Minetest logo

IRC log for #minetest-dev, 2017-02-17

| Channels | #minetest-dev index | Today | | Google Search | Plaintext

All times shown according to UTC.

Time Nick Message
00:01 ircSparky joined #minetest-dev
00:11 ircSparky_ joined #minetest-dev
00:16 paramat i forgot to say, i could not merge #5234 , did not apply. sofar
00:16 ShadowBot https://github.com/minetest/minetest/issues/5234 -- Re-add Droid Sans as fallback font (Reverts part of eb490) by sofar
00:17 sofar yeah, that's git for ya
00:17 sofar I'll do it myself later tonight
00:17 paramat i was using 'git am'
00:17 paramat ok
00:17 sofar yup, git am b0rks on that one
00:47 DonBatman joined #minetest-dev
01:12 ircSparky_ joined #minetest-dev
01:52 sofar how do I measure mapgen time performance wise?
01:52 sofar is there such a way?
01:54 VanessaE dtime?
01:54 VanessaE in register-on-generated
01:54 VanessaE probably isn't accurate enough though
01:54 VanessaE but that's what I use in biome_lib to split-up the mapgen stuff into small steps
02:37 paramat joined #minetest-dev
02:39 paramat sofar in a mapgen uncomment these lines https://github.com/minetest/minetest/blob/master/src/mapgen_v7.cpp#L233 https://github.com/minetest/minetest/blob/master/src/mapgen_v7.cpp#L275 https://github.com/minetest/minetest/blob/master/src/mapgen_v7.cpp#L31
02:41 paramat note that doesn't include liquid update and light calculation, so i like to move the second line to after those
03:41 kaeza joined #minetest-dev
03:51 deepbook5broo joined #minetest-dev
03:51 deepbook5broo left #minetest-dev
04:05 sofar maybe I just want something that generates a known amount of mapchunks (large number) and times it
04:21 Hunterz joined #minetest-dev
05:58 xerox123 joined #minetest-dev
06:01 Hunterz joined #minetest-dev
06:50 Player_2 joined #minetest-dev
07:07 xerox123 joined #minetest-dev
08:09 red-001 joined #minetest-dev
08:18 Darcidride joined #minetest-dev
08:56 Gael-de-Sailly joined #minetest-dev
09:21 kaeza lol running luacheck on builtin
09:21 nerzhul joined #minetest-dev
09:22 kaeza even disabling "safe" warnings (whitespace/shadowing/unused stuff), it's like 5 pages of warnings
09:22 kaeza it wants to dump everything in _G
09:26 blaze joined #minetest-dev
09:27 nore kaeza: well, you should tell it about the known globals (core, ItemStack, etc.)
09:27 kaeza already did :P
09:27 nore hmm
09:28 nore what it is complaining about then?
09:28 kaeza https://gist.github.com/kaeza/8c6bbda05ed1d19234874fe69656bdca
09:29 kaeza most of those look fairly safe
09:30 nore ow
09:30 nore yeah, would be good to fix that one way or the other though
09:31 kaeza and that one's not including builtin/mainmenu
09:32 kaeza I can go through the code and clean up stuff, but is it worth the effort?
09:33 kaeza i.e. anybody except me and maybe ShadowNinja cares?
09:34 nore I think it's worth it
09:37 Darcidride joined #minetest-dev
10:18 exio4 builtin/game/item.lua:193:1: inconsistent indentation (SPACE followed by TAB)
10:21 * kaeza hugs exio4
10:21 kaeza okay, I have a few questions
10:22 kaeza first, is basic_dump meant to be global? doesn't look so, looks like it is a helper for dump and dump2
10:22 kaeza and second, why is mainmenu-only code in builtin/common ? O.o
10:57 kaeza #5253
10:57 ShadowBot https://github.com/minetest/minetest/issues/5253 -- [WIP] Run builtin through LuaCheck and fix warnings. by kaeza
11:05 proller joined #minetest-dev
11:06 DonBatman joined #minetest-dev
11:08 TheReaperKing2 joined #minetest-dev
11:08 kilbith joined #minetest-dev
11:12 proller joined #minetest-dev
11:19 red-002 joined #minetest-dev
11:19 red-002 joined #minetest-dev
11:20 red-002 #5252
11:20 ShadowBot https://github.com/minetest/minetest/issues/5252 -- Add support for the new arguments of `request_shutdown` to the `/shutdown` chatcommand. by red-001
11:38 Fixer joined #minetest-dev
11:40 crazyR joined #minetest-dev
11:48 shymega joined #minetest-dev
12:23 xerox123 joined #minetest-dev
12:26 Fixer joined #minetest-dev
12:44 betterthanyou711 joined #minetest-dev
12:48 betterthanyou711 joined #minetest-dev
12:49 betterthanyou711 joined #minetest-dev
12:49 betterthanyou711 joined #minetest-dev
12:52 betterthanyou711 joined #minetest-dev
13:09 xerox123 joined #minetest-dev
13:14 burli joined #minetest-dev
13:19 glut32 joined #minetest-dev
14:11 ircSparky joined #minetest-dev
14:14 Hunterz joined #minetest-dev
14:24 red-002 Did I misunderstand #244 or are players floating in mid-air after they die a feature not a bug?
14:24 ShadowBot https://github.com/minetest/minetest/issues/244 -- Strange effects with dead players
14:26 lisac joined #minetest-dev
14:45 VanessaE Interesting effect I never noticed before:  if I happen to be "just standing there" after signing on, and the more distant parts of the map haven't loaded because they're too far, eventually they do start to load as night falls.
14:45 VanessaE (even if it's been several minutes since I signed on)
14:47 Taose joined #minetest-dev
14:47 Taose left #minetest-dev
15:06 Zeno` VanessaE, what are you smoking and can I have some?
15:06 VanessaE I'm serious.
15:06 VanessaE and if I had it, I'd share ;)
15:06 Zeno` lol
15:06 rubenwardy joined #minetest-dev
15:10 garywhite joined #minetest-dev
15:12 octacian joined #minetest-dev
15:12 octacian joined #minetest-dev
15:13 VanessaE on another note, is there a "sane" way to scan a world, from outside of Minetest, for bad/corrupt blocks?
15:15 sfan5 VanessaE: i guess my mapblock-parse.py script should returns errors if you have a corrupt block
15:16 sfan5 i would require you to invoke it <number of mapblocks> times though
15:17 VanessaE just curious.   rogier5's minetestmapper returns tons of "incorrect header check, corrupt?" errors, but in-world everything's normal.
15:18 VanessaE figured a mass check-for-errors, like a map-level fsck, might be a good idea
15:18 VanessaE (if such a thing existed)
15:19 sfan5 >Sorry, forked repositories are not currently searchable.
15:19 sfan5 for fucks sake github
15:20 VanessaE wayt
15:20 VanessaE wat
15:20 sfan5 this incorrect header msg doesn't seem to be part of rogiers code
15:20 sfan5 most likely returned by zlib
15:21 VanessaE ah, could mne
15:21 VanessaE be*
15:21 sfan5 are you running -dev?
15:21 VanessaE lemme check that even have zlib installed
15:21 sfan5 if so then minetestmapper needs changes to function correctly
15:21 sfan5 you do
15:21 VanessaE you mean -dev as in minetest engine?  yes.
15:22 diemartin joined #minetest-dev
15:22 sfan5 yes i meant that
15:23 VanessaE which of us gets to file an issue against his repo? :)
15:23 sfan5 not me
15:23 VanessaE (I'd just use upstream, but the leaflet mapper needs a couple of features only rogier5's fork has, I guess)
15:24 sfan5 which ones?
15:24 VanessaE that's just it, I don't remember for sure
15:24 VanessaE est31 would know
15:25 VanessaE what sort of changes does it need?  I'll mention this in the bug report.
15:25 sfan5 $mapperpath ${MAPPERPARAMS} -i ${MAPDIR} --geometry ${posx},${posy}+${tilesize}+${tilesize} -o ${tiledir}/20/map_${x}_${y}.png
15:26 sfan5 i don't see anything
15:26 VanessaE http://daconcepts.com/vanessa/hobbies/minetest/Server-scripts/update-server-maps-leaftest.sh
15:27 est31 joined #minetest-dev
15:28 VanessaE ah there's the man of the hour now :)
15:29 sfan5 i still don't see anything incompatible
15:29 VanessaE est31: why did I need rogier5's mapper for leaflet?
15:32 sfan5 also your script is copying colors.txt needlessy
15:32 VanessaE oh?
15:32 red-002 #5255
15:32 ShadowBot https://github.com/minetest/minetest/issues/5255 -- Add option to also check the center to `find_node_near` by red-001
15:33 sfan5 VanessaE: it does $ cp /home/vanessa/www/vanessa/hobbies/minetest/colors.txt $workdir/leaftest   but then passes the original colors path to mapper.sh
15:33 VanessaE oh heh
15:33 VanessaE good catch
15:34 VanessaE can you give me a brief description of the "changes" the mapper would need, and maybe the commit that made it necessary?  I'd like to include those in this bug report.
15:35 sfan5 https://github.com/minetest/minetest/commit/f17c9c45dc30a388675d46418d278a4a029206e2 -->>https://github.com/minetest/minetestmapper/commit/21444d00c35dc6896ffc27af7da38d8a296178fd
15:37 VanessaE thanks
15:40 VanessaE https://github.com/Rogier-5/minetest-mapper-cpp/issues/19
15:42 VanessaE still racking my brain trying to remember why rogier5's mapper was needed for leaftest *pokes est31 again :) *
15:42 Fixer more about stutter/jitter, I loaded up old minetest files from 0.2.2011 era
15:42 Fixer Minetest 0.2.20110618_0 does not have visible stutter at all on my PC, yet 0.2.20110704_0 version does
15:43 VanessaE Fixer: did you control for things like CPU spikes outside of minetest, e.g. from a browser or screen update elsewhere?
15:44 Fixer VanessaE: cpu was quit, i played those some time, I'm pretty sure about this, stutter was visible in 20110704_0
15:44 Fixer quite
15:45 paramat joined #minetest-dev
15:45 VanessaE Fixer: ok.
15:56 Taoki joined #minetest-dev
16:00 Gael-de-Sailly joined #minetest-dev
16:04 Fixer i'm looking at this part right now: https://github.com/minetest/minetest/commit/02006a95402c41833779e3825c44b39d600e566b
16:04 Fixer dat sleep
16:04 Fixer wanna try it on newest to slow down it
16:10 rubenwardy set it to 1000
16:10 rubenwardy wait, is that ms or s?
