Minetest logo

IRC log for #minetest-dev, 2015-09-07

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

All times shown according to UTC.

Time Nick Message
00:49 paramat joined #minetest-dev
00:55 paramat i updated #3146 to calculate displacement 'disp' only once for use with leaves and/or plants, much more efficient because these 2 options are usually used together. also, as tuned yesterday, much less compression/stretching of leaves. can anyone review/approve?
00:55 ShadowBot https://github.com/minetest/minetest/issues/3146 -- Shaders: Use triple-frequency waving for leaves and plants by paramat
00:56 paramat now i will reduce the heat/humidity spreads of mgv5/v7 biomes back to 750, as biomes get too big
01:18 RealBadAngel paramat, feels ok, but on the other hand, plants seen from above makes too visible square patterns imho, twisting movement you have eliminated helped there a bit
01:19 RealBadAngel you should think about rotating plants to avoid that and bring more variety to the world
01:19 paramat thanks. yes i noticed the twisting before
01:20 RealBadAngel just set randomly param2 of plants
01:21 paramat as you know i'm not keen on plant rotation :)
01:21 RealBadAngel yeah yeah you love squares, yet you now like round minimap ;)
01:22 paramat yes i prefer the round map because it rotates around the player and has no player marker
01:23 paramat #3152
01:23 ShadowBot https://github.com/minetest/minetest/issues/3152 -- Mgv5/mgv7 biomes: Reduce heat and humidity noise spreads to former value of 750 by paramat
01:24 RealBadAngel but seriously plant jungle grass on 10x10 area and look at it from above
01:24 RealBadAngel they make a grid
01:25 paramat i like the diagonal pattern, i find it suitable for a voxel world
01:25 RealBadAngel without twisting it looks even worse
01:26 RealBadAngel what you called more natural look is non changing diagonal pattern
01:26 RealBadAngel all over the area
01:27 paramat guess i should disturb hmmmm because of my PR above (i miss you)
01:27 RealBadAngel you should rather say "more natural voxel look" ;)
01:28 RealBadAngel anyway, im ok with the code. i will just apply rotation to the plants to avoid those patterns
01:29 paramat thanks for looking
01:30 RealBadAngel btw, another issue, why the heck if i update meta for one node, server updates meta for every node around?
01:32 paramat neighbours or whole mapblock?
01:32 paramat well i dunno
01:34 RealBadAngel propably whole mapblock
01:35 Sketch2 joined #minetest-dev
01:35 RealBadAngel i noticed that with signs, i have 3 signs next to each other, and when i edit just one, client is deserializng meta for all of them again
01:37 RealBadAngel awww lol
01:37 RealBadAngel thats pretty funny
01:37 RealBadAngel ive put there 6 chests
01:38 RealBadAngel editing one signs is causing server to send all 9 metas to the client
01:46 RealBadAngel no wonder inventory operations are so laggy
01:54 paramat now merging 3152
01:59 paramat complete
02:30 Warr1024_ joined #minetest-dev
02:36 exio4 joined #minetest-dev
03:09 dzho joined #minetest-dev
03:19 Siva_AndroIRC joined #minetest-dev
03:42 paramat i might merge the waving changes later, i *think* i have enough approval and testing now, and have reduced the squishing on est31's request
04:02 deltib joined #minetest-dev
04:28 paramat left #minetest-dev
04:34 OldCoder joined #minetest-dev
04:55 Siva joined #minetest-dev
05:31 hmmmmm joined #minetest-dev
05:43 diemartin joined #minetest-dev
05:45 paramat joined #minetest-dev
05:54 paramat ~tell est31 thought i'd mention i had a 'cmakefiles' folder pop up unexpectedly in one of my branches in 'src/threading/', it contained 'cmakedirectoryinformation.cmake' and 'progress.marks'. git didn't gitignore it
05:54 ShadowBot paramat: O.K.
