Minetest logo

IRC log for #minetest-dev, 2017-01-31

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

All times shown according to UTC.

Time Nick Message
00:24 DI3HARD139 joined #minetest-dev
00:30 Miner_48er joined #minetest-dev
00:56 STHGOM joined #minetest-dev
00:56 STHGOM joined #minetest-dev
00:56 STHGOM joined #minetest-dev
01:00 crazyR joined #minetest-dev
01:05 octacian joined #minetest-dev
01:07 STHGOM joined #minetest-dev
01:07 STHGOM joined #minetest-dev
01:11 octacian joined #minetest-dev
01:13 STHGOM joined #minetest-dev
01:18 STHGOM joined #minetest-dev
01:18 STHGOM joined #minetest-dev
04:16 STHGOM joined #minetest-dev
04:16 STHGOM joined #minetest-dev
04:25 Hunterz joined #minetest-dev
05:18 Lunatrius` joined #minetest-dev
06:20 Hunterz joined #minetest-dev
06:38 nrzkt joined #minetest-dev
06:51 proller joined #minetest-dev
06:54 proller joined #minetest-dev
07:11 proller joined #minetest-dev
07:16 paramat joined #minetest-dev
07:25 lisac joined #minetest-dev
07:29 Lunatrius` joined #minetest-dev
07:31 ^v joined #minetest-dev
07:36 paramat #5146 simple and seems good
07:36 ShadowBot https://github.com/minetest/minetest/issues/5146 -- Fix fog weirdness by numberZero
07:42 nrzkt paramat, i agree too, merging
07:43 paramat great
07:43 nrzkt paramat, what do you think about #5151 ?
07:43 ShadowBot https://github.com/minetest/minetest/issues/5151 -- Tell on_punch to expect a return value by duane-r
07:58 paramat not sure yet
08:01 kaeza it fixes broken API
08:15 ^v joined #minetest-dev
08:39 nrzkt joined #minetest-dev
08:45 ^v joined #minetest-dev
09:10 ^v joined #minetest-dev
09:22 Darcidride joined #minetest-dev
09:25 Darcidride joined #minetest-dev
09:31 Darcidride joined #minetest-dev
09:35 Darcidride joined #minetest-dev
09:37 ^v joined #minetest-dev
09:40 Darcidride joined #minetest-dev
09:42 xerox123 joined #minetest-dev
09:43 Lunatrius` joined #minetest-dev
09:44 ^v joined #minetest-dev
09:45 Darcidride joined #minetest-dev
10:01 ^v joined #minetest-dev
10:06 Darcidride joined #minetest-dev
10:10 ^v joined #minetest-dev
10:16 ^v joined #minetest-dev
10:47 ^v joined #minetest-dev
10:49 YuGiOhJCJ joined #minetest-dev
10:51 juhdanad joined #minetest-dev
11:00 ^v joined #minetest-dev
11:06 An0n3m0us joined #minetest-dev
11:22 ^v joined #minetest-dev
11:25 red-001 !seen davisonio
11:25 ShadowBot red-001: I saw davisonio in #minetest-dev 20 weeks, 2 days, 20 hours, 31 minutes, and 23 seconds ago saying "Perhaps it can be solved by setting a maximum limit of 14 in the light source definition"
11:27 paramat ^v Darcidride your join / leave messages are spamming this channel, anything you can do to reduce?
11:28 paramat phew :P #5148 updated and seems to work
11:28 ShadowBot https://github.com/minetest/minetest/issues/5148 -- Mgvalleys: Fix missing decorations and incorrect function order by paramat
11:30 troller joined #minetest-dev
11:32 paramat #1475 updated any comments? this adds sand resources throughout the world instead of just near sea level
11:32 ShadowBot https://github.com/minetest/minetest/issues/1475 -- Client lag improvement by sapier
11:32 paramat game#1475
11:32 ShadowBot https://github.com/minetest​/minetest_game/issues/1475 -- Ores: Add silver sand blob ore, relocate other blob ores by paramat
11:35 paramat input needed on game#1532 i'm unsure
11:35 ShadowBot https://github.com/minetest​/minetest_game/issues/1532 -- Bushes: Add leafdecay to bush leaves by paramat
11:37 Fixer joined #minetest-dev
11:47 paramat game#1539
11:47 ShadowBot https://github.com/minetest​/minetest_game/issues/1539 -- Creative: Cache creative mode setting by paramat
11:58 ^v joined #minetest-dev
11:59 juhdanad joined #minetest-dev
12:03 red-001 for CSM is isolating mods from each other really needed?