16:10 rubenwardy ms surely
16:11 rubenwardy oh, misread, I'm an idiot
16:11 Fixer ms
16:12 Fixer want to experiment slightly
16:13 paramat maybe you could bisect to find the commit that causes jitter
16:13 Fixer i doubt I will compile 2011 build succesfully today with mingw64
16:14 Fixer and I think I put sleep in a wrong place %) but thats ok, will try more
16:16 nore Fixer: just decide it is bisect good if it doesn't compile, and then see if the commit you get is the first that compiles or if it causes jitter
16:19 Fixer I was looking at this part https://github.com/minetest/minetest/commits/81535f6277b66c869cda7125b2aeaf99e8078b92/src/client.cpp (since Jun18 and up)
16:19 celeron55 Fixer: you don't need to experiment with that because we all already know your jitter solely depends on the number of mapblock meshes being updated per frame
16:19 Fixer celeron55: exactly
16:19 celeron55 and increasing that sleep obviously slows down the updates
16:24 red-002 updated #5255 it should be more readable now
16:24 ShadowBot https://github.com/minetest/minetest/issues/5255 -- Add option to also check the center to `find_node_near` by red-001
16:34 Fixer and indeed, sleep(50ms) eliminated the stutter
16:34 Fixer totally
16:35 VanessaE I'm curious why a sleep is even needed in the first place
16:35 VanessaE shouldn't meshes be sent in a separate thread as fast as possible?
16:36 Fixer http://pastebin.com/raw/D44biexY
16:36 VanessaE oh. they are. :)
16:37 VanessaE (still doesn't explain why a sleep is used.  thread priority problems maybe?)
16:37 Fixer VanessaE: it was used like this in mid of 2011 and then refactored to current look I think
16:41 Fixer https://i.imgur.com/qWEoGdH.png
16:42 VanessaE you really wanna see stutter, load HDX256 and go over to VE-S around the spawn
16:42 VanessaE you'll see pauses as long as a couple of seconds.
16:43 Fixer now, lets speed it up to 25ms
16:43 Fixer or 20...
16:47 Fixer sl20 here i go
16:48 Fixer very good
16:48 Fixer acceptable
16:49 sfan5 the correct way to solve this is probably to add a setting
16:51 Fixer lazy chunk load enabled by default with some pretty good sleep value
16:52 Fixer i see some rare bump of draw time to 0.05 rarely from different source, apart from that it is damn smooth in huge jungle
16:52 Fixer lets try 10 ms
16:53 VanessaE Fixer: please go try VE-S with HDX256
16:53 VanessaE I'm curious how it performs
16:54 Fixer VanessaE: i got that 0.7 sec freeze in loop_other, even with default textures, hard to imagine HDX256, i can join to see it, but with HDX128
16:55 VanessaE that's fine
16:55 Fixer VanessaE: on default minetest, have not tested with sleep
16:55 red-002 #5254 doesn't really seem like something that should be on a bug tracker
16:55 ShadowBot https://github.com/minetest/minetest/issues/5254 -- This code should fill a mapchunk
16:56 paramat looking
16:56 paramat is there a disadvantage to increasing that sleep time slightly, finding a good balance?
16:57 Fixer paramat: decreasing sleep time will increase stutter, and vice versa (versa?)
16:57 paramat 5254 is code error, will reply
16:57 Fixer currently trying 10ms
16:58 nore hmm, I think either a setting or a way to automagically adapt the sleep time would be the correct solutions
16:58 Warr1024 joined #minetest-dev
17:00 Fixer hmmm, 10ms is close to edge of my liking
17:00 Fixer feels somewhat agreesive at 10ms, still pretty smooth
17:01 VanessaE Fixer: aggressive?
17:01 VanessaE surely no stutter is the ideal?
17:01 Fixer VanessaE: somewhat more jitter at 10 ms
17:01 VanessaE oh
17:02 Fixer VanessaE: like this part: https://i.imgur.com/AmQfEyl.png it is not critical
17:02 Fixer 0.037
17:02 Fixer avg is ~0.01
17:02 Fixer but even 10ms is quite good
17:02 Fixer as compromise between speed and jitter
17:03 Fixer hmmm, i feel like 15 ms sleep test is needed
17:04 Fixer 10 ms sometimes gives some jitter I dislike (like shallow fps drop from 100 to 70fps)
17:04 Fixer and back to 100
17:05 Fixer VanessaE: i will join your server with sleep 15
17:05 VanessaE ok
17:05 Fixer dreambuilder survival
17:05 VanessaE yeah
17:05 VanessaE at the spawn is the worst of course
17:07 sfan5 is this trampoline thing a feature? https://a.uguu.se/N2SgPMiYWgbE.mp4
17:07 Fixer yes, 10ms is harsh, preferring 15ms
17:07 paramat i notice it's worse over dense forest than empty terrain, i guess because the meshes are more complex
17:08 Fixer paramat: yes, but sleep eliminates even that
17:08 Fixer it is very weak after that
17:08 Fixer like you loose 1-2 fps at most with sleep 50ms
17:08 Fixer when it is 100fps
17:08 paramat that would be worth it for me
17:09 paramat jitter is really bad recently, hmmmm noticed too
17:09 paramat the denser, taller forests of mgv7 probably make it worse
17:10 paramat plus all those snow nodeboxes on pines
17:10 Fixer sweet spot near sleep 15-20ms on my PC
17:10 nerzhul joined #minetest-dev
17:10 Fixer very good in forest
17:10 paramat i might test this ..
17:10 Fixer map loading will be somewhat slower, but acceptable
17:10 Fixer i can still run
17:12 red-002 should #5218 be a warning?
