Time  Nick        Message
02:30 Zeno`       label removed
07:26 nrzkt       hi all
07:49 OldCoder    nrzkt, hi back
09:56 Zeno`       merging #5115 shortly
09:56 ShadowBot   https://github.com/minetest/minetest/issues/5115 -- Plantlike: Fix visual_scale being applied squared by paramat
09:56 Zeno`       it might cause minor problems, but as sofar said we're early in the release cycle
09:59 * Krock     waits for it to upload his newest binary
10:00 Zeno`       Krock, I'll give it 10 minutes wait
10:07 Zeno`       Krock, ok
10:08 Krock       may the linking begin!
13:26 red-001     !seen davisonio
13:26 ShadowBot   red-001: I saw davisonio in #minetest-dev 19 weeks, 6 days, 22 hours, 32 minutes, and 13 seconds ago saying "Perhaps it can be solved by setting a maximum limit of 14 in the light source definition"
14:56 sapier      I'm gonna rebase and merge #5085 in a view minutes so last chance for objections
14:56 ShadowBot   https://github.com/minetest/minetest/issues/5085 -- Make entity on_punch same signature and behaviour as player on_punch by sapier
15:01 sapier      https://github.com/minetest/minetest/pull/5045 nerzhul, sfan5 numberZero are there open issues? blocking this pull request?
15:01 sfan5       it's only missing approvals
15:02 nrzkt       except the code style issue no problem for me
15:02 sapier      which code style issue is left?
15:02 nrzkt       sapier, we need you on 5100
15:03 nrzkt       it should be fixed and merged in CSM branche asap
15:03 sapier      Ahh I see :-) I'm gonna fix
15:05 sapier      nrzkt what exactly on 5100?
15:06 nrzkt       your points about security
15:08 sapier      yes? did someone check or fix them? I don't see a answer there or a change? (maybe I'm blind I sometimes get confused by githubs style of updating)
15:08 OldCoder    sapier, Hi.
15:08 sapier      hello oldcoder
15:09 sapier      nrzkt shall I really change L1670 in tile cpp? while I agree that's not coding style it's done this way everywhere in this file
15:09 nrzkt       i don't know exactly what block this PR
15:09 nrzkt       sapier, our coding rules indicate if you modify a line with wrong style, please fix the style
15:10 sapier      me too 5100 is a improvement for sure and as long as csm ain't in master I see no reason for not merging improvements to that branch
15:10 sapier      that's what the branch is for
15:10 nrzkt       but do you see issue on security on this PR or is this sufficient ?
15:11 red-001     well as far as I can see mods only have access to other mods
15:11 sapier      ok I'm gonna follow the coding style even I believe this is one of the cases where "follow to the text of the rules" is contrary to the intention of the rules ;-)
15:11 red-001     readonly access
15:11 red-001     but I haven't tested it
15:11 nrzkt       red-001, can you rebase your two CSM PR we need to merge them asap
15:12 red-001     nrzkt, I reabsed 5100
15:12 red-001     ok I will rebase the other one then
15:13 red-001     I will rebase the other one as soon as smart git finnishs updating
15:15 sapier      by the way what's result of the file access discussion for what I see there's two opinions "allow (limited) file access" and "no file access at all"
15:16 red-001     I think it's allow special api for file access
15:17 sapier      so "no direct file access" ?
15:17 nrzkt       sapier, i think have a virtual FS could be nice
15:17 red-001     that way we can stop mods from wasting disk space
15:17 nrzkt       mod://<file> in a mod only virtual fs and shared:// for sharing between mods
15:17 sapier      that's what I meant with "no direct file access"
15:17 red-001     e.g. generating 1gb files
15:17 nrzkt       and yes, maybe limit size
15:17 sapier      btw that would be helpfull on server too
15:17 red-001     ^
15:18 sapier      I recently discovered there's still no minetest way to store world specific data is it?
15:18 red-001     are local mods suppose to say or just serve as a test?
15:18 sapier      the only way is store some file to world folder ... of course that's simple too
15:18 red-001     stay not say
15:18 Krock       well, mods can write their config stuff to the world
15:19 sapier      well Krock there's not only config data to be stored to world
15:19 red-001     I think there should be a version of the player storage for worlds
15:19 Krock       yes I know. all other data is serialized in some way
15:20 sapier      e.g. mobf stores advanced spawning data to avoid recalculating it over and over again, quest_engine stores player quest information ... i think there are plenty of other examples where worldspecific gamedata need to be saved
15:20 red-001     then after a release or two direct access could be blocked
15:21 nrzkt       since recent merce you have a way to store player extra attributes, it solve the behaviour for players (ex: player achievement progress, or quest progress)
15:21 sapier      I believe I had some sort of minetest.world_settings_set() pull request years ago ... not sure if I can find it ... well I guess it'd not be a lot of work to reimplement it anyway
15:21 OldCoder    Shara, I'd like to proceed with a text console PR. PM me when here.