05:57 nrzkt joined #minetest-dev
06:00 paramat now merging #3146
06:00 ShadowBot https://github.com/minetest/minetest/issues/3146 -- Shaders: Use triple-frequency waving for leaves and plants by paramat
06:00 Hunterz joined #minetest-dev
06:03 paramat ah bums 'patch does not apply'
06:03 paramat will do this another time
06:45 paramat fractal mapgen is ready for merge in a basic form, but i have hit a bemusing bug https://github.com/minetest/minetes​t/pull/3088#issuecomment-134002097
06:45 paramat left #minetest-dev
06:45 red1 joined #minetest-dev
07:03 red1 joined #minetest-dev
07:42 rubenwardy joined #minetest-dev
07:50 nrzkt joined #minetest-dev
07:51 rubenwardy #3153
07:51 ShadowBot https://github.com/minetest/minetest/issues/3153 -- Override groups by rubenwardy
07:51 rubenwardy just rebasing to remove bad commit
07:54 rubenwardy done
07:54 rubenwardy #3153
07:54 ShadowBot https://github.com/minetest/minetest/issues/3153 -- Add minetest.override_item_groups by rubenwardy
08:00 Yepoleb_ joined #minetest-dev
08:02 julienrat joined #minetest-dev
08:03 julienrat left #minetest-dev
08:04 leat joined #minetest-dev
08:08 T4im joined #minetest-dev
08:16 Siva_AndroIRC joined #minetest-dev
08:29 kilbith joined #minetest-dev
08:46 rubenwardy nvalid block data in database - https://gist.github.com/rube​nwardy/a0362922ebe85e6b27b6
08:46 rubenwardy Invalid block data in database - https://gist.github.com/rube​nwardy/a0362922ebe85e6b27b6
08:46 rubenwardy hmmmm
08:53 rubenwardy https://www.dropbox.com/s/jn​atised6nya4fg/world.zip?dl=0
08:54 rubenwardy #2610
08:54 ShadowBot https://github.com/minetest/minetest/issues/2610 -- Invalid block data in database: unsupported NameIdMapping version
09:02 julienrat joined #minetest-dev
09:02 julienrat left #minetest-dev
09:12 leat joined #minetest-dev
09:18 leat joined #minetest-dev
09:28 leat joined #minetest-dev
09:31 rubenwardy joined #minetest-dev
10:02 rubenwardy I'm experiencing those 1/2 second jolts, but only when looking at unloaded blocks...
11:49 rubenwardy joined #minetest-dev
11:57 WSDguy2014 joined #minetest-dev
11:57 WSDguy2014 left #minetest-dev
12:01 Darcidride joined #minetest-dev
12:15 rubenwardy is #184 still present?
12:15 ShadowBot https://github.com/minetest/minetest/issues/184 -- Map-related FPS performance regression
12:45 julienrat joined #minetest-dev
12:59 julienrat left #minetest-dev
13:03 Megaf joined #minetest-dev
13:03 Megaf joined #minetest-dev
13:29 Darcidride joined #minetest-dev
13:35 julienrat1 joined #minetest-dev
14:10 book` joined #minetest-dev
14:22 rickmcfarley joined #minetest-dev
14:43 Siva joined #minetest-dev
14:49 hmmmm joined #minetest-dev
15:16 CraigyDavi joined #minetest-dev
15:36 TenPlus1 joined #minetest-dev
15:36 TenPlus1 Hi folks...
15:36 TenPlus1 any devs in the house today ??
15:38 nrzkt ?
15:39 TenPlus1 hi nrz, a quick question about the serialize.h error ppl are having
15:39 TenPlus1 https://forum.minetest.net/v​iewtopic.php?f=6&amp;t=13177 <--- is 4aiman correct in his statement?
15:40 nrzkt please build your servers as a release to avoid the assert
15:41 nrzkt for a server it's better.
15:41 TenPlus1 so using 0.4.13 stable would stop the serialize.h error altogether ???