12:06 red-001 I mean if mods are written correctly and set all their internal functions to locals etc I don't see why having them share the environment is a problem
12:15 ^v paramat, sorry lol my VPS is having issues
12:23 blaze joined #minetest-dev
12:35 troller joined #minetest-dev
12:36 nore updated #5143
12:36 ShadowBot https://github.com/minetest/minetest/issues/5143 -- Fix anticheat resetting client position after the client is teleported. by Ekdohibs
12:52 juhdanad joined #minetest-dev
12:54 celeron55 red-001: most mods are not written correctly and never will be
12:55 celeron55 that's a silly presumption
12:55 red-001 still if the worst thing that can happen is a another mod breaking...
12:55 red-001 plus only the mods that are really badly written will break
12:56 nrzkt celeron55: the main problem with your approach is the callbacks
12:56 nrzkt and the fact that builtin should be instancied in every mod
12:56 nrzkt but it's possible to sandbox them in multiple Lua states
12:56 red-001 I mean it's not that hard to make sure your internal stuff is local
12:56 celeron55 assuming client-side mods never have any special privileges, even local ones if they run in the same environment, then there is no security benefit; mods will just break each other
12:57 celeron55 people will probably start to want to have client-side mods that aren't sandboxed; is there a plan for that?
12:57 celeron55 just say no?
12:57 red-001 I think so
12:58 celeron55 people will want client-side mods that interface with the menu; is there a plan for that?
12:58 celeron55 just say no to that too?
12:58 red-001 which menu?
12:58 red-001 main menu?
12:58 celeron55 yes
12:58 nrzkt main menu is already in its own lua state
12:58 nrzkt client state != main menu
12:58 red-001 thats already scriptable
12:59 nrzkt celeron55: i think not sandboxed mods is a good candidate for a modpack
12:59 nrzkt you want interaction with some mods, use a modpack, it's designed for
12:59 nrzkt but okay, we will think about sandboxing each mod, this should be too have for init, it just add complexity on callbacks
13:00 celeron55 i don't think we need to think about that
13:00 celeron55 we just need to decide what the security model is and if it involves that or not
13:01 nrzkt what do you want, it's not clear, 1 mod(pack) = 1 lua state or 1 state for all mods ?
13:02 Fixer on side note, modpack system is so bad, can someone redesign it? https://github.com/minetest/minetest/issues/4183
13:02 red-001 is one mod crashing suppose to just print an error to the console then?
13:03 celeron55 neither; i want that people don't just disregard these issues and assume everything will be fine
13:03 red-001 I mean if thats still going to cause the client to disconnect then this will be pointless
13:04 nrzkt red-001: absolutely not, if a mod is crashing it's user responsibility for CSM, like in many games
13:04 nrzkt if a mod is crashing and it's not in its own lua state it's nearly imposisble to recover safely from error
13:04 nrzkt ignoring it can trigger many bad things
13:05 red-001 I meant if sandboxing gets added
13:05 troller joined #minetest-dev
13:05 nrzkt i don't think it's a good idea, it's safer, because we can disable mod, but what about user experience ? mod suddenly disappear ? it's strange
13:06 red-001 surely if faulty code in one mod make other mods crash we are back were we started
13:06 nrzkt no, because user knows which mod failed and can disable it
13:06 nrzkt like in every game
13:07 celeron55 how about error reporting
13:07 nrzkt error is reported on screen and in logs
13:07 celeron55 currently mod errors end up in the server owner's hands and they are gnerally somewhat knowledgeable about how to go about fixing them
13:08 celeron55 csm errors end up in the player's computer unless a system is added that allows them to be sent to the server which provided them
13:09 nrzkt celeron55: CSM doesn't download anything from server it's not in the first mergeable version of the PR
13:09 celeron55 as the mods run in a sandbox, they shouldn't have access to anything that should be private so the errors can be sent without issues about what the data contains
13:09 celeron55 yes but this will surely happen in the future
13:10 nrzkt maybe, but reporting should be done in two parts, for CSM pure client maybe send the trace to modstore for reporting to modders, for CSM from server yes report to server owner
13:11 celeron55 well, in any case "send the trace to wherever the code came from" seems to make sense
13:11 nrzkt celeron55: yes
13:13 STHGOM joined #minetest-dev
13:13 celeron55 people have had much hopes about having multithreaded mods on the client though given that the single environment on the server has problems keeping up with things
13:13 celeron55 i guess that's not happening
13:14 red-001 well that's a good reason for multiple lua states
13:14 nrzkt MT mods is not required client side atm
13:14 nrzkt MT mods = please use all client CPU
13:14 red-001 better then saving modders from their badly written code
13:15 nrzkt just optimize mods, it's mod, not a IA
13:15 nrzkt neither an heavy loaded server
13:15 nrzkt red-001: did you ahve some PR for CSM to review ?
13:16 celeron55 my opinion is that it probably doesn't make sense to multithread everything by default; however it might make sense to have multiple "micro-environments" for different things like my lua client-side physics thing and eg. lua map generators on the server possibly
13:17 celeron55 maybe even essential
13:17 nrzkt MT mods has sense server side yes, especially for lua mapgen, but it's not the CSM discussion
13:18 celeron55 well i try to think about the design generally
13:18 paramat would be nice to be able to multithread the main generation loop of a lua mapgen, which tends to be the processing of a lua voxelmanip
13:19 nrzkt paramat: it's not easy, and very dangerous to implement, you should be careful about that, the main problem will be the return callbacks, not the calls to lua state
13:19 nrzkt calls from lua to core when generating
13:20 paramat yes hmmmm was considering it but afaik ran into issues
13:20 celeron55 i don't have much reasons to be nitpicky about CSM details but i do have reasons to try to form an idea about what might happen in the future
13:22 red-001 well it might be a good idea to make all variables created by mods local by default
13:22 red-001 if thats possible
13:22 celeron55 the way to make lua mapgens is to completely decouple it from the regular lua environment (aside from some messaging system if needed) and run it in sync with the emergethread just like the C++ mapgen runs
13:22 nrzkt it's only possible mod side
13:22 red-001 and add some sort of supported way for mods to add public API's
13:22 nrzkt celeron55: this is a good and difficult idea
13:23 celeron55 which is exactly the same as with my lua physics demo
13:23 nrzkt red-001: we can only write guidelines to help modders
13:23 celeron55 it's not very difficult at all, you just have to live with the limitations
13:24 celeron55 which are the same limitations that the C++ mapgen has
13:25 STHGOM joined #minetest-dev
13:25 STHGOM joined #minetest-dev
13:25 celeron55 i probably could implement a demo of that in a day if i wanted to
13:26 celeron55 i might want to if someone else would like to seriously try using it
13:27 STHGOM joined #minetest-dev
13:28 paramat if anyone coded threaded lua mapgen i would use it heavily
13:29 paramat i have so many lua mapgens that are problematic to use due to the lag they create, especially on servers
13:29 celeron55 also connecting a fast path from farmap to it would be interesting
13:30 nrzkt personnaly i have many things to do on the CSM, modstorage and player management server side :p
13:30 paramat if you consider that a lua mapgen is running in 'on generated' and only modifies the current chunk, maybe it's easier to thread
13:31 paramat maybe only if no other mod is doing stuff in 'on generated'
13:31 celeron55 it wouldn't work in that way
13:32 celeron55 we are talking about a separate file not loaded in to the regular environment with no other APIs than on_generated, probably not even get_node or such
13:32 celeron55 just the VM
13:32 blaze joined #minetest-dev
13:32 celeron55 and noise things
13:33 celeron55 and anything else that i forget that is absolutely necessary
13:35 red-001 node ids?