15:22 Shara       I'm here, but was  hoping on working on it myself.
15:22 Shara       Feel free to PM.
15:22 Shara       OldCoder: ^
15:22 sapier      hmm player attributes ... true would solve the player relevant part ... I'll check and use it. But wouldn't help for orher parts like spawning data
15:23 red-001     nrzkt, I think we need to add player name to on_chat_message_recive callback
15:23 sapier      is there any limitation to the amount of data stored as attribute?
15:24 red-001     the point at which the player file stops loading normaly>
15:24 red-001     *?
15:24 sapier      I read that as "logicaly no, technicaly of course"
15:24 nrzkt       red-001, of course but its impossible with current chat protocol, it send a raw shit :)
15:25 kaeza       wasn't there some pull request adding SQLite access to Lua?
15:25 nrzkt       i need to finish #5117 to permit this
15:25 ShadowBot   https://github.com/minetest/minetest/issues/5117 -- Chat protocol rewrite by nerzhul
15:25 nrzkt       kaeza, atm no
15:25 kaeza       spawning data could use that
15:25 kaeza       nrzkt, :/
15:25 sapier      I remember that too kaeza there was something similar quite some time ago
15:26 red-001     nrzkt, isn't there some could that separates out the playername?
15:26 red-001     nrzkt, isn't there some code* that separates out the playername?
15:26 nrzkt       red-001, absolutely not, and i'm against adding shitty parser to solve the current chat protocol problems
15:27 sapier      me too fix it correct or don't fix it at all
15:27 red-001     just saying that minetest is alreadying using one so it could be reused for messages that use the old protocol
15:27 sapier      if it was using something like that we should get rid of it instead of increasing that hacks usage ;-)
15:28 red-001     well it even has a comment saying that
15:30 sapier      does "dismiss review" on github mean "yes I fixed it" or "no I don't accept that comment"?
15:30 red-001     I have no idea
15:31 red-001     try it?
15:31 sapier      that's somehow strange do they believe comments are always valid or why don't they provide a simple accept/deny  mechanism
15:31 sapier      well I don't want to offend my reviewers by unintentionally using wrong button ;-)
15:31 Zeno`       <OldCoder> Shara, I'd like to proceed with a text console PR. PM me when here.
15:32 Zeno`       OldCoder, please let Shara have an attempt first?
15:32 kaeza       it's probably for "this comment is no longer valid" a.k.a. "commenter is full of sh¡t" :P
15:33 Zeno`       he did all the hard(ish) work by asking if it might be accepted after all
15:34 Zeno`       s/he/Shara
15:36 sapier      I'm gonna merge #4983 soon
15:36 red-001     nrzkt, there is a typo in the name of 22516eee29126179194e693b89ab8ba20597af3e
15:36 ShadowBot   https://github.com/minetest/minetest/issues/4983 -- from_table: Fix crash for missing inventory or field by SmallJoker
15:37 red-001     on_damage_taken is misspelled
15:38 sapier      merging now
16:02 Fixer       sapier: that entity fix, people in boats/carts can be finally attacked?
16:03 red-001     so can 5100 be merged?
16:04 red-001     nrzkt, sapier ^
16:06 sapier      red-001 to clientscripting branch? of course, but it doesn't mean security is complete
16:06 sapier      fixer I don't know what issue prevented ppl to be attacked in boats, carts
16:07 sapier      but I don't believe this fixes the issue on it's own, at best it makes it possible to be fixed
16:07 Fixer       sapier: there was some deep reason, it was discussed somewhere, still not fixed to this day, you can go in cart and you are immune
16:08 sapier      I don't know but If you're able to get a list of attached entities from the cart damage handler you could just pass the damage to the attached player
16:09 sapier      but that would have to be done by cart/boat mod
16:23 Amaz        Is it a bug or a feature that entities with nametags show up on the minimap like players do?
16:24 nrzkt       merged #5100
16:24 ShadowBot   https://github.com/minetest/minetest/issues/5100 -- [CSM] Improve security by red-001
16:24 sapier      Amaz it's a feature
16:24 OldCoder    Zeno`, I'd already made the patch, months ago. But Shara and I have chatted about the matter.
16:25 Amaz        sapier: Okay :)
16:25 Zeno`       ok
16:26 nrzkt       red-001, are you interested about the VFS issue for CSM ?
16:26 nrzkt       oh #5113 needs rebase
16:26 ShadowBot   https://github.com/minetest/minetest/issues/5113 -- [CSM] Add function to get the name of the player and functions to colour chat. by red-001
16:28 crazyR      Who maintains the servers list?
16:28 nrzkt       crazyR, serverlist server itself or servers in the list ?
16:28 nrzkt       the list is published by servers themselves
16:31 crazyR      nrzkt: serverlist server
16:31 nrzkt       sfan5, if i remember
16:39 crazyR      thanks
16:41 red-001     VFS?