15:42 nrzkt For the assert itself, a backtrace is needed to find which call send the wrong value
15:42 red1 joined #minetest-dev
15:42 TenPlus1 I've run standalone server for a while and could not get that error to appear, until I teleported far beyond map limits
15:43 TenPlus1 which leads me to believe the updatelistener bug on linux builds may be the cause
15:43 TenPlus1 it teleports you far outside maps also which could crash server and cause same error
15:43 nrzkt strange thing
15:43 TenPlus1 nothing else causes it to happen
15:45 TenPlus1 but... saying that, if 4aiman is right and a check statement could stop error happening and throw up an error instead of a crash...  would it help future testing of builds ?
15:46 red1 I added a node to a group but minetest.get node group still returns 0. What am I doing wrong?
15:46 TenPlus1 red1, show your code plz
15:47 TenPlus1 use pastebin
15:49 Player_2 joined #minetest-dev
15:50 red1 http://pastebin.com/eeZSqgG5
15:53 TenPlus1 the command you are using is deprecated red1
15:53 TenPlus1 use minetest.registered_nodes[node].groups.groupname instead
15:54 TenPlus1 e.g. minetest.registered_nodes["defa​ult:water_source"].groups.water
15:54 red1 oh ok thanks
15:54 TenPlus1 np :)
15:56 red1 I though that the wiki meant get_node_level is deprecated
15:57 TenPlus1 double-checking page...
15:58 TenPlus1 apologies, you are right... I misread the command page...
15:59 TenPlus1 although try the minetest.registered_nodes command to see if that works
16:00 red1 yeah it works
16:00 TenPlus1 :)
16:08 TenPlus1 bye folks
16:22 nrzkt joined #minetest-dev
16:27 Amaz joined #minetest-dev
16:28 ElectronLibre joined #minetest-dev
16:31 technics joined #minetest-dev
16:35 Hunterz joined #minetest-dev
16:45 Robert_Zenz joined #minetest-dev
16:46 Puma_rc_ joined #minetest-dev
16:57 zupoman joined #minetest-dev
17:02 Calinou joined #minetest-dev
17:16 Krock joined #minetest-dev
17:32 rubenwardy joined #minetest-dev
17:35 younishd joined #minetest-dev
17:37 rubenwardy #3153
17:37 ShadowBot https://github.com/minetest/minetest/issues/3153 -- Add minetest.override_item_groups by rubenwardy
18:11 TenPlus1 joined #minetest-dev
18:11 TenPlus1 nrz, thanks for help earlier but their is no 0.4.13 stable build available for ubuntu/mint at this time
18:15 TenPlus1 and the 0.4.13 ppa available on launchpad still has serialize.h error
19:04 rickmcfarley joined #minetest-dev
19:10 FR^2 joined #minetest-dev
19:18 rickmcfarley joined #minetest-dev
19:18 rickmcfarley joined #minetest-dev
19:27 rickmcfarley joined #minetest-dev
19:32 RealBadAngel guys, whos familiar with network code? ive found serious issue with metadata
19:33 RealBadAngel im not familiar with it, and can see it would require lotsa changes
19:35 RealBadAngel this is real nasty bug affecting both server and clients, more developed world (and more complex mods) the things get worse
19:36 RealBadAngel well, its not a bug, its planned by whoever wrote that code
19:36 RealBadAngel problem is that changing single node meta means sending whole block to the client
19:37 RealBadAngel then client erases whole block metas, deserialize got ones
19:37 RealBadAngel and do the same for all neighbouring blocks
19:37 RealBadAngel then all the meshes for those blocks get updated
19:38 RealBadAngel this is trigerred by simple moving something in a chest, or editing the sign, whatever operation on meta
19:38 RealBadAngel mods like technic, pipeworks are then practically unuseable on servers
19:39 RealBadAngel thx to that mechanism VEs spawns have few fps
19:39 RealBadAngel server is busy sending blocks all over, clients regenerating meta and meshes
19:46 est31 joined #minetest-dev
19:46 est31 paramat: will push #3146 for you
19:46 ShadowBot https://github.com/minetest/minetest/issues/3146 -- Shaders: Use triple-frequency waving for leaves and plants by paramat
19:46 est31 ... if I remember how to apply it xD
19:47 sfan5 git am
19:47 RealBadAngel i just used copy paste
19:47 est31 git am fails :)
19:47 RealBadAngel est31, have you read what i wrote just before you came in?