13:37 red-001 #5152
13:37 ShadowBot https://github.com/minetest/minetest/issues/5152 -- [CSM] Add `get_wielded_item` by red-001
13:37 proller__ joined #minetest-dev
13:42 Player_2 joined #minetest-dev
13:47 STHGOM joined #minetest-dev
13:49 STHGOM joined #minetest-dev
13:50 STHGOM joined #minetest-dev
13:53 nrzkt red-001: you are very active :)
13:53 red-001 well this is mostly adding lua bindings to c++ stuff
13:54 nrzkt red-001: can you give me the variable output printed by your example ?
13:54 nrzkt the PR is trivial and looks great, porting is easy as it seems
13:55 red-001 {
13:57 red-001 {
13:57 red-001 metadata = ""
13:57 red-001 count = 1
13:57 red-001 name = "default:sword_steel"
13:57 red-001 wear = 15232
13:57 red-001 }
13:58 proller__ joined #minetest-dev
14:06 nrzkt okay nice
14:06 * red-001 starts working on porting meta-data stuff to client
14:06 STHGOM joined #minetest-dev
14:06 STHGOM joined #minetest-dev
14:07 nrzkt red-001: warning, metadata is only server side, it's possible to have an interface client side, but it will not be persistent
14:07 red-001 STHGOM, You need to fix your connection
14:07 nrzkt red-001: can you fix the style on 5152 it's trivial i can merge it
14:07 red-001 nrzkt, I only want to add meta-data reading
14:08 lumidify joined #minetest-dev
14:08 nrzkt but metadata is not send from server to client ?
14:08 nrzkt sent*
14:08 red-001 I'm pretty sure it is
14:09 est31 it is sent
14:09 est31 all formspecs for chests are in metadata
14:09 est31 and more thing
14:09 red-001 the chest formspec is made using metadata
14:10 est31 in fact, good thing to add would be a client side function that autogenerates formspecs so that in future it doesnt have to be stored in the metadata
14:10 est31 (but I guess some people want to eliminate formspecs entirely)
14:10 est31 its quite funny, some mods store passwords in metadata
14:11 nrzkt est31: it's not like if we store password in memory during client object lifetime :(
14:11 est31 lol
14:11 est31 I think we do
14:12 est31 I more refered to mods like ones where you can add a password so that you can access a chest
14:12 nrzkt yeah
14:12 proller__ joined #minetest-dev
14:12 red-001 well thats a pretty bad way to do it
14:13 est31 you have to store them *somewhere*
14:13 est31 a good way would be to store a hash
14:13 est31 and have a server side secret
14:13 nrzkt est31: maybe having it in modstorage when merged ? :)
14:13 nrzkt storage.get("vault_x_y_z_password")
14:13 nrzkt :D
14:14 est31 you compute the hash via hash(server side secret, password)
14:14 est31 then it can still be stored in metadata
14:14 nrzkt est31: we don't have a password type ? could be good to generalize this model
14:14 est31 as you cant do a brute force attack, since you dont know the server side secret
14:15 est31 nrzkt: perfectly it would be stored outside of metadata, in some file or database
14:15 red-001 nrzkt, updated the PR
14:15 est31 or we introduce a server side only metadata concept
14:15 STHGOM joined #minetest-dev
14:15 STHGOM joined #minetest-dev
14:15 est31 but idk
14:16 est31 other than for passwords there are no good use cases
14:16 celeron55 how about a proposal from 2012 :D http://c55.me/minetest2/wiki/doku.php?​id=roadmap:nodemetadata_proposal#todo
14:17 STHGOM_ joined #minetest-dev
14:17 est31 :)
14:18 celeron55 well also other things that are supposed to be hidden inside a node, like inventories of locked chests
14:19 STHGOM joined #minetest-dev
14:19 Hunterz joined #minetest-dev
14:20 est31 https://github.com/ac-minetest/basic_machines/c​ommit/fb5aca17487ccd3f5bf41dfc3385a4b051f4f70f
14:21 est31 hash makes it at least non-brute forceable
14:21 est31 err
14:22 est31 hashing makes it at least non trivial
14:22 est31 but easy passwords still can be brute forced
14:22 est31 and yeah, probably some people don't want inventories of locked chests to be sent
14:22 * red-001 wonders what sort of person would brute force a password for a chest
14:24 est31 maybe the chest contains an admin pick, which doesnt check for server privilege because "I never lose it"
14:26 red-001 good point
14:28 nrzkt joined #minetest-dev
14:31 est31 with inventories of nodes in fact its quite hard because some players do need them while others dont
14:31 est31 so you need to dispatch the per node check to a mod
14:32 est31 or have a whitelist of players allowed to access in the metadata
14:32 est31 neither is a good solution i think
14:34 est31 but idk
14:35 paramat game#1540
14:35 ShadowBot https://github.com/minetest​/minetest_game/issues/1540 -- Trees: Add 'snowy' group for pine sapling snow detection by paramat
14:54 nrzkt red-001: there is one remaining issue on 5152
14:56 red-001 fixed
15:00 octacian joined #minetest-dev
15:09 lumidify joined #minetest-dev
15:17 VanessaE I'd like to suggest a feature (or rather, an API change):  if drawtype is set to framedglass, this forces param2 to create a "leveled" appearance to one of the node's textures.  If paramtype2 is set to "color", I'd like this to override the leveled behavior and make it behave like any other node (but the "leveled" part of the node would always be "full")
15:20 VanessaE also the infill of that drawtype needs shifted inward slightly, to eliminate Z-fighting.
15:36 STHGOM joined #minetest-dev
15:36 STHGOM joined #minetest-dev
15:39 paramat currently they 'level' like liquid in a tank afaik, is this levelling a useful feature? always seemed weird to me
15:39 paramat well i guess it can't be removed now
15:40 VanessaE I find it a useless feature as well
15:40 VanessaE however, I'm just proposing an override if paramtype2 is set to "color".
15:41 VanessaE that way one can apply param2 coloring to the central part of the node.
15:41 VanessaE (or whatever)
15:41 VanessaE (as I recall, RBA created that feature with Pipeworks' water tanks in mind, but it was never put to use)
15:55 red-001 what sort of minimap handlers would be useful for the client?
15:58 ^v joined #minetest-dev
16:01 VanessaE I imagine being able to enable it only when the user wields a certain node, maybe a similar function for the radar mode
16:02 VanessaE s/node/item/
16:02 nore red-001: #3747 could give you some ideas
16:02 ShadowBot https://github.com/minetest/minetest/issues/3747 -- Moddable Minimap by est31
16:04 red-001 thanks for the link
16:08 Hunterz joined #minetest-dev
16:18 STHGOM joined #minetest-dev
16:18 STHGOM joined #minetest-dev
16:18 ^v joined #minetest-dev
16:19 nrzkt red-001: position size, type
16:24 nrzkt red-001: maybe change the texture around it, permitting to customize it from mod
16:44 VanessaE paramat:  bug in mt_game,
16:44 VanessaE local dir = minetest.facedir_to_dir(param2)
16:44 VanessaE local p = {x = bed_pos.x + dir.x / 2, y = bed_pos.y, z = bed_pos.z + dir.z / 2}
16:44 VanessaE this occasionally returns dir = nil (like 90, mt_game/mods/beds/functions.lua)
16:44 VanessaE line*
16:45 VanessaE I think it falls over if the target bed isn't normal facedir
16:45 VanessaE (as opposed to colorwallmounted, in this case)
17:02 nore sfan5, sofar, could you please recheck #5143 and approve the last commit, so it can be merged?