16:41 red-001     oh virtual filesystem?
16:42 nrzkt       red-001, hmm i'm working on it atm in fact :)
16:42 nrzkt       red-001, if you want to help maybe adding core.get_node function permit to do nice things client side :)
16:45 red-001     so how are you planning to make vfs work?
16:45 nrzkt       removal of all io method if it's not done in your PR and first add core.write_file and core.read_file, after enhance it to sandbox per mod
16:46 red-001     oh I see, I was thinking of something more high level
16:47 nrzkt       explain me what is your idea ?
16:47 nrzkt       and also limit file size per mod
16:47 nrzkt       per file first and per mod after
16:48 red-001     I was thinking of a database-like system to which mods can save data
16:48 red-001     so core.store("mymod:player_data", table_with_data)
16:49 nrzkt       hmm... using a sqlite db instead of pure files ?
16:49 red-001     and core.load("mymod:player_data")
16:49 sapier      what about something similar to settings minetest.get_settings("somevirtualfilename")
16:49 nrzkt       the problem with sqlite db is the db size, you cannot really limit it
16:49 sapier      and then mysettings:set_(name,value) and mysettings:get(name)
16:49 red-001     well my idea was that it could be first implemented with files, and then maybe a more advance system
16:50 crazyR      https://github.com/ethic-rush/database  <-- thats my lua implementaion of it
16:50 red-001     so take a hash of the name and use that as the file name
16:50 crazyR      its not the best but it works
16:51 nrzkt       red-001, in a file mode, is this better to have one file per mod with a K/V store using json ?
16:52 sapier      crazyR: problem is this relies on io beeing available which we want to remove
16:52 red-001     well I would say one file, make it easier to check file size
16:52 nrzkt       red-001, without doubt :)
16:52 red-001     also we want to make sure that servers can't bypass the file size limit by creating lots of mods
16:53 sapier      imho the api should not provide any information about how the data is stored
16:53 nrzkt       red-001, we should have global limitation & local mod limitation
16:53 sapier      this way we can improve it whenever we want
16:53 red-001     ^
16:54 sapier      just provide a clean api to store key value (as first step)
16:54 crazyR      sapier: i understand. i was referancing that mod as an idea to maybe how the methods could work... structure wise?
16:54 nrzkt       sapier, okay, as a json format ?
16:54 nrzkt       1 file per mod ?
16:54 red-001     I say so
16:54 crazyR      so is there a reason why leveldb cant be used as the storage mechansm?
16:54 red-001     well one file per user account would be better in someways
16:55 sapier      nrzkt: no don't say anything about the format data is stored
16:55 nrzkt       crazyR, leveldb is not a good storage for relational data
16:55 red-001     how unique is the playername salt?
16:55 crazyR      i know. but its gotta be better than file
16:55 nrzkt       sapier, i didn't said i will call it write_json, but i want to know how i store the data really for the first version :p
16:56 red-001     ehh I meant how unique is the salt used for hashing the password
16:56 sapier      for simplicity just store it in lua text form same as settings file?
16:56 red-001     maybe that could be used as the filename
16:56 red-001     or a hash of the playername and salt
16:57 sapier      we should not create hundreds of files containing 10 bytes ;-)
16:57 nrzkt       sapier, i will use CPP as an intermediate, i can like player extended attributes do a json with lua string data in values :)
16:58 nrzkt       red-001, like in World Of Warcraft we should have two storages areas: 1 global & 1 per server
16:58 sapier      I don't understand why you convert lua string data to json just to convert it back to lua but you'll have your reasons
16:58 nrzkt       this permit to store global configurations & local server things
16:58 nrzkt       sapier, it's a K/V interface
16:58 sapier      what about just creating a second sqlite/leveldb (whatever db backend is selected) and store the date in world folder
16:58 sapier      jason ain't a key value interface
16:59 sapier      it's a plain description how to store information
16:59 nrzkt       not exactly but can be used for taht :)
16:59 crazyR      sapier: thats what i was thinking storage wise
16:59 sapier      it could be key subkey1 value1 subkey2 value2 too
17:00 sapier      most modders will store lua tables within the values anyway
17:00 nrzkt       yeah, my idea is  if i do core.save_key('default:var1', lua__variable) it stores it in a json or a DB
17:00 nrzkt       sapier, not a problem, look at player exnteded attributes
17:00 sapier      noone wil manage hundreds of keys when it's more easy to just load the table
17:00 red-001     ^
17:01 nrzkt       sapier, you mean instead of permitting multiple keys a mod store raw variable directly ?
17:01 sapier      well as I said I don't care how the data is stored (unless it's bloated by factor 1000 or something like that ;-)
17:01 nrzkt       and it's its problem to manage keys ?
17:01 sapier      no I'd not make a limitation on number of keys
17:01 red-001     well we should allow a mod to use multiple keys
17:02 nrzkt       K/V and V = raw lua table
17:02 nrzkt       or variable or whatever :p
17:02 sapier      e.g. if a mod just needs two parameters it's stupid to force it to write a table ... on the other hand if a mod needs 100 variables anyway it's (most likely) not gonna use 100 keys
17:02 nrzkt       if we offer K/V modder can use multiple a key or a unique key with its dirt :)
17:02 nrzkt       like modder prefers
17:02 sapier      yes
17:03 nrzkt       okay then K/V call
17:03 nrzkt       after the storage, a DB size is difficult to limit
17:03 sapier      and same api can be used on client and server side
17:03 nrzkt       sapier, yeah, i write it for CSM but it should be ported to server too
17:04 nrzkt       i will create ModApiModStorage class in cpp
17:04 red-001     whats the problem with doing part of the sandboxing in builtin?
17:05 sapier      I'd recommend writing it for master as the api most likely will be identical
17:05 red-001     there is no way to undo seting a variable to nil as far as I know
17:05 nrzkt       sapier, okay, in fact l_modstorage will be called from both sides, i will cp it and send the PR when ready
17:06 nrzkt       sapier, setting to limit data size per mod and globally ?
17:06 red-001     wouldn't ModApiStorage be a better name?
17:06 nrzkt       red-001, ofc
17:06 sapier      red-001: if there's any leak within our sandbox you can escape it way more easy if the code is part of the sandbox then if it's not even loaded
17:07 sapier      hmm I'd start with global at the beginning unless it's no big difference in work to implement it for both
17:07 red-001     but doesn't mod security do  just that?
17:07 red-001     remove access to functions
17:07 red-001     not remove them
17:08 sapier      yes but security was initially done for server where security had to be added to existing environment
17:09 sapier      there it had to preserve some (unwanted) functionallity which has been present before. On client side we have the opportunity to do it as correct as possible right from the begining
17:10 sapier      it's a pitty chrome has gotten that fast compared to firefox
17:12 red-001     why is setlocale a problem?
17:12 juhdanad    nrkzt: is there anything I should change in #4421?
17:12 ShadowBot   https://github.com/minetest/minetest/issues/4421 -- Expose getPointedThing() as Raycast by juhdanad
17:13 sapier      because locale is used for string/number conversions causing sideefects hard to judge
17:14 sapier      and as long as there's no use for it it just spares us from doing that judgement disallowing it
17:14 red-001     well builtin uses it
17:14 red-001     not sure what for
17:14 sapier      for what?
17:14 red-001     os.setlocale("C", "numeric")
17:14 sapier      ohh I see ... yeah I guess it's even been me who added it
17:15 sapier      formspecs string to number conversions
17:16 red-001     so is it ok to just set it to nil in lua it that case?
17:16 sapier      juhdanad: isn't there a better name instead of "raycast" ? I know this is tecnicaly correct but I guess noone using it will ever care about if it's done by ray by woodstick or by anything else ;-)
17:16 juhdanad    Do you have any ideas?
17:16 sapier      red-001: I think so
17:17 sapier      maybe something like get_pointed_at(...)
17:17 juhdanad    And what would be the name of the returned iterator?
17:18 sapier      or get_target() ... well target might be a little bit to generic
17:18 nrzkt       how can we get modname after load ? at mod load lua_rawgeti(L, LUA_REGISTRYINDEX, CUSTOM_RIDX_CURRENT_MOD_NAME); return it, but after it's nil
17:19 sapier      nrzkt: that's difficult
17:19 sapier      especially if mods use functions of other mods (library mods) it's almost impossible
17:19 nrzkt       sapier, then how can i sandbox a mod from another mod in our storage ? :)
17:19 sapier      maybe ShadowNinja knows a way to do it
17:19 red-001     nrzkt, is mod sandboxing needed?
17:20 red-001     I think we should just use the same system as formspecs for this
17:21 nrzkt       maybe not, but in client it's important to have different storages, to prevent rogue mod modifying other mods
17:21 red-001     if mod wants to access data from another mod it should be able to
17:21 nrzkt       access data why not, but storage no
17:21 nrzkt       each mod should have its own storage
17:21 sapier      juhdanad: hmm difficult to find a name for it
17:21 juhdanad    It is mathematically a ray.
17:21 juhdanad    And I like mathematics!
17:22 sapier      As I said it is correct but it's not "speaking" for non-math ppl ;-)
17:23 sapier      btw noone else did mention this by now maybe it's something only I am uncomfortable with in this case don't bother about it
17:25 sapier      nrzkt: on client side distinguishing between server sent mods and manually installed mods is way more important
17:25 sapier      imho server side sent mods shall not be able to store any persistent data on client
17:26 juhdanad    Do you mean that users will be able to play on servers with their own set of additional client mods?
17:26 sapier      juhdanad: yes, but those mods will only be able to change ui things
17:27 juhdanad    I just thouhkt it would be easy to publish a 'killaura' mod.
17:27 nrzkt       sapier, i agree, but server side sent mods is not my priority for first CSM
17:27 sapier      and there's some way to go till this is merged juhdanad
17:27 nrzkt       having a working local mod API and strong API , after we will look at server sent mods
17:28 juhdanad    Off-topic: how do I measure the memory consumption of mapblock meshes?
17:28 sapier      nrzkt: well security is most important for server sent ones. For others it's same as for current mods the one installing it has a certain responsibility for them.
17:30 sapier      but as we don't want two mod engines on client side we have to get the one we want done correct
17:33 sapier      juhdanad: isn't the raycast something like a shootingpath?
17:33 juhdanad    Yes, it is. (but unaffected of wind, Coriolis force etc.)
17:34 sapier      well if you use a railgun those effects are ignorable ;-)
17:34 sapier      maybe call it shoot_at() and shootingpath
17:35 sapier      any other opinions about that?
17:35 juhdanad    I think modders would think that this method actually places a projectile in the environment.
17:35 sapier      I don't wanna be the only one doing suggestions
17:35 sapier      get_shootingpath_to()
17:35 sapier      ?
17:36 sapier      but to be honest my name suggestions usually aren't best ;-)
17:37 juhdanad    Well, it is better. (but to be honest I don't really want to push more commits to that branch)
17:37 sapier      Yes but there are already some nice features noone uses because of name beeing missleading I just want to save your feature from same fate
17:38 nrzkt       sapier, red-001 it's a little draft #5131
17:38 ShadowBot   https://github.com/minetest/minetest/issues/5131 -- Add ModStorage Lua API by nerzhul
17:38 nrzkt       atm no storage as i don't know how to retrieve mod name (the mentioned call return nil)
17:38 juhdanad    The problem is that line_of_sitht is already reserved.
17:38 juhdanad    *sight
17:39 sapier      I'd mention the data being world specific too nrzkt
17:40 sapier      juhdanad: yes but you could add a param to it for example defining the mechanism to be used
17:40 sapier      similar to pathfinding there you can/could specify the pathfinding mechanism
17:42 juhdanad    that would make line_of_sight really tricky. With one parameter it returns a bool and a vector; with another it returns an iterator.
17:43 sapier      true not best idea
17:44 juhdanad    sapier: "there are already some nice features" what are these?
17:44 sapier      let me have a look at lua api i never can remember their names
17:45 sapier      texture modifier is an example noone used it (if someone would have the bugs would've been reported)
17:46 nrzkt       sapier, yes it's an evidence :)
17:46 nrzkt       world private
17:47 sapier      damage and digging mechanisms for sure are used by a very very small group of ppl too ... it's exremely hard to understand
17:47 sapier      imho it's overdesigned to many things put together instead of creating a more speaking while little bit more redundant api
17:48 sapier      not sure but i believe voxelmanip isn't used that often too (but I could be wrong about that)
17:49 sapier      juhdanad: what about path_of_sight?
17:52 sapier      or viewpath?
17:55 red-001     yeah VM is pretty hard to understand
17:56 sapier      btw not relate does someone know how to tell chrome to do a dns lookup prior sending content of address box to search engine?
17:57 red-001     I'm not sure if it does a dns lookup
17:58 sapier      of course noone does care about privacy today but imho it's plain stupid to replace a proven mechanism like dns by something unpredictable like a search engine
17:58 red-001     I think it just checks it the address could be a site
17:58 juhdanad    sapier: path_of_sight is good enough. But are you really sure that 'raycast' is misleading?
17:59 sapier      juhdanad: it's not misleading it's just non speaking to a lot of ppl
17:59 sapier      for others it's gonna tell precisely what is done but I believe first ones are majority ;-)
18:00 juhdanad    What about "nodes_on_line"?
18:01 sapier      yes would be better than raycast to
18:01 sapier      too
18:01 juhdanad    It tells that it returns an iterator.
18:01 sapier      yes that name would tell me what it does too
18:02 juhdanad    Unfortunately it does not tell that entities may also be returned.
18:02 juhdanad    Maybe objects_on_line()
18:03 sapier      thought about that too but term objects is already used for luaentity and player baseclass
18:05 sapier      the function does only provide nodes/objects within the bounding box of pos1/pos2?
18:05 juhdanad    Yes, why?
18:05 sapier      so it basically provides everything the line does cross
18:06 juhdanad    Only if you ask.
18:06 sapier      yes but it could (if you call the iterator to the end)
18:06 juhdanad    If you stop the iteration after the first pointed thing, the remaining area is not computed.
18:06 sapier      line_crosses()
18:06 sapier      or get_line_crosses()
18:06 juhdanad    This would return a bool to me.
18:07 juhdanad    What about Then the only remaining option is things_on_line()
18:07 sapier      get_crossed_by_line
18:07 juhdanad    Sorry... things_on_line.
18:07 sapier      "things"? :-)
18:07 juhdanad    After all, you get pointed_things
18:08 sapier      hmm true we already use the (non speaking) term "thing"
18:08 juhdanad    And everybody knows the word "thing", not like "raycast".
18:08 sapier      ok why not thing is generic and already used
18:09 juhdanad    Minetest's sets: thing=(node+object=(entity+player))
18:14 juhdanad    sapier: I asked it on the PR's thread.
18:15 red-001     !seen paramat
18:15 ShadowBot   red-001: I saw paramat in #minetest-dev 20 hours, 22 minutes, and 15 seconds ago saying "indeed index calculation doesn't check bounds or out-of-volume"
19:02 red-001     nrzkt, could you make ModStorage userdata like the player object?
19:03 red-001     then you could use rubenwardy's suggestion
19:03 red-001     minetest.get_mod_storage() will just return the userdata/object
19:06 juhdanad    Could you help me? I would like to make a mesh buffer that has a reference to its material. This would require copying and modifying CMeshBuffer.h from Irrlicht. How do I do that legally?
19:16 nrzkt       red-001, it's not as simple as player object, if i need to do that i need to instanciate a new object, but it's an interesting thing
19:16 red-001     I though the player object did that?
19:17 red-001     ever player object refers to a diffrent player?
19:17 red-001     every*
19:19 red-001     how do I only rebase one commit?
19:19 juhdanad    cherry pick it.
19:19 sapier      juhdanad: do not modify irrlicht code ;-)
19:19 juhdanad    Why not?
19:20 sapier      because you'll never get it merged
19:20 sapier      minetest does not include irrlicht thus any changes to their code are useless ;)
19:20 juhdanad    Every mapblock uses the same grass material, yet each mapblock has its own instance...
19:21 juhdanad    No wonder that mapblock meshes consume a lot of memory...
19:21 sapier      you can change this as long as you modify minetest files only but it's no use to change irrlicht itself
19:22 juhdanad    Yes, I wanted to do that. But thi will be mostly Irrlicht code, I just make the material member apointer.
19:23 sapier      what exactly do you mean with "irrlicht code" most ppl don't compile irrlicht themselfs but just use the distribution version thus those changes would not be there for them
19:23 juhdanad    That I copy-paste unchanged methods into Minetest's code.
19:24 red-001     wow why did I do that before....
19:24 juhdanad    It is inevitable, since the only way to remove a member from a class is to re-define it.
19:24 sapier      uarg ... I know this is already done sometimes but that's worst thing you can do
19:24 red-001     that was so much easier
19:24 sapier      it's gonna break whenever irrlicht changes it's internal structures
19:25 sapier      of course it'd be even more easy to include irrlicht but it's still worst solution
19:25 red-001     rebased #5113
19:25 ShadowBot   https://github.com/minetest/minetest/issues/5113 -- [CSM] Add function to get the name of the player and functions to colour chat. by red-001
19:25 juhdanad    Well, I'll try anyways...
19:27 sapier      I strongly encourage you to find a way not to copy code ;-)
19:32 nore        sapier: well we did that already for colored chat I remember
19:32 nore        and yes, it was ugly
19:32 nore        so, if you have another solution, please find one
19:32 nore        but if there is no other way, be aware it has already been done
19:33 nore        copying an irrlicht class and hacking its own code
19:33 sapier      I know it's been done multiple time but doing bad things multiple times doesn't make them better
19:33 nore        (and due to Irrlicht compiled without rtti, it was even more awful)
19:33 nore        sapier: yes, of course
19:34 sapier      the more we copy the more we risk incompatibility to new code and it depends what type of code you copy gui code for example is way more separated then rendering code (usually)
19:37 rubenwardy  juhdanad, iirc Irrlicht is MIT
19:37 rubenwardy  so should be fine
19:37 sapier      it's zlib but that's fine too
19:40 juhdanad    I don't want to break anything! I just create a new class that derives from IMeshBuffer.
19:41 sapier      just be carefull ;-)
19:43 juhdanad    How can I measure the memory usage of the meshes?
19:44 Fixer       vram usage?
19:44 juhdanad    No, RAM usage.
20:09 red-001     nrzkt #5113
20:09 ShadowBot   https://github.com/minetest/minetest/issues/5113 -- [CSM] Add function to get the name of the player and functions to colour chat. by red-001
20:10 red-001     you can remove the rebase need tag
20:19 red-001     so should I make a PR to remove the io library so the security is finished?
20:20 sapier      well finished is a big word but it'd ne another step towards it
20:21 rubenwardy  there's also dofile etc
20:21 red-001     well thats not possible
20:21 rubenwardy  VFS
20:21 red-001     well I guess
20:21 rubenwardy  worth leaving that for later though
20:21 rubenwardy  removing io is much easier
20:24 red-001     I think it should wait for server to client code transfer
20:26 nrzkt       i also think server to client should not appear too fast
20:26 nrzkt       in fact not with 5088
20:26 nrzkt       first have a working CSM purely client side, with a mod store
20:26 sapier      true we should add that feature after client side scripting has prooven to be save
20:27 nrzkt       improve API client side, adding features from client, interaction with local Hud
20:28 sapier      still if we don't add it within reasonable time ppl will abuse mechanisms present e.g. abuse chat for sending code
20:28 red-001     true
20:29 nrzkt       sapier, hmmm yeah it could be a wrong thing, but chat should not permit to execute code directly
20:29 red-001     nrzkt, someone could make a mod to do that
20:29 sapier      well imho that's nothing to be discussed about but an absilute requirement
20:30 sapier      yet you can always encode code in something pure text and decode it by a special mod
20:30 nrzkt       yeah, but without cleint mod you will have a very ugly user experience, and it's against the server owner
20:30 nrzkt       then very risky for server owner to do it, as he can break his user's experience
20:30 sapier      you just have to write a simple client mod receiving and decoding chat
20:31 sapier      user won't even notice
20:31 sapier      client side mod sits in between chat handler and can hide the code communicatio
20:31 sapier      n
20:32 nrzkt       yeah, but without the mod, message will not be ate and shown to user's chat
20:33 nrzkt       then bad user experience, and this restraint the server owner's to do shit a little bit :p
20:33 nrzkt       but i agree we should think about it
20:33 sapier      yes so mod will fitst send a message "hey i'm installed at user x's client" and only those clients get the code messages
20:34 sapier      if done correct noone will see this unless looking at network traffic
20:34 red-001     sapier, mods can't send chat
20:34 sapier      can they show formspecs?
20:34 nrzkt       CSM doesn't permit chat message sending, especially to prevent bots and flooding
20:34 rubenwardy  to be useful, CSM need to be able to send info to server. And vise-versa
20:35 nrzkt       rubenwardy, yeah, but not chat, chat is for human, not bots
20:35 sapier      @rubenwardy not in first step
20:35 nrzkt       i prefer mod channels to discuss between mod users
20:35 nrzkt       and server acts as a bridge between mod users
20:36 sapier      I don't even think communication as of in game communication would really provide what csm is supposed to do
20:36 sapier      it'd suffer of same issues current minetest suffers LAG ;-)
20:36 sapier      csm should be able to do things without talking to server
20:36 nrzkt       sapier, with mod channel communication, you don't need server modding, just retransmit messages to subscribers, it's a pure C++ and lightweight usage
20:37 nrzkt       example of usage, minimap player detection shared between players
20:37 sapier      a very very very dangerous feature
20:37 sapier      direct connection between scriptable client engines
20:37 nrzkt       it's used un world of warcraft massively
20:38 red-001     sapier, why is client <-> server communication bad?
20:38 nrzkt       you should only limit the message size and maybe words too
20:38 red-001     or do you mean client <-> client?
20:38 sapier      red-001 it's not bad but I don't see the benefit for client <-> server (in game) communication it's still limited by lag
20:38 sapier      bad is client <--> client
20:39 nrzkt       not bad, it permit to share some informations between clients, but should be very very limited
20:39 sapier      nrzkt: you can't
20:39 nrzkt       imagine a calendar formspec, shared between mod owners
20:39 sapier      do this in server
20:39 nrzkt       it's not mandatory server side
20:40 sapier      never allow a client to send any data to another client
20:40 sapier      not even tunnled through server
20:40 nrzkt       sofar, maybe you have more informations to give to us about that
20:40 sapier      problem is if there's any bug in the sandbox any malicious client can attack the other client
20:40 nrzkt       sapier, the client <-> client channeling feature is not mandatory for CSM release, we don't need it at first
20:40 sapier      if you do not allow this you just have to fix the server to prevent it
20:41 sapier      I know but client <--> client communication was suggested and imho we should strongly discourage anyone who wants to do this
20:42 sapier      it's dangerous you open up a direct attack vector to our sandbox for a unknown number of clients
20:43 nrzkt       i need to re-study blizzard API for this, there are soem things to learn on implementation, and, i repeat, at CSM start it's absolutely not required
20:43 nrzkt       we only want to manipulate client from client
20:43 nrzkt       itself
20:43 sapier      yes that's first step
20:44 sapier      first step no server <-> client communication and for sure no client <-> client communication (talking about lua engines only)
20:48 sofar       it's optional imho
20:49 sofar       I don't see the dangers per se
20:49 sapier      define "it"? ;-)
20:49 sofar       client-client messaging
20:49 sapier      well sofar think about some buggy mod which could be exploitet to execute arbitrary code
20:50 sapier      if you have client client communication every connected client could do this
20:50 sofar       can always happen irregardless
20:50 sofar       if a server can send data to a client you can get that
20:50 rubenwardy  I don't see the use of c<->
20:50 rubenwardy  c
20:50 sapier      yes but it's a difference user can decide to connect to a server
20:50 sofar       for instance, in WOW specifically
20:50 sapier      but user cannot decide whom other is connected to the server
20:50 sofar       it allows loot sharing information between clients directly
20:51 sapier      and server does not have any chance to protect it's clients
20:51 sapier      wow is different
20:51 sofar       doesn't matter
20:51 sofar       client A has a mod which logs all received loot
20:51 sapier      those guys do have a valid testing AND even more important they do have update mechanisms in place
20:51 nrzkt       sapier, it's not different, wow also permit to write to disk and also uses lua
20:51 sofar       client B requests the data from client A
20:51 sofar       client A sends the loot to client B
20:52 sofar       (it all goes through the server anyway, even in wow)
20:52 sapier      as I said they are different the "solved" or at least reduced the impact by controling the code on client side
20:52 red-001     does WoW have issues with rec?
20:52 sofar       pipe dream, anyone can edit their client to add more client mods
20:53 sapier      well I don't know what wow does "through server" isn't same as "through server"
20:53 sapier      there's a major difference if server just relays data or controls content of data
20:53 nrzkt       sapier, you can cheat on wow by modifying memory to unlock blizzard lua and it works
20:54 sapier      yes you can and if you do and get cought your account will be closed and your money is lost (not sure how big chances are)
20:54 sapier      well what's worst to happen on minetest server?
20:54 nrzkt       sapier, i did it at a moment 4 years ago, no problem
20:54 nrzkt       it depend on what you are doing
20:54 rubenwardy  you would get "banned" and return the next day with a different username and ip
20:55 sapier      exactly yet there's nothing minetest server owners could do nor would there be any risk for criminals to be cought
20:55 sapier      both things are usually present for wow
20:55 nrzkt       sapier, and cheating Lua API is ofter used by many top guilds ...
20:55 sofar       client-to-client bitcoin transerv
20:55 sofar       client-to-client bitcoin transfer*
20:55 nrzkt       sofar, client to client bitcoin mining :p
20:56 sapier      see there's a  major difference if you're blizzard with a lot of lawyers testers and things like that or if you're a small minetest server owner
20:56 red-001     sapier, WoW is a way bigger target
20:56 sapier      of course wow is a more valuable target but a way more risky one too
20:57 sapier      on the other hand minetest has exactly zero risk while providing a little benefit
20:57 nrzkt       say that to chinese cheaters and gold farmers :p
20:57 sapier      especially as it's open source too so bugs are readable
20:57 nrzkt       it's why we should have a strong API when it will come
20:58 sapier      I'll never understand why ppl got used that much to running unverified unknown code but obviously big companies managed to make ppl accept it
20:59 sapier      you can call me pessimistic but I believe we'll never have a sandbox hard enough to be capable of surviving direkt attack
21:03 red-001     dunno I think it's different with lua
21:47 red-001     #5133
21:47 ShadowBot   https://github.com/minetest/minetest/issues/5133 -- [CSM] Block access to the `io` library by red-001
22:06 red-001     could someone review it?
22:16 red-001     !tell nrzkt could you review #5133
22:16 ShadowBot   red-001: O.K.
22:38 ShadowNinja New threading fix: #5134 (sfan5?)  What do you think about the alternative solution?  It would be a bit more complicated but it may be faster when not building with C++11.
22:38 ShadowBot   https://github.com/minetest/minetest/issues/5134 -- Fix synchronization issue at thread start by ShadowNinja
22:46 ShadowNinja Oh, there's also another tiny thing I noticed: The AIX conditional block seemed te be broken (people play Minetest on AIX?  I guess not since nobody noticed.).
22:47 red-001     AIX?
22:49 ShadowNinja ~g AIX
22:49 ShadowBot   ShadowNinja: IBM AIX, UNIX for IBM Power Systems - IBM AIX Power Systems operating system- Offers the highest level of performance, security and reliability of any UNIX operating system. </l/?kh=-1&uddg=http%3A%2F%2Fwww.ibm.com%2Fsystems%2Fpower%2Fsoftware%2Faix%2F>, IBM AIX - Wikipedia - AIX (Advanced Interactive eXecutive, pronounced / ˌ eɪ aɪ ˈ ɛ k s /) is a series of proprietary Unix (5 more messages)
22:49 sapier      ?
22:50 sapier      maybe we should create a c64 port ;-)
22:50 sfan5       ShadowNinja: i'm guessing the threading impl. was initially copied from somewhere with aix support
22:53 ShadowNinja Probably not. kwolekr added that code, but he based his rewrite on my rewrite.  That's a minor issue though.  Fine if I go ahead and merge that little commit now (Maybe we'll remove all the esoteric OS support code later).
23:29 paramat     #5109 simple PR well tested
23:29 ShadowBot   https://github.com/minetest/minetest/issues/5109 -- OpenAL sound: Use a simpler distance model by paramat