17:12 ShadowBot https://github.com/minetest/minetest/issues/5218 -- "ERROR[Server]: Pathfinder: failed to update cost map" floods console
17:12 red-002 or do you think it should stay an error?
17:12 paramat unless there's some big disadvantage i would support a new balanced value
17:14 * VanessaE watches Fixer's PC begin to smoke
17:15 Fixer sfan5: trampoline should remine everyone that minetest also has UGLY unfixed falling related bugs
17:16 Fixer VanessaE: 40fps at spawn, there was freeze, but it was as usual of unrelated loop_other origin
17:16 VanessaE 40 fps?!
17:16 VanessaE how the fuck?
17:16 VanessaE I can't even maintain 30
17:16 Fixer 30fps
17:16 paramat falling issues have been discussed and are not easy to fix
17:16 Fixer VanessaE: it loads slightly slower
17:17 Fixer VanessaE: let me do a walk
17:18 Fixer whats up with mainloop_other 1.6 sec on vanessa servers only?
17:18 VanessaE idk
17:19 VanessaE signs maybe
17:19 VanessaE (entities that is)
17:19 Fixer i seen this only on your servers... including basic one
17:20 VanessaE hopefully that text-on-signs idea pans out (and can replace signs_lib)
17:21 Fixer VanessaE: 20-30 fps on spawn, what is good, is it is pretty solid fps, not quick changes, only unrelated rare freezes due to signs or whatever, will upload gif
17:21 VanessaE ok
17:22 ge joined #minetest-dev
17:24 Fixer VanessaE: notice how smooth it is on your server https://imgur.com/a/9qd0T
17:24 VanessaE far smoother than it is for me, for sure
17:25 Fixer fps may be low, but at least it is consistent without insane jitter
17:27 Fixer num_processed_meshes was at 5 on that server
17:27 Fixer pretty good value
17:28 Fixer now if I try default client... that will be disaster %)
17:29 Fixer yesterdays gif from same place, probably shorter https://imgur.com/a/bWJoe
17:29 VanessaE ew
17:29 VanessaE that's almost as bad as it is for me
17:30 Fixer yes, and compare with almost smooth line for sleep 15ms
17:31 Fixer it just feels better, i don't want to throw up anymore, it is just low fps, but without this jitter shit on top
17:37 paramat i might make a PR and we can then test to find a good balanced value
17:38 VanessaE go with sfan5's idea, make it configurable
17:38 Fixer i already tested 10-15-20-50ms sleeps, will spent some time with 15ms one, which I feel is quite good balance
17:38 VanessaE (with a sane defautl)
17:38 VanessaE default*
17:38 paramat just remembered you can do PRs now
17:39 Fixer nah, i'm noob with this
17:39 Fixer it needs more code
17:39 Fixer to check settings value and apply it to sleep
17:39 paramat why more?
17:39 paramat ok
17:39 Fixer do you want it configurable?
17:39 paramat having a setting is quite easy to code
17:40 paramat erm maybe
17:40 Fixer lets say we want 15ms, add setting with default 15ms, and in code it just checks that settings and uses for sleep :)
17:40 Fixer people who want agressive map load can easily decrease it down to 1 ms...
17:40 Fixer or 0 for disabled
17:41 paramat it's important so a setting is probably needed
17:42 sfan5 different idea: have a setting agressive_map_loading, defaults to false; in the code have sleep_ms(g_settings->get_bool("agressive_map_loading") ? 1 : 25);
17:42 sfan5 then the setting is more user friendly and thus be added to be settings menu
17:44 Fixer i hope this checking will not slowdown anything
17:45 Fixer it will be cached settings value, right?
17:46 rubenwardy joined #minetest-dev
17:46 Fixer feels sooo good on servers too
17:46 VanessaE sfan5: 1ms?? way too low
17:46 VanessaE maybe you want 15 and 100 ?
17:46 VanessaE (or whatever the higher value was)
17:46 Fixer VanessaE: 1ms - most agressive, 100ms - slowgen
17:47 VanessaE it's better to just allow a custom value imho
17:47 VanessaE Fixer: yeah but you said that for you, below ~10ms is too aggressive
17:47 Fixer VanessaE: yes
17:47 sfan5 well currently the sleep_ms() doesn't seem to exist at all
17:47 Fixer VanessaE: and default minetest don't have this sleep in the first place %)
17:47 Fixer VanessaE: so it is nearly as agressive as default
17:47 VanessaE ohhhhh
17:47 sfan5 and i don't have any jitter problems
17:48 VanessaE sfan5: go try VE-S then.
17:48 VanessaE you will have bad jitter problems there :)
17:48 Fixer sfan5: with sleep or without?
17:48 Fixer i have jitter since 2011 basically
17:48 Fixer on two different computers
17:48 sfan5 without
17:49 sfan5 where did you add the sleep anyway
17:49 Fixer P4 2.8GHz + R9600 when I played on 0.2.2011/0.3/0.4dev and today with core i3 + R6870
17:49 sfan5 VanessaE: you mean dreambuilder survival?
17:49 * VanessaE pokes Zeno`
17:49 VanessaE sfan5: yeaH
17:49 Fixer sfan5: http://pastebin.com/raw/D44biexY
17:50 VanessaE (zeno had some stuff to say a while back about jitter/stutter and mesh gen lag)
17:50 sfan5 Fixer: that's not exactly the same as with the old code though
17:50 Fixer yeah, but helps %)
17:51 sfan5 VanessaE: where do i look for jitter
17:51 VanessaE near the spawn
17:51 VanessaE just walk say toward the tunnel
17:51 Fixer really big freezes don't count, they are of different origin
17:53 Fixer example of worst kind, 0.4.13 on discontinued just test server: http://i.imgur.com/F4jOrj3.png
17:53 sfan5 fps is pretty variable
17:53 sfan5 frametime increases if mesh updates happen
17:53 sfan5 this has been known since forever
17:54 Fixer correct
17:54 Fixer sleep just spreads it out more nicely, in lesser chunks
17:54 Fixer this eliminates ingame jitter
17:55 celeron55 but then the client is incapable of processing lots of meshes
17:55 celeron55 it should be spread out, not limited
17:55 sfan5 the problem here is that "aggressive" mesh updates presents no problem at all in many cases
17:56 celeron55 i.e. a dynamic limit
17:56 sfan5 so it would be a bad idea to just add a sleep_ms thats enabled by default
18:01 paramat spread out seems better
18:04 rubenwardy sofar: https://gist.github.com/rubenwardy/5ebfc6e6a15976118d487a1a51f9601e
18:04 paramat #5258 fixed some more world-edge crashes
18:04 ShadowBot https://github.com/minetest/minetest/issues/5258 -- Voxelmanip: Do not emerge or blit to blocks over map gen limit by paramat
18:05 rubenwardy likely related to font_path = fonts/liberationsans.ttf in minetest.conf
18:05 rubenwardy it shouldn't segfault
18:10 sofar rubenwardy: fonts and 'run_in_place' do not get along at all
18:10 sfan5 that's not the problem here
18:10 rubenwardy :D
18:10 sofar I had the same crashes
18:10 sofar oh, hmm, I understand
18:10 rubenwardy as a general rule, Minetest should never segfault
18:11 rubenwardy :P
18:11 sofar welp, I gotta go to work
18:11 * sofar afk
18:11 nerzhul no program should segfault, only assertions in debug builds are  acceptable
18:15 Fixer sfan5: no idea in what cases, any mesh update that is more than say 5 gives jitter for me, and usually in singleplayer it goes as high as 40 :(
18:16 Fixer same for vanessa server
18:16 Fixer with limiter it is much better experience
18:16 Krock joined #minetest-dev
18:16 Krock joined #minetest-dev
18:16 Fixer map still loads pretty fast, yet smoother experience without jitter
18:21 Fixer if you want to spread it... there still must be limit to not affect fps much, like 2-3(at most) mesh updates per frame
18:25 Fixer it is fraps time
18:32 Fixer joined #minetest-dev
18:35 VanessaE just as a reminder, someone please fix the screwdriver. https://github.com/minetest/minetest_game/issues/1567
18:37 YuGiOhJCJ joined #minetest-dev
18:42 red-002 rubenwardy, does #5255 look ok to you now?
18:42 ShadowBot https://github.com/minetest/minetest/issues/5255 -- Add option to also check the center to `find_node_near` by red-001
18:46 rubenwardy yeah, although ){
18:48 red-002 should I fix the code style in the function?
18:52 garywhite joined #minetest-dev
18:52 Fixer my fly test ended up with much higher fps due to slightly slower map load, results: https://i.imgur.com/F8Hqpwk.png Look standard deviation and frame distribution, huge difference in jitter
18:52 red-002 done
18:55 rubenwardy someone knows their statistics
18:57 garywhite Someone suggested that a feature be added to export the chat or F10 to a file. Is this possible? (I don't think so, cause then it would have to be added to the engine to save it to a file by default
18:57 Fixer some hardware sites also started measuring frame times, not fps, to uncover various problems, like jitter/etc
19:12 paramat joined #minetest-dev
19:32 celeron55 so now loading less map is better? lol
19:32 Fixer it is acceptable
19:35 paramat i guess it is not just a case of number of meshes processed per frame but how complex they are
19:38 red-002 #5071
19:38 ShadowBot https://github.com/minetest/minetest/issues/5071 -- Rename functions in the minetest namespace to be more consistent. by red-001
19:38 Fixer if someone have better idea that eliminates jitter, please do, meanwhile i;m sticking with sleep 15 as standard for my buids
19:39 rubenwardy if it makes a difference, then it may be worth it as a hot fix
19:39 rubenwardy leave better load balancing for latter
19:39 rubenwardy although, for later usually means never
19:39 Fixer it is still faster than +fast on ground on my hardware
19:40 Fixer in terms of mapblocks load speed
20:09 DonBatman joined #minetest-dev
20:30 celeron55 sleeping a different amount makes zero sense; there are much better ways to throttle the mesh generation rate
20:31 Fixer could be
20:31 Fixer i'm sticking with sleep 15, who knows how many decades will pass when someone comes up with a solution
20:33 celeron55 https://github.com/minetest/minetest/blob/master/src/client.cpp#L535
20:33 celeron55 this is where you should throttle it
20:33 celeron55 limit the number of iterations that loop makes
20:34 celeron55 it'll make it even smoother than you can get it with sleeps
20:34 celeron55 for example, just replace while -> if
20:34 celeron55 you'll get exactly one mesh per frame
20:35 celeron55 a limit for that could be made a configuration option i guess
20:40 sfan5 i remember someone complaining about the design https://github.com/minetest/master-server/pull/15
20:40 sfan5 or not whatever
20:40 sfan5 ^ Fixer you sound like you might have time to look at this
20:41 Fixer sfan5: about design?
20:41 sfan5 yes
20:45 Fixer sfan5: that was probably not me, but i have problems with last column - it is hard to understand what exactly are those numbers for ping, lag
20:45 paramat c55 thanks
20:45 sfan5 Fixer: i wasnt trying to imply that it was you, just needed someone to look at it
20:45 Fixer sfan5: after comma digits, few persons may check server uptime to see when there will be reset?
20:45 sfan5 heres how it looks btw https://kitsunemimi.pw/tmp/sldesign/
20:46 Fixer hmm
20:46 Fixer second column needs improvement
20:47 sfan5 not aligned properly?
20:47 Fixer https://i.imgur.com/JC3g6mO.png
20:47 paramat c55 i might work on this
20:47 Fixer sfan5: maybe make it players/max and below it average/top so it matches the pattern below
20:48 Fixer sfan5: this looks crappy https://i.imgur.com/76h0SdB.png
20:48 sfan5 oh
20:48 sfan5 your screen/window is too narrow
20:48 sfan5 can't do anything about this
20:49 Fixer sfan5: 1280*1024
20:49 sfan5 the old design would look similarily bad
20:49 Fixer sfan5: old design looks better on my screen, https://i.imgur.com/1CGKC9m.png
20:49 Fixer sfan5: for that particular part
20:50 Fixer on my LCD*
20:50 Krock age is wrapped to a new line here
20:50 Krock (with old style)
20:50 Fixer hhmmmm, actually i can resize it in firefox to 1600*1200 as example, to see how it looks
20:50 sfan5 accomodating for different sizes while keeping the nice alignment is pretty hard
20:50 Fixer web dev tool
20:50 Krock it's just too much information for narrow screens. hard to make it fit well
20:51 Fixer ok, i'm looking at 1600*1200 version
20:51 Fixer it looks neat
20:52 Fixer sfan5: on 1600x1200/1920*1080/1920*1280/etc it should look nice
20:52 sfan5 it does look nice on 1080p
20:52 sfan5 since that is what i dev'ed it on
20:53 Krock 1280*1024 screen: http://i.imgur.com/nru6iCf.png
20:53 Fixer also nice for 1366*768
20:54 Fixer below 1364 it starts to look ugly with second rows...
20:54 Fixer sfan5: not a lot of people have 1280*1024 tbh
20:54 sfan5 yeah most people have 16:9
20:54 sfan5 i'll let the PR sit 1 or 2 days and then merge it
20:55 sfan5 (since SN probably won't come around to comment on it anyway)
20:57 Fixer no wait, now it starts second row on 1366
20:58 sfan5 its quite tolerateable on phones too
20:59 sfan5 celeron55: isn't doUpdate() countinously called by the thread while the queue is not empty?
21:00 paramat nore sfan5 rubenwardy when you can please have a look at game#1552 , good new leafdecay system we have been working on
21:00 ShadowBot https://github.com/minetest/minetest_game/issues/1552 -- Leafdecay: node timer based implementation, API. by sofar
21:00 sfan5 does the leafdecay abm really take that much processing power?
21:00 Fixer sfan5: i think you should trim it to 1366*768, sometimes it starts second row on 1366px...
21:01 sfan5 Fixer: the table layout is auto decided by your browser
21:01 sfan5 i haven't set any specific limits in the css
21:01 VanessaE sfan5: it can take a lot if it has a shitton of leaves to process i.e. moretrees
21:02 Krock Fixer, what if the font size would be reduced by one pt?
21:03 Fixer sfan5: if you have firefox you can go into dev panel and type "resize to 1366 768"
21:04 Fixer shift+f2
21:05 Fixer serverlist looks cleaner now
21:05 sfan5 i mean i could (try to) fix the size at 1366/768 but that would be a rather bad idea
21:06 paramat it's the most intensive ABM, and of course for the vast majority of the time it's running for no reason
21:07 paramat it only needs to be active when changes are being made to trunks
21:10 paramat we recently removed the leafdecay trunk cache which was eating huge amounts of memory, so then the leafdecay ABM became more intensive and it had to be slowed down because of that
21:15 Fixer sfan5: this make confuse https://i.imgur.com/VhfwvYl.png
21:16 Fixer make=can
21:16 rubenwardy I'd per separate columns for those things
21:17 rubenwardy maybe with colspan for headers, if needed
21:17 rubenwardy just so they're all aligned
21:17 sfan5 Fixer: indeed, it kinda sucks that those two need to be in a signle columns
21:17 sfan5 s/s$//
21:24 Fixer Calinou can also look at that I guess
21:25 rubenwardy paramat, reviewed
21:28 Fixer celeron55: with if method, it seems I have more jitter
21:29 Fixer compared to sleep 15
21:32 Fixer will try to benchmark it
21:37 paramat ok
21:40 Fixer jitter indeed pretty close to unfixed one with IF method, will try again
21:40 Fixer also visible by eye compared to sleep 15
21:40 Fixer map loading is faster
21:41 Fixer but jitter ... :(
21:41 Fixer one more try
21:47 Fixer joined #minetest-dev
21:51 Fixer with if (!m_mesh_update_thread.m_queue_out.empty()) jitter is slightly better than original game (has visible stutter), much worse than sleep 15, after looking at fraps data
21:52 paramat i'm testing too
21:54 Fixer standard dev was ~5.3 both two runs with this IF method, standard game is about ~5-6 too, so not much win
21:54 celeron55 what does the mesh update graph look like?
21:54 Fixer sleep 15 gives stand dev 1.65
21:55 Fixer one sec, imgur is slow
21:57 Fixer celeron55: https://i.imgur.com/yupq0AO.png
21:58 paramat my code https://github.com/paramat/minetest/tree/meshprocess
21:58 paramat will test with different values
21:59 paramat with max 5 my graph is at solid max like that too, when flying
22:00 Fixer mainloop_draw should be without any big spikes
22:00 Fixer i still have them when it is set to 1
22:01 Fixer i will try this code, with 5
22:01 Fixer or 3
22:02 celeron55 wait where was the code that sleeps?
22:03 celeron55 i can't even find it anymore
22:03 Fixer celeron55: in 2011? yes
22:03 Fixer one sec, will give you link
22:04 Fixer celeron55: https://github.com/minetest/minetest/commit/02006a95402c41833779e3825c44b39d600e566b
22:05 Fixer or this https://github.com/minetest/minetest/blob/02006a95402c41833779e3825c44b39d600e566b/src/client.cpp
22:06 Fixer it was 50, then reduced to 3ms, then refactored/removed I think
22:06 celeron55 so where do you add sleep_ms(15) now?
22:07 Fixer celeron55: here http://pastebin.com/raw/D44biexY / modify it to 15 though
22:08 paramat i'm just a little concerned about what happens to the queue if the client can't process meshes fast enough for a long time. it seems to prioritise nearby meshes
22:09 celeron55 Fixer: what does the graph look like with the sleep, without other limits?
22:13 garywhite celeron55: The guys on -project just brought up an interesting topic: Adding sound to the main menu. Thoughts?
22:13 celeron55 well i guess you'll try something better than adding demoscene music to it
22:15 paramat oh crumbs the music topic. this has been discussed a lot, music is a subgame thing, not an engine thing. i oppose it in mtgame for many reasons
22:15 paramat other subgames can use it if they want
22:15 celeron55 s/to it/to the engine/
22:17 Fixer celeron55: sleep 15 only on Vanessa survival creative: https://imgur.com/a/9qd0T Singleplayer has much better graph, will post in a sec
22:18 Fixer celeron55: in singleplayer https://i.imgur.com/lKkvRHs.png
22:18 Fixer celeron55: it can hit 2 or even 5 meshes depending on server, but it is damn smooth
22:18 rubenwardy I support main menu music for MTG, but it should follow and suit ingame music
22:19 Fixer i already checked few servers with sleep 15 and had very good experience
22:19 garywhite Why are you showing the profiler graph from F5? (& yes, I'm not a coder or a developer-type person, but I do know that it's a profiler graph if you press F5 x2)
22:19 Fixer fps_max = 100
22:19 Fixer will try paramats one
22:21 celeron55 this indicates the spike is somehow caused by mesh generation (something inside MapBlockMesh::MapBlockMesh), not by something that is ran in the main thread, which seems weird
22:21 Fixer nope nope nope nope
22:21 Fixer i tried paramat IF code with limit at 5
22:22 Fixer lots of stuttering
22:22 Fixer no go
22:22 Fixer will try 3
22:22 YuGiOhJCJ joined #minetest-dev
22:22 proller joined #minetest-dev
22:23 Fixer 3 - also stuttering
22:25 Fixer 1 - has less stuttering, but much much more compared to sleep 15
22:25 Fixer i would prefer sleep 15 any day
22:28 Fixer yeah, it is noticable even with limit at 1
22:30 celeron55 it seems there could be something in MapBlockMesh::MapBlockMesh that somehow pauses rendering or the main thread
22:31 celeron55 it's supposed to be able to be run self-contained in a thread without any outside references, which should allow rendering to happen at regular fps alongside of it
22:33 celeron55 and the only point the result should get in touch with the main thread and rendering is the part i originally suggesting throttling
22:36 celeron55 time to read through all of the relevant code in irrlicht i guess...
22:36 celeron55 well, maybe some day
22:43 * VanessaE growls at celeron55
22:46 proller joined #minetest-dev
23:09 butchster joined #minetest-dev
23:26 paramat joined #minetest-dev
23:26 paramat odd, my patch creates quite smooth behaviour for me
23:26 paramat apart from an occasional momentary freeze
23:27 red-002 #5260
23:27 ShadowBot https://github.com/minetest/minetest/issues/5260 -- Use the `ARRLEN` macro in more places and remove an unused macro. by red-001
23:27 Fixer paramat: with what limit?
23:28 paramat 5
23:28 paramat need to test more though
23:28 Fixer for me even 1 was not good enough
23:29 red-002 5260 is quite trivial
23:29 red-002 it just removes some code duplication
23:31 paramat mtgame is too characterless and universal to have menu music, it woud be good for more specialised subgames though
23:31 paramat (would)
23:31 Fixer paramat: do you have stutter/jitter in the game without patch?
23:34 paramat yes lots
23:34 Fixer now try sleep 15 one
23:34 paramat i have intel integrated graphics
23:34 Fixer do you have nvidia?
