Time Nick Message 00:35 MTDiscord it basically just lets you simulate C code and run C libraries tektatietwte 00:35 MTDiscord at least in LuaJIT 01:21 repetitivestrain lhofhansl (if he is around): ... as I mentioned that is the original PR... The initial fix would be to increase server_unload_unused_data_timeout, which defaults to 29s. ... it simply doesn't work, and it is too dicey for a map generator that cannot tolerate errors at all 01:21 repetitivestrain one error of this character can destroy a 5x5 block area of a map 09:42 sfan5 I would be worrying about a mapgen that takes more than 29s to generate a given area 10:13 repetitivestrain 29 seconds is simple: a 29 second lag spike, inopportune swapping, or hibernation would be more than sufficient to induce premature unloading 10:13 repetitivestrain and 40 seconds/chunk is par for the course under PUC Lua 10:15 sfan5 I would also worry about the user's hardware and or the game if 29 second lag spikes ever happen 10:16 repetitivestrain i've personally seen multiple-minute lag spikes on a server that was thrashing desperately 10:16 repetitivestrain it was misconfigured, but no accidental misconfiguration or run of ill fortune should damage the map 10:17 sfan5 your mapgen should fail safely if something unexpected happens 10:17 [MatrxMT] a mod can get stuck pretty badly without it being fatal, and so, yes, not damaging the map is important in that case 10:17 sfan5 but I guess the fault is on the engine here 10:18 sfan5 (though I don't think any map damage happens out of the box with e.g. mgv7) 10:20 repetitivestrain Not possible without visible interruptions in features--since features generated in an incompletely generated chunk might also extend and depend on the contents of neighboring chunks 10:22 repetitivestrain sfan5: it would be worse if the player should chance to build something in a partially generated chunk... 10:22 repetitivestrain at least my mapgen protects all chunks that have not been completely post-processed 10:23 repetitivestrain so that only generated features sustain damage from such engine errors 10:23 repetitivestrain might also extend into* 10:24 sfan5 the player could potentially destroy nodes in the overlapping blocks if there are structures there 10:24 sfan5 but not build something new due to the ignore nodes 10:24 sfan5 the risk of this ever being a problem in practice is practically zero if your mapgen is fast 10:25 repetitivestrain I actually saw this in practice all the time, even on hardware where it generates chunks in 700-1000 ms 10:25 repetitivestrain or I would not have written this patch 10:25 sfan5 what is "this"? 10:26 repetitivestrain The engine repetitiously generating chunks that have already been reported to register_on_generated callbacks after "couldn't grab block we just generated" 10:27 sfan5 I'm not talking about that 10:27 repetitivestrain ah you mean building in partially generated mapchunks 10:28 repetitivestrain i would not be so sanguine, though it's been months since i last attempted anything substantial with the built-in mapgens 10:29 repetitivestrain since ignore nodes should only replace mapblocks that failed to generate, not every mapblock such a mapchunk 10:30 [MatrxMT] ignore nodes are routinely placed when decorations extend into neighbouring ungenerated chunks though 10:31 [MatrxMT] the rest of the block gets filled with ignore temporarily 10:32 repetitivestrain this is supposed to be impossible with my map generator, which doesn't post-process (execute decoration placement in) MapBlocks until all mapblocks that could thereby be overwritten have already been reported to on_generated 10:32 repetitivestrain and larger decorations (structures etc.) are laid out long before the base terrain leaves the emerge thread 10:33 sfan5 the edge blocks will exist anyway and given enough time the player could possibly use a pistonto push a node into them 10:33 sfan5 (assuming mesecons doesn't prevent this) 10:33 repetitivestrain mcl_pistons (we don't use mesecons any longer) respects area protection 10:33 repetitivestrain therefore that should be impossible 10:34 repetitivestrain mobs also will not spawn or move till their surroundings are generated and cannot pathfind or interact with ungenerated MapBlocks 10:34 repetitivestrain cannot pathfind into or interact with ungenerated mapblocks* 10:36 repetitivestrain liquid transforms are the only remaining mechanic that is liable to alter ungenerated mapblocks, and as mineclonia will soon gain a lua-based liquid transformation system, this won't be an issue for much longer 10:39 repetitivestrain caves are largely generated from noise, and during the generation of each mapblock, the generator searches for non-noise cave systems which originate within a distance of 8 16x384x16 columns from the mapchunk being generated and replays their generation within the VM's emerged area 10:40 repetitivestrain which is also luanti's own cave generators's MO, AFAIK 15:06 bracket x 18:45 lmat Good afternