Minetest logo

IRC log for #minetest-dev, 2017-04-05

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

All times shown according to UTC.

Time Nick Message
00:01 benrob0329 http://semver.org/
00:04 Warr1024 joined #minetest-dev
00:13 cheapie How is the settings thing for CSM supposed to work? It says "It can be created via `Settings(filename)`.", but settings, Settings, minetest.settings, and minetest.Settings are all nil.
00:15 cheapie VanessaE: I can suggest some database tweaks that might help with rollback, but I think the map editing part is what takes forever.
00:18 cheapie VanessaE: "CREATE INDEX `actionActorIndex` on `action`(`actor`,`timestamp`);" *might* help.
00:19 cheapie (the index name can be whatever you want, of course)
00:19 VanessaE that looks similar to the command nrz mentioned
00:19 VanessaE cheapie: looks like you can also do:
00:19 VanessaE "minetest.setting_get(name) and minetest.setting_getbool(name) can be used to read custom or existing settings at load time, if necessary"
00:20 cheapie From minetest.conf, yeah, and read-only. I need to store settings.
00:20 VanessaE hm
00:21 VanessaE maybe,   foo=settings(filename)    foo:get(key)  and such?
00:21 cheapie VanessaE: I haven't looked too deeply into it, but I'd assume /rollback's query is probably selecting by actor and timestamp.
00:21 cheapie ...that's what I tried, but "settings" is nil, so it crashes.
00:21 VanessaE hm
00:21 TC01 joined #minetest-dev
00:22 cheapie I think rollback.sqlite used to have an index on actor, but got a bit of "tunnel vision" when I was fixing /rollback_check and didn't think that /rollback could use that index.
00:23 * cheapie looks
00:23 cheapie Heh, looks like I was still the last person to touch the rollback DB schema.
00:24 cheapie Anyway, it does appear that there were two separate indexes before that, one on action and one on timestamp.
00:25 cheapie Either one would probably help /rollback, although you most likely won't get any additional gains from having both. Having one index that indexes both together (as I pasted above) is likely to be fastest.
00:26 cheapie "actor" could probably also be dropped from the existing index (the one for /rollback_check), but it's not likely to matter a whole lot.
00:30 cheapie Putting an index on the "name" column of the "actor" table would possibly lead to a slight speed boost on servers with tons of players, but it's most likely not worth it.
00:31 cheapie The "node" table is only ever queried by "id", and that's the primary key which I think implies it's indexed.
00:50 benrob0329-phone joined #minetest-dev
00:55 VanessaE #5517
00:55 ShadowBot https://github.com/minetest/minetest/issues/5517 -- Feature request: specify alternate node for display, and scale the displayed mesh
01:27 benrob0329-phone http://m.imgur.com/a/909KD
01:42 Player_2 joined #minetest-dev
01:46 paramat joined #minetest-dev
01:49 ssieb joined #minetest-dev
01:54 nnnn20430 joined #minetest-dev
02:41 turtleman joined #minetest-dev
03:08 benrob0329 joined #minetest-dev
03:22 octacian What is a debug build? Is it not the same as pulling the latest master and compiling?
03:23 benrob0329 debug would have extra symbols to help debuggers to find variables n such
03:23 benrob0329 its also a faster build (less optimised)
03:24 octacian Because apparently minetest.get_player_information only returns the client version on debug builds, which is REALLY important to know for my mod
03:24 benrob0329 perhaps that should be moved to all builds?
03:24 octacian It really should
03:25 octacian As older clients have chat prediction while newer ones don't, my mod would cause all chat messages to be duplicated for the sender if they were using an older client
03:26 benrob0329 I found a bug
03:26 benrob0329 http://imgur.com/a/eGz7V
03:26 benrob0329 its known, but i havent seen it like this yet
03:27 paramat translucent tardis textures?
03:29 paramat but then, the top seems opaque
03:32 benrob0329 the texture is opaque
03:32 benrob0329 but i do have use_texture_alpha set to true
03:33 benrob0329 I have a question
03:34 benrob0329 Would the idea of moving to Antarctica be, even entertained by the devs?
03:35 benrob0329 Some of the STK devs sound willing to help (I dont know to what extent) with the move, but i dont know if it would even be possible at this moment
03:38 QwertyDragon joined #minetest-dev
03:40 benrob0329-lapto joined #minetest-dev
03:44 paramat ?
03:45 benrob0329- irc is not happy currently
03:52 kaeza octacian, a always, version-based checks are bad. I would suggest something like get_client_features() or such (like core.features, but for things strictly client-side)
03:59 diemartin joined #minetest-dev
04:10 benrob0329 nobody?
04:29 benrob0329-phone joined #minetest-dev
04:34 xtremehacker_ joined #minetest-dev
04:50 lhofhansl joined #minetest-dev
04:51 lhofhansl Hi... Does anybody have recommendation about how to do cross-platform read/write locks for Minetest (since Boost is not allowed)?
04:52 octacian !tell kaeza I couldn't find any documentation for get_client_features(). Do you have a link? If all else fails, I'll play around with it later and figure out what it returns.
04:52 ShadowBot octacian: O.K.
04:53 lhofhansl A lot of the jitter comes from the calling the voxelmanipulator from the main thread when new blocks are loaded from the server. In order to do that in another thread efficiently I'd need read/write locks; but they'd have to work on all platforms that Minetests supports.
04:57 Player_2 joined #minetest-dev
04:57 paramat if only hmmmm were here he might know
04:58 lhofhansl I can build read/write locks on top of the mutexes we have, or on top of the semaphores. Not necessarily the most efficient.
04:58 paramat if no-one can help his email is available
05:00 paramat celeron55 might be able to help?
05:02 paramat will merge #5513 #5515 in an hour, trivial
05:02 ShadowBot https://github.com/minetest/minetest/issues/5513 -- Mapgen: Remove '#include treegen.h' from non-mgv6 mapgens by paramat
05:02 ShadowBot https://github.com/minetest/minetest/issues/5515 -- Mapgen documentation: Add descriptions to noise parameters by paramat
05:05 Hunterz joined #minetest-dev
05:30 lhofhansl left #minetest-dev
06:03 nerzhul joined #minetest-dev
06:04 epoch joined #minetest-dev
06:25 burli joined #minetest-dev
06:28 paramat merging #5513 #5515 in a moment
06:28 ShadowBot https://github.com/minetest/minetest/issues/5513 -- Mapgen: Remove '#include treegen.h' from non-mgv6 mapgens by paramat
06:28 ShadowBot https://github.com/minetest/minetest/issues/5515 -- Mapgen documentation: Add descriptions to noise parameters by paramat
06:31 nerzhul paramat, why do you add many useless slashes in your issues/PR ?
06:32 paramat it's a nice divider :] merging
06:33 nerzhul it's just ugly and divide nothing when it's at the end of your comment :)
06:37 paramat merged. yes it defines the end of the commit message, but i get into the habit and add it when nothing comes after
06:39 paramat the mapgen docs were a mess, missing descriptions, missing flags, finally this unpleasant job is done, large amounts of white text make my head spin, coloured code is much nicer
06:42 nerzhul we have a nice HTML + CSS interface it's not vim
07:03 paramat just found some highlight modes in gedit, will try some
07:16 kilbith joined #minetest-dev
07:46 nrzkt joined #minetest-dev
09:02 Megaf[m] left #minetest-dev
09:15 Megaf joined #minetest-dev
09:51 proller joined #minetest-dev
10:03 nrzkt omfg
10:03 nrzkt LocalPlayer is NEVER removed
10:04 nrzkt i didn't found any palce which remove it and a breakpoint on destructor shows we never call it
10:08 Fixer joined #minetest-dev
10:12 nrzkt joined #minetest-dev
10:15 nrzkt i test the fix now as it's very trivial and push the fix to master when ready as it's just a delete
10:17 behalebabo joined #minetest-dev
10:21 nrzkt joined #minetest-dev
10:24 paramat good find
10:25 nrzkt yeah it's not very critical, but i need to test it
10:31 nrzkt okay it works as inteneded without any side effect, pushed
10:31 nrzkt https://github.com/minetest/minetest/commi​t/54ffc4e915fcdb6f6e7556bc2d1f8b27fd567d9f
10:35 paramat +1
10:41 proller joined #minetest-dev
11:18 YuGiOhJCJ joined #minetest-dev
11:29 Taoki joined #minetest-dev
11:39 blaze joined #minetest-dev
11:44 lisac joined #minetest-dev
11:53 Megaf joined #minetest-dev
11:55 Darcidride joined #minetest-dev
12:10 Player_2 joined #minetest-dev
12:18 YuGiOhJCJ joined #minetest-dev
12:22 nrzkt joined #minetest-dev
13:28 nrzkt joined #minetest-dev
13:36 YuGiOhJCJ joined #minetest-dev
13:41 Megaf joined #minetest-dev
13:41 epoch joined #minetest-dev
13:43 RobbieF joined #minetest-dev
13:53 Darcidride joined #minetest-dev
14:05 DS-minetest joined #minetest-dev
14:13 octacian joined #minetest-dev
14:32 XtremeHacker joined #minetest-dev
14:34 proller joined #minetest-dev
14:41 proller joined #minetest-dev
14:43 benrob0329 joined #minetest-dev
14:51 Taoki joined #minetest-dev
15:08 twoelk joined #minetest-dev
15:32 Megaf joined #minetest-dev
16:44 Hunterz joined #minetest-dev
17:02 twoelk|2 joined #minetest-dev
17:35 Krock joined #minetest-dev
17:58 kilbith joined #minetest-dev
18:06 ssieb joined #minetest-dev
18:11 Fritigern joined #minetest-dev
18:16 troller joined #minetest-dev
18:20 nrzkt joined #minetest-dev
18:31 Player_2 joined #minetest-dev
18:44 troller joined #minetest-dev
18:57 rubenwardy joined #minetest-dev
19:09 nerzhul joined #minetest-dev
19:15 Taoki joined #minetest-dev
19:26 Megaf joined #minetest-dev
19:28 nerzhul we need a second approval on #5480 please
19:28 ShadowBot https://github.com/minetest/minetest/issues/5480 -- [CSM] Localplayer api lua by Dumbeldor
19:29 nerzhul sofar, you want it :)
19:29 sofar sorry, I've been trying to get to testing those
19:29 sofar but i spent 2 days on entity_ai refactoring
19:29 nerzhul sofar, ambitious
19:29 sofar maybe tonight I'll review the CSM patches
19:29 nerzhul maybe you should review my players to db too then
19:30 sofar well, I want to at least test all of them
19:30 nerzhul it seems the 0.4.16 will be a huge release, bigger than previous we did last two years
19:31 nerzhul CSM rampup, performances fixes, light fix, sneak controversy, players to database (i hope), UI enhancements
19:31 sofar yup
19:31 sofar CSM is getting close to really needing server sending of mod code
19:32 nerzhul i really think we should release MT without it first
19:32 nerzhul then maybe think about a release date next month ? before 15 may ?
19:33 nerzhul this will permit to cover existing part with most user feedbacks because it's released and properly think about this part without being in rush mode because we need a release in june :p
19:34 sofar idk, I'm all for more releases though
19:34 nerzhul this release also fix breath cheat, interesting code refactors, mod metadata, mesh shading, player attribut backend, item meta
19:34 nerzhul may means 6 months since last release i think it's good
19:34 nerzhul we have many things in this release
19:35 nerzhul just close the remaining PR on CSM permitting to close the second roadmap for CSM and start third after release with more interactions with server (channels, mods from server)
19:35 nerzhul i also want to fix the chat before release
19:35 nerzhul this will permit users to do better client side mods with chat
19:37 nerzhul i also want to ahve players in database for this release, as PR is now complete for classic usage, it only needs support for migration data between sqlite & pg, but it's not mandatory for the beginning
19:38 nerzhul sofar, sfan5 rubenwardy etc, the player to database is an interesting change for server owners, how can we communicate about this ? Should be merge this just before release to have a better communication for all people or should we merge this asap ?
19:38 sofar does it affect running servers?
19:38 sofar if not then we can merge it
19:39 nerzhul not running, but after a restart an automatic upgrade will be performed to default backend (sqlite)
19:39 sofar well that's something we should communicate then
19:39 nerzhul then if you want pg you need to configure server before restart (world.mt)
19:39 sofar right, then for sure
19:39 nerzhul i just don't know how to properly communicate to end users, IRC is not sufficient
19:39 sofar forum post
19:39 sofar we don't have much more
19:39 nerzhul yeah nice shot
19:39 sofar could make a formal announcement mailinglist
19:40 nerzhul how do you think about release preparation announcement including this change ?
19:40 nerzhul what*
19:41 benrob0329 joined #minetest-dev
19:41 sofar is paramat going to do the release again? maybe he wants a break :)
19:41 sfan5 can't the migration be optional?
19:41 nerzhul sfan5, no it's mandatory
19:41 sfan5 why
19:42 nerzhul look at the PR https://github.com/minetes​t/minetest/pull/5475/files
19:43 nerzhul files are jsut read for migration (and it's relatively secure, we don't remove it, just move them, permitting to rollback MT version if needed)
19:43 nerzhul but the old serializaiton, loading code was just a card skyscraper
19:43 sfan5 i am asking WHY it isn't designed in such a way that the old "flat file" format can also be used as a backend
19:44 sofar honestly I'm not against it going away
19:44 sfan5 me neither
19:44 nerzhul because files are not reliable and it's a very agressive backend to disks, corrupted files happens sometimes, some server owners complains about files reliability
19:45 sfan5 but it would be better if we didn't force people to switch to sqlite directly
19:45 sofar how hard would it be to keep it (with the risk remaining the same as it is now)
19:45 sofar ?
19:45 sfan5 those are they reasons why the flat files are bad
19:46 nerzhul yeah it's a difficult decision to take
19:46 sfan5 small hint: there are NO reasons why we can't support the old files
19:46 sofar I mean, please tell me that the default for new worlds is sqlite...!
19:46 sofar :)
19:46 troller joined #minetest-dev
19:47 nerzhul sfan5, i don't think it's not possible, in the PR it's just dropped because i rewrote the player loading to be correctly done (the previous loading mode was jsut crappy and works, because of chance ? xD), and didn't re-test old mode
19:47 nerzhul sofar, with PR the default mode is sqlite everytime, because i don't support players to files
19:47 sfan5 yes it's possible and i think we should keep supporting the old way at least for this release
19:48 sofar I agree with sfan5
19:48 Tmanyo joined #minetest-dev
19:48 sofar this is one of those cases where I think even nerzhul agrees :)
19:48 nerzhul okay, it's not a nice design to have (unlike db :p) but we can add it, communicate about you should use it, and switch defnitively on next release ?
19:49 nerzhul next +1 i mean, october*
19:49 sofar I'm fairly sure a lot of server operators will like it
19:49 nerzhul it's reasonable, permits to have feedback about it (especially in doc, the process is stable i'm using it on my server since 2 years)
19:49 sofar plus the code looks pretty complete
19:50 sofar I'm not to worried about the new code in it tbh, I should put it on my "test" list
19:50 nerzhul shit, i removed crap code and i should now look how to re-integrate it in nice code :p
19:51 nerzhul is reasonable to add a warningstream permitting users to know they should switch ?
19:51 sofar that should be easier than before, since now your code is reasonably organized ;)
19:51 nerzhul could be a good communication for future deprecation
19:51 sofar yeah, a warning that they could switch to a better backend is fine with me
19:51 nerzhul we don't have a such processus except in lua
19:52 sofar porting the removePlayerFromDatabase() back to old files should be trivial
19:52 sofar that's just an unlink() call
19:52 sfan5 why are the players in the same database as the blocks if they are differentiated by DATABASE_TYPE_PLAYERS anyway?
19:53 sofar isn't that just how the database class works?