23:34 Fixer oh
23:35 paramat i spend a lot of time flying at high speed over terrain, this probably doesn't help
23:35 paramat (while deving mapgens)
23:36 red-002 yeah I suppose there isn't anything generic enough for mt-game
23:37 paramat mods could supply music though, if there's a lightweight way to play it
23:38 paramat in-game music is best done generatively from short samples, to avoid huge music files
23:38 red-002 this seems pretty generic https://soundcloud.com/essa-1/generic-game-loop-5-free-to-use?in=essa-1/sets/generic-game-music-loops
23:39 * red-002 goes off on a search for the most generic video game music possible
23:39 paramat hehe well any music by definition can't be generic
23:39 VanessaE originally RBA wanted to suggest Mercury Rain, if I remember right
23:39 VanessaE and that was an S3M
23:39 VanessaE or some such
23:41 red-002 it might be nice to change the speed or something like that when the player gets damaged
23:41 red-002 not sure how hard that would be through
23:42 Fixer every time i join vanessas servers my GPU fans go crazy
23:43 paramat ugh i can't stand this type of stereotyped mainstream movie rigid linking of music to events
23:43 paramat the best movie music is fairly independent from on-screen events
23:43 paramat like MC's music
23:44 VanessaE I don't know what MC music is like but to my mind, music that fits the "theme" of a given area would be enough
23:44 VanessaE go into a cave, the music gets "dark" and "broody"
23:45 paramat that's also really cheesy
23:45 VanessaE up on the surface, bright and happy.  at night?  gloomy but not "scary"
23:45 red-002 that's what people expect music to do
23:45 VanessaE paramat: games have been doing this since the early 80's...
23:45 paramat i hope not
23:45 paramat inly crap games
23:45 paramat (only)
23:46 VanessaE Super Mario Brothers is a crap game?
23:46 VanessaE was*
23:46 paramat that kind of arcade game is different
23:46 red-002 don't see how
23:47 paramat because the whole game is comical and lighthearted, and 'arcady' it works
23:47 VanessaE paramat: I assume you would be free to turn the music off if you don't like it...
23:47 VanessaE wait what?
23:47 VanessaE have you ever played SMB or one of the later iterations?
23:47 red-002 since when is minetest anything but lighthearted?
23:48 red-002 I mean it's kinda hard to tell the theme
23:49 paramat we should stop being concerned with what we imagine people expect from a game, that leads to mainstream game clones, MC broke many expectations and became one of the most successful games ever
23:49 VanessaE paramat: "we" should focus on what makes players enjoy the game.
23:49 red-002 I'm pretty sure linking music with action has been done since like forever
23:50 paramat SMB has a character that works with playful event-based game music, that's not suitable for a sandbox
23:50 VanessaE short of somehow improving gameplay (which there isn't that much room for), in-game music generally makes a game more enjoyable
23:50 VanessaE um, no
23:50 VanessaE SMB's music is not event based (not precisely)
23:50 VanessaE as I recall there's only two event that trigger a change in music - super-star power (change in theme), and running out of time (speeds up the main theme, otherwise it's the same), and
23:51 red-002 I mean if you don't want music affected by the game then you can just mute the game music and play your own music
23:51 paramat 'done since forever' (by others) is no good reason in itself
23:51 VanessaE otherwise the music is themed according to the environment - underground, above-ground, in water, and a few others.
23:51 VanessaE and only the *sound effects* are event-based aside from the two things above.,
23:51 paramat 'you can mute the music' is also no argument for adding a terrible idea
23:52 red-002 you don't seem to have a good agrument for why it's a bad idea other then that you think a lot of people do it
23:52 Fixer music? you don't even have sound for flowing water, nor your hand display anything when viewed outside /music...
23:52 VanessaE a thousand game companies say "terrible idea" is probably wrong.  Hell, even first-person shooters have (or had) music.
23:52 paramat music reacting to biomes would be stomach-churningly cheesy. MC rightly doesn't do this
23:53 VanessaE paramat: no one said biomes.
23:53 red-002 and we don't have to copy mt
23:53 red-002 mc*
23:53 VanessaE just menu, above-ground, underground, maybe something that plays when you're fighting something
23:53 VanessaE doesn't need to be much
23:53 red-002 ^
23:53 red-002 music is meant to set the tone
23:53 paramat red depends what you mean by 'it', music can be excellent if done well
23:54 red-002 The use of incidental music dates back at least as far as Greek drama.
23:54 red-002 if wikipedia is correct
23:55 paramat a thousand game companies create mainstream clones with no artistic merit, doesn't make them right
23:55 VanessaE I wasn't talking about cloners
23:55 VanessaE I was talking about companies that make or made original games
23:55 paramat i'm not arguing against all use of music, read back
23:55 VanessaE I know, but you said "terrible idea"
23:56 red-002 ^
23:56 VanessaE that sounds like an argument against music.
23:56 paramat no, that was for music being closely tied to events or environments
23:57 red-002 ...
23:57 paramat MC has the right idea, the music is completely independent from events or environments
23:57 red-002 or they couldn't be bothered to add it
23:58 paramat this reminds me of arguing with RBA :] he had the worst imaginable approach to music in games
23:58 VanessaE heh
23:58 red-002 really I dom't think there is any point to music otherwise
23:58 red-002 don't*
23:58 VanessaE looking at http://minecraft.gamepedia.com/Music  it would appear you're not entirely right about MC music
23:59 VanessaE paramat: ^
23:59 VanessaE some of MC's music IS environment-oriented.
23:59 paramat visuals too we were polar opposites
23:59 paramat oh ok
23:59 red-002 well good background music is not obvious
23:59 paramat not in an overly obvious stereotyped way

| Channels | #minetest-dev index | Today | | Google Search | Plaintext