19:48 MinetestForFun joined #minetest-dev
19:48 est31 RealBadAngel, reading now
19:48 RealBadAngel this single change is worth .14 imho
19:48 RealBadAngel thats way too serious issue
19:49 est31 first doing paramat's merge, ok?
19:49 RealBadAngel no hurry, theres no fix for it yet lol
19:51 Amaz Would it fix a bit/lot of server lag?
19:51 RealBadAngel lot? shitload of lots
19:51 RealBadAngel that issue renders well developed worlds unplayable
19:52 Amaz That would make a lot of people happy :)
19:52 RealBadAngel whats most funny approach to send whole block instead of single meta is called inside source as fix
19:52 RealBadAngel thats because of compability with older clients
19:54 est31 RealBadAngel, why do you have low FPS if you have to regenerate meshes?
19:54 RealBadAngel i was always repeating where keeping compability with yrs old code leads
19:54 RealBadAngel to serious fuckups in the end
19:54 est31 whats your proposed behaviour?
19:55 RealBadAngel est31 imagine technic machine and pipeworks moving item in and out of the inventory
19:55 RealBadAngel theres shitload of machines around
19:55 RealBadAngel each meta change triggers block update on client side
19:55 RealBadAngel each block update is causing its neigbours update
19:56 RealBadAngel result is 5-6 fps on VEs servers
19:56 RealBadAngel on very decent machines
19:56 est31 <RealBadAngel> each block update is causing its neigbours update <---- we had this, this is due to things like connected glass or raillike drawtype, to work on block border
19:56 RealBadAngel hold on
19:56 est31 RealBadAngel> each meta change triggers block update on client side <---- this is no bug, but a performance problem
19:56 RealBadAngel take an item out of the chest
19:57 RealBadAngel should it regenerate meshes for every block around?
19:57 est31 no.
19:58 RealBadAngel even regular furnace cooking items is causing mesh updates each cooked item
19:58 H-H-H joined #minetest-dev
19:58 RealBadAngel solution is to send to client just single node meta
19:58 RealBadAngel not the whole block
19:59 est31 well that doesnt solve the actual problem
19:59 RealBadAngel that will solve the problem
19:59 est31 you say the problem is that meshes get regenerated when its not needed eg if you take out sth
20:00 RealBadAngel hold your horses
20:00 RealBadAngel lemme prepare some links for you
20:01 RealBadAngel server.cpp line 926
20:02 RealBadAngel here we have a reason - single node meta change - action = send whole block with all nodes meta
20:03 est31 well there are two issues here
20:03 est31 1. meshes get regenerated when its not needed
20:03 est31 2. if you change metadata, the changes are sent for the whole block
20:04 RealBadAngel server.cpp line 2101
20:04 RealBadAngel the explanation why its done like that
20:04 RealBadAngel 1. is not the issue
20:05 RealBadAngel regeneration of meshes is triggered because block was sent
20:05 RealBadAngel thats normal action for block being sent and received
20:05 RealBadAngel problem is in this case block should not be sent at all
20:05 est31 to address 2. we should perhaps change protocol and add a message that only changes the metadata for a block
20:06 RealBadAngel not add the message
20:06 RealBadAngel fix the existing one
20:06 RealBadAngel the meaning of it is clear
20:06 RealBadAngel but what it does is Monthy Python joke
20:07 RealBadAngel it does exactly the same like adding/removing node
20:07 RealBadAngel event.type = MEET_BLOCK_NODE_METADATA_CHANGED;
20:08 est31 ah I see now
20:08 RealBadAngel this event should not in any case send a block
20:08 RealBadAngel just a single meta
20:08 est31 we can perhaps indeed adjust the change msg.