19:53 * sofar not too familiar
19:53 nerzhul sofar, the call was intended to be used for databases, but yeah i can add a removefile support for old backend
19:53 nerzhul sfan5, no it's not same database
19:53 nerzhul we have 2 databases instancied
19:53 nerzhul (class)
19:53 sfan5 yes but it's in the same c++ class
19:54 sfan5 despite storing 2 completely different datasets, in two different databases
19:54 red-001 could you fix the old backend being able to crash windows servers if you are going to re-add it?
19:54 nerzhul not exactly, we have database and playerdatabase interface inside it
19:55 nerzhul sfan5, i was thinking about using same connector for a common database, it's why i choose this design, but it could be enhanced later to have a generic postgresql & sqlite interface and have SqliteMapDatabase using sqlite + mapdatabase interfaces and SqlitePlayerDatabase using sqlite + playerdatabase
19:56 sfan5 sofar: when a "Database" instance is created the DATABASE_TYPE_PLAYERS or DATABASE_TYPE_MAP is used to differentiate what you actually want the class to do; to me it would be more logical to have two different classes
19:56 sofar sfan5: yeah, agreed on that
19:56 sfan5 nerzhul: reusing the utilties is good but in this case i think a different way would be better
19:57 nerzhul okay, then i should split the specialized database in a generic way and a super class using the backend + the mt interface
19:57 sfan5 for example what are you going to put into getBlock() for a class that reads the players for flat files?
19:57 sfan5 nothing?
19:57 sfan5 uh how would that work
19:58 nerzhul you can call it from code yes if you add code for this, it's exact, but it's not the case currently
19:58 nerzhul then okay i will refactor this part :)
19:59 sfan5 you intend on having a DatabaseSQLite and a PlayerDatabaseSQLite + MapDatabaseSQLite that are child classes of that, nerzhul?
19:59 nerzhul yeah
20:00 nerzhul playerdatabasesqlite = DatabaseSQLite + PlayerDatabase
20:00 sfan5 if you can make that work that's great
20:00 nerzhul i can, it's not very difficult, just need to take some time, i will do this now, it's less borring than adding prvious file backend xD
20:01 sofar lol
20:01 sofar I wish I could c++ :)
20:01 * sofar laughs
20:10 turtleman joined #minetest-dev
20:32 nerzhul okay refactor for SQlite is done, need to do same thing on pg :)
20:41 Fixer VanessaE: on newest build: animated lamp from homedecor has no animation: https://i.imgur.com/l4NEpJK.png
20:41 Fixer lava lamp has broken animation after some recent commit
20:57 nerzhul sofar, sfan5 i did the refactor
20:57 nerzhul now we have three base interfaces, Database, MapDatabase & PlayerDatabase
20:58 nerzhul Database_Sqlite inherits from Database, MapDatabaseSqlite from dbsqlite + mapdb and analog for player
20:58 nerzhul also i removed the Database_Sqlite function exposure, it's masked by MapDatabaseSqlite & Player
21:00 VanessaE Fixer: that's the "framed glass" drawtype; the plasma animation is the "liquid" inside the node.  I'm gonna redo that to not use that damn drawtype
21:01 Fixer VanessaE: so not a bug, right>
21:01 Fixer ?
21:01 VanessaE it's a bug.
21:02 VanessaE in the engine
21:02 VanessaE (that lamp hasn't been changed in 2 years)
21:04 VanessaE (maybe the bulb is burned out ;) )
21:04 sfan5 nerzhul: i'll take a look at it tomorrow
21:05 nerzhul no problem, i didn't restore the file backend and i don't have time to reinsert it this evening, i will do it later this week
21:05 nerzhul i'm adding the warning now and preparing it
21:39 Fixer https://github.com/minetest/minetest/issues/5522
21:47 benrob0329 joined #minetest-dev
21:53 Puka joined #minetest-dev
22:04 rubenwardy joined #minetest-dev
22:18 kaeza joined #minetest-dev
23:03 twoelk left #minetest-dev
23:07 diemartin joined #minetest-dev
23:08 benrob0329 joined #minetest-dev
23:16 Warr1024 joined #minetest-dev
23:32 octacian_ joined #minetest-dev

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