minetest/pull/12389 that creating a vector with the specified values does nothing – but IIRC putting it on a stack for a function does something, namely, crash the game. 21:26 erle IMO this is still a typing error in the end, positions that take or emit player positions do not exactly agree what a position looks like. 21:27 erle i.e. once those are *exactly* aligned, there should be no crash (according to my understanding of the code) 21:28 erle well, not for player positions only, for vectors in general. 22:55 Zughy[m] Ladies and gentlemen, I hereby inform you that Minetest has now reached less than 100 PRs, after a very very long time. 90 to be precise 22:56 Zughy[m] LET'S GO, PEOPLE 22:56 * Zughy[m] uploaded an image: (10KiB) < https://libera.ems.host/_matrix/media/r0/download/matrix.org/TZCUvrdQoUQgLMarPxWfgUtb/image.png > 23:01 MTDiscord this could either be a very bad thing or a very good thing 23:04 Zughy[m] See you in about 10 days for hopefully <1000 issues 23:16 MTDiscord oh god 23:16 MTDiscord please dont tell me you're purging everything 23:17 MTDiscord "too many issues, lets just get rid of them and pretend they dont exist" 23:23 Zughy[m] I'm not purging anything, just filter for "Possible close" and see how there are 30+ of them, against 1020 issues