20:08 RealBadAngel msg?
20:08 est31 no need for .14 here, just a higher proto version
20:08 est31 pkt sorry not msg
20:08 RealBadAngel yes
20:09 RealBadAngel but thats spreaded all over there
20:09 RealBadAngel even in rollback
20:09 RealBadAngel i meant .14 for the effects it can bring
20:10 RealBadAngel that will allow servers to deploy complex mods again
20:11 est31 I dont think 2. is that much of an issue...
20:11 est31 but yeah it should be fixed
20:12 RealBadAngel its damn huge blocker
20:12 RealBadAngel with signs mod its even worse
20:12 VanessaE est31: imagine a well-built-up area (spawn, perhaps) with hundreds of signs, machines, chests, and other stuff moving things around and causing singular meta updates.  Seems to be that with the existing behavior, dozens of mapblocks are being sent for every meta update in such an area.
20:12 RealBadAngel textures got deleted only on client shutdown
20:12 est31 RealBadAngel, with signs mod you do need regular mesh updates, no?
20:12 est31 I mean this is yet another problem
20:13 RealBadAngel when metas get refreshed all the time without a reason textures are made all over
20:13 est31 you propose to change features of the node depending on its meta
20:13 est31 so how can that be achieved without updating the mesh?
20:13 RealBadAngel dont look at the problem from the ass side
20:13 est31 VanessaE, I see that it might cause network traffic, but I still doubt that this is the cause of the low FPS on such spawn areas
20:14 est31 RealBadAngel, I have to look at every side of the problem, in order to solve it
20:14 RealBadAngel server is spamming each client with blocks data
20:14 VanessaE est31: it's just my theory.  if metas are being updated and causing dozens-at-a-time mapblock refreshes on the client, at the very least it'll cause choppy FPS.
20:14 RealBadAngel client are continuously regenerating meshes
20:14 est31 VanessaE, mesh updates are done in mesh update thread not in main thread
20:14 RealBadAngel need a proof join VE server
20:15 RealBadAngel est31, it happens dozen of times each second
20:15 VanessaE RealBadAngel: you benchmarked it?
20:15 RealBadAngel each meta change means send a block
20:16 RealBadAngel each block will cause 9 mesh updates
20:16 RealBadAngel also, erase whole meta on client side for those blocks
20:16 RealBadAngel and deserialize it
20:17 RealBadAngel dont need to benchmark it, i can see clearly shitload of work needed to be done all over again
20:17 RealBadAngel without a single reason for it
20:17 RealBadAngel there are no magic dwarves here to do that all shit
20:19 VanessaE still need to benchmark it.  cut off all meta updates on your client and go visit VE-S
20:19 est31 <RealBadAngel> dont need to benchmark it, i can see clearly shitload of work needed to be done all over again <----- thats the wrong way of approaching things
20:19 VanessaE see how it compares to the normal client behaviopr
20:19 VanessaE -p
20:19 RealBadAngel i can easily disable that command and join your server VanessaE
20:20 VanessaE RealBadAngel: that's the idea.
20:22 RealBadAngel est31, thats absolutely right approach. when we do know that code is doing things in this case absolutely unnecessary
20:22 RealBadAngel whatever time the code takes it means always NOT NECESSARY
20:23 Calinou joined #minetest-dev
20:23 est31 RealBadAngel, we just know that it sends unneccessary data
20:23 RealBadAngel sending is less the problem
20:23 RealBadAngel what it triggers is
20:23 est31 but we cant say whether it is at all an issue on VE's servers
20:23 est31 or whether the low fps is caused by sth else
20:24 RealBadAngel wait, sending blocks all over is not the problem?
20:24 RealBadAngel how much takes well developed block with lotsa machines, signs, inventories?
20:24 est31 RealBadAngel, prove me wrong xD
20:25 RealBadAngel lets say we have 10 clients
20:25 est31 perhaps 4 kb of data
20:25 VanessaE est31: well regardless of which thread mesh updates take place in, every time there IS one, it causes a tiny FPS lag, for me at least.  add up enough of those, happening often enough and you've got a recipe for slow FPS
20:25 RealBadAngel lets say meta changes (for one node) happens 10 times a second
20:25 RealBadAngel 10*10*4kb
20:26 est31 RealBadAngel, have you measured that?
20:26 RealBadAngel we just had here 10 machines
20:26 est31 dont just say some number
20:26 RealBadAngel and just 10 clients
20:26 RealBadAngel thats very light scenario
20:26 est31 but still it isn't *that* stupid
20:26 RealBadAngel sending whole block?
20:27 RealBadAngel thats DUMB
20:27 est31 if you run the machine ABMs, and indeed every of those machine's inventories changes, the block is only sent once for all 10 machines changes, not ten times
20:27 RealBadAngel not to use stronger words
20:27 RealBadAngel add 20 signs to that
20:27 RealBadAngel and 20 new textures on each refresh :P
20:27 est31 well, gtg sleep. bye
20:34 ElectronLibre left #minetest-dev
20:36 Wayward1 joined #minetest-dev
20:36 Wayward1 joined #minetest-dev
20:46 leat joined #minetest-dev
21:01 VanessaE ok well since RealBadAngel is gonna hem and haw over it:  here's his screenshot showing vastly-improved FPS by disabling... something that he hasn't said just what:  https://imgrush.com/M9Ma6966sJt6.png
21:02 nrzkt VanessaE: you made cross realm chat ?
21:02 RealBadAngel lol
21:02 VanessaE heh
21:03 RealBadAngel we all do as it seems ;)
21:03 VanessaE RealBadAngel: so...explain that screenshot
21:03 VanessaE (also, don't mind the wrongly-oriented text on the right, foreground.  signs_lib bug.)
21:05 RealBadAngel the client is ignoring meta changes
21:05 RealBadAngel thus it is pissing on continous request to update blocks
21:06 RealBadAngel result = about 40 fps instead of 5-7
21:06 kahrl_ how does it detect whether a sent block is a meta change or a change of the bulk data?
21:06 RealBadAngel theres special case
21:06 RealBadAngel meant to send only meta changes
21:07 RealBadAngel but whoever coded it, decided to send whole block there to keep compability with older clients
21:07 RealBadAngel see comments in source
21:07 kahrl_ where exactly?
21:13 kahrl_ you can't mean the MEET_BLOCK_NODE_METADATA_CHANGED handler in server.cpp, because changing that would have zero effect when you connect your client to someone else's server
21:24 VanessaE [09-07 16:01] <RealBadAngel> server.cpp line 926
21:24 VanessaE [09-07 16:04] <RealBadAngel> server.cpp line 2101
21:25 RealBadAngel RealBadAngel> the client is ignoring meta changes
21:28 RealBadAngel VanessaE, after all this time i made the screenshot im still connected with 40fps
21:29 RealBadAngel never saw your spawn so smooth
21:46 Wayward1 RealBadAngel, would you mind pastebinning your changes here? I'd like to test them myself
21:46 domtron joined #minetest-dev
22:12 domtron_ joined #minetest-dev
22:18 Lunatrius` joined #minetest-dev
22:25 MC_Pro joined #minetest-dev
22:32 Lunatrius joined #minetest-dev
22:36 Lunatrius joined #minetest-dev
22:47 red1 joined #minetest-dev
23:05 domtron_ joined #minetest-dev
23:23 est31 joined #minetest-dev
23:29 est31 joined #minetest-dev
23:30 domtron_ joined #minetest-dev

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