17:02 ShadowBot https://github.com/minetest/minetest/issues/5143 -- Fix anticheat resetting client position after the client is teleported. by Ekdohibs
17:02 sfan5 sure
17:03 paramat joined #minetest-dev
17:03 paramat perhaps colorfacedir is causing the problem
17:03 paramat facedir_to_dir may now need to only look at the first 5 bits
17:05 paramat it's an engine issue
17:05 sfan5 nore: you could use MYMAX instead of the if(lag_pool_max < LAG_POOL_MIN)
17:05 sfan5 also that particular if is missing spacing
17:05 nore sfan5: ah, indeed
17:05 paramat please open an issue :]
17:05 nore fixing that
17:07 VanessaE paramat: consider this my open issue ;)
17:07 paramat #5148 is simple and tested, can anyone +1?
17:07 ShadowBot https://github.com/minetest/minetest/issues/5148 -- Mgvalleys: Fix missing decorations and incorrect function order by paramat
17:07 paramat reporting a bug in IRC will get forgotten
17:08 VanessaE too lazy to open an issue :)
17:08 paramat ok
17:09 nore #5153
17:09 ShadowBot https://github.com/minetest/minetest/issues/5153 -- Fix facedir_to_dir and wallmounted_to_dir for coloured nodes. by Ekdohibs
17:09 paramat i'll quickly do it
17:09 nore paramat: no need :p
17:09 paramat heh
17:09 VanessaE heh
17:09 paramat thanks
17:10 sfan5 lua has modulo operators?
17:10 sfan5 i guess it does
17:10 sfan5 ¯\_(ツ)_/¯
17:10 paramat then there's this #5109 which is simple, well tested, but maybe a reviewer would want to test (that's the boring part)
17:10 ShadowBot https://github.com/minetest/minetest/issues/5109 -- OpenAL sound: Use a simpler distance model by paramat
17:11 nore sfan5: yep, it does, but it doesn't have bitwise operators :'(
17:11 paramat no rush i guess
17:13 nore sfan5: done the change to use MYMAX
17:15 proller__ joined #minetest-dev
17:20 sapier joined #minetest-dev
17:28 nrzkt joined #minetest-dev
17:51 STHGOM_ joined #minetest-dev
17:53 STHGOM joined #minetest-dev
17:54 STHGOM joined #minetest-dev
18:00 red-001 could some give #5152 a WIP tag?
18:00 ShadowBot https://github.com/minetest/minetest/issues/5152 -- [CSM] Add `get_wielded_item` by red-001
18:00 red-001 Turns out I need to fix some stuff as l_item.cpp uses getServer often
18:07 sofar nrzkt: how often does the lua client sandbox run?
18:07 sofar nrzkt: does it have the same concept of server_step?
18:07 sofar client_step?
18:07 nrzkt yeah it's already implemented, but i seem you don't read preview mod
18:08 nrzkt red-001, and me add each functional feature to preview mod permitting user to see them
18:08 sofar nrzkt: no time to test, sorry
18:08 nrzkt just read the mod
18:08 sofar in between one mt project I'm working on now and the house building design work, 0 free time left
18:08 sofar :P
18:09 nrzkt and you also have client_doc_api.txt
18:09 sofar where's the mod? :D
18:09 nrzkt everything is documented
18:09 sofar minetest-mods? :P
18:09 nrzkt no
18:09 nrzkt in the branch itself in clientmods
18:09 nrzkt https://github.com/minetest/minetest/​blob/47574ac72e53c3a45a9ed4c4eeabfeb6​fd6488de/clientmods/preview/init.lua
18:10 sofar found it
18:11 sofar question again
18:11 sofar can I run something every step and get player controls? e.g. mouse down?
18:11 sapier Imho clientstep should be removed lua handling in serverstep is one of the major design errors of server
18:12 nore sapier: what do you mean?
18:12 sapier if you need somesthing asynchronously to be done create a separate lua thread but don't allow broken mods to cause rendering to be delayed
18:13 sfan5 yes rendering should never be delayed by an client side mod stuff
18:13 sofar rendering should never be affected by lua execution period
18:13 nore ah yes
18:13 sapier yes but exactly this happens if you implement on_clientstep
18:13 nore I thought you meant removing on_serverstep
18:14 nore (hm, on_globalstep actually)
18:14 sofar hell first test I want to do is make a lua client mod that spins in a loop for 1 second
18:14 nrzkt sapier, the problem is the rendering itself as it's not, it should be threaded
18:14 nrzkt client should have its life and renderer too
18:14 sapier that one too but it's way more hard there then not implement it client side at all
18:14 nore though, can we really do that without having map concurrency access problems?
18:15 nore *concurrent map access
18:15 sapier nrzkt: of course but right now rendering is mixed to ui control adding lua to it too wont make things better
18:15 sapier client doesn't have map access
18:15 sapier at least not writeable
18:15 nore hmm
18:15 nrzkt we already know that
18:15 nrzkt client should not modify the map
18:15 nrzkt but lua should have access to control, can be interesting
18:15 sapier client MAY NOT modify map
18:16 nore yes, but couldn't there be concurrent free or things like that?
18:16 nore I mean, let's say that sector of the map hits unload time limit and is unloaded
18:16 nore couldn't it cause problems if the map is accessed at the same time?
18:17 sapier right now we're talking about on_clientstep and imho adding it is a very bad idea because there WILL BE broken mods and modders will claim "but you made the interface available"
18:17 sofar #define will be broken
18:17 sapier nore map access is not topic here
18:17 nrzkt sapier, i got an idea, participate on #5088 and add your comments
18:17 ShadowBot https://github.com/minetest/minetest/issues/5088 -- Client side scripting/modding by nerzhul
18:17 nrzkt this permit to prevent repeating things everytime :p
18:18 nore sapier: the thing is, we *can't* execute lua asynchronously in that case
18:18 nore like, every callback can take 1s
18:18 nore so the thing is, adding on_clientstep will not break more things than can already be
18:19 sapier serverside globalstep is a continous source of lag for about 5 years now I'm absolutely against doing same error clientside once again ;-)
18:19 sapier of course there are other callbacks causing lag too (doesn't make the main step thing better)
18:20 nrzkt sapier, the server source of lag is ABM not globalstep
18:20 sapier nrzkt: only for very very very bad written mods
18:21 nrzkt in fact no
18:21 sapier and abm lag is quite easy to track down the global or entitystep lag is extremely hard to find
18:21 nrzkt just use a valgrind on a classic server modpack and you will see 70% server time spent is on ABM
18:21 sapier un fact yes because noone reads the important message "ABM's have to be one line functions"
18:22 nrzkt but they solve many things
18:22 sapier doing fancy stuff in abm is stupid (sorry if this offends someone)
18:22 nrzkt with a lag cost, yes
18:22 nrzkt dirt replacement, flowers , fields etc
18:22 sapier abms are only supposed to do small stuff or trigger things
18:23 sapier and the triggering is where globalstep comes in ... as it's not separated from main step there's no good way to do asynchronous things
18:23 nrzkt in majority of case yes, but the problem is... algorithm is huge, looping on everynode consume some time
18:23 sapier imho we shouldn't do same mistake once again
18:23 nrzkt sapier, maybe you can participate on this PR
18:24 nrzkt just IRC is not sufficient
18:24 sapier nrzkt: yes but if there's only a one line action in abm it's not that bad (compared to functions reading whole map for each abm)
18:24 sapier ppl do that especaly those who abuse abm for spawning
18:25 nrzkt it's normal there is no other mechanism to do it
18:25 nrzkt the problem is here, ABM are the only way to do things and it consume many time
18:25 nrzkt globalstep is not very used, except for after functions
18:25 nrzkt and it's load is tiny
18:26 nrzkt it's not the problem
18:28 sapier yes it's normal but not because there's no other mechanism to do it. It's because it's easiest way to do it. See I wrote "advanced spawning" mod exactly to do the spawning. There's no need to use abm's for spawning at all
18:29 nrzkt then why it's not used ?
18:30 sapier I guess because of well known "not invented here" syndrom ;-)
18:30 nrzkt or less documented ? and can you proof it's lightweight ? also don't forget one thing with global step, async = many locking issue, in fact everytime, the benefit is not here client side
18:30 sapier well documentation is a little bit short too yes but it'd be enough to start
18:30 sapier it's not lightweight, but it's lag free
18:31 sapier basicaly adv_spawning impelments it's own scheduling
18:31 sapier it ensures there (usually) won't be consumed more the a couple of ms per server step and delays any other work to be done to future steps
18:33 sapier sadly there's no preemption in lua so user can break the delayed work mechanism in adv spawning too ... but you have to break it on purpose
18:33 sapier https://github.com/sapier/adv_spawning
18:33 numZero joined #minetest-dev
18:35 sapier but back to client on globalstep Imho there's no reason to run async thread within main client thread
18:35 sapier on server it's required to have map r/w access but that's not required in client anyway
18:37 sapier same for core.after that one could be run in a separate thread too (well I quess it'd be automaticaly there anyway)
18:58 Krock joined #minetest-dev
19:00 numZero_ joined #minetest-dev
19:05 celeron55 is anyone in the current csm developers willing to try those?
19:11 nrzkt no not actually
19:18 STHGOM joined #minetest-dev
19:18 STHGOM joined #minetest-dev
19:26 rubenwardy joined #minetest-dev
19:28 STHGOM joined #minetest-dev
19:30 STHGOM joined #minetest-dev
19:39 red-001 ok fixed all the issues I could find with the get_wielded_item pr but it will need to be reviewed again and it's a way larger change now
19:48 rubenwardy I'm currently working on meta data. I've made NodeMetadata inherit from Metadata, and made item metadata mostly work (I'm just working on serialisation atm)
19:48 sapier metadata? clientside?
19:50 rubenwardy see my shared_metadata branch
19:50 rubenwardy gtg now
19:51 sapier ahh paramat merged the multiply pull ;-) thanks
19:53 * red-001 was working on client-sided metadata
19:56 sapier red-001 what are client side metadata supposed to do?
19:57 red-001 it's a copy of the server-sided metadata
19:57 red-001 for use by mods
19:57 sapier but there's no way back to server?
19:57 red-001 no
19:57 red-001 there isn't even a way to change it localy
19:57 red-001 just a read-only copy of the server metadata
19:58 sapier hmm well you'll have a usecase for it for sure as you spend time on it
19:58 red-001 easy stop the digging of a chest if it has items in it
19:59 sapier wouldn't that be better done in core instead of csm?
19:59 red-001 stops the chest from disappearing and reappearing  as soon as the server catches up to the client
20:00 red-001 why so?
20:00 sapier seems like exactly same usecase as protection
20:01 red-001 well that could also be improved my client-sided modding
20:01 red-001 by*
20:01 sapier no
20:01 red-001 why not?
20:01 sapier protection is nothing to be done in mods as protection has to be enforced by server
20:02 red-001 yes server enforces it, but the client can make it seem less laggy by not digging nodes that are protected
20:02 sapier of course everything can be done in mods ... but question is always does it make sense to do it in mods
20:03 sapier basicaly this is check a flag "can node be digged" ... it#s quite simple to check in c++ but a lot of overhead in lua
20:18 octacian_ joined #minetest-dev
20:23 juhdanad joined #minetest-dev
20:28 STHGOM joined #minetest-dev
20:31 Fixer joined #minetest-dev
20:34 DonBatman joined #minetest-dev
20:41 STHGOM joined #minetest-dev
20:52 DFeniks joined #minetest-dev
21:32 STHGOM_ joined #minetest-dev
21:32 red-001 closed #4798 as it should be done using client-sided modding
21:33 ShadowBot https://github.com/minetest/minetest/issues/4798 -- Allow overriding and disabling the death formspec. by red-001
21:36 STHGOM_ joined #minetest-dev
22:06 turtleman joined #minetest-dev
22:10 octacian joined #minetest-dev
22:22 proller__ joined #minetest-dev
22:30 proller joined #minetest-dev
22:44 DI3HARD139 joined #minetest-dev
22:47 proller joined #minetest-dev
23:00 Karazhan joined #minetest-dev
23:03 sapier1 joined #minetest-dev
23:08 rubenwardy joined #minetest-dev
23:25 QwertyDragon joined #minetest-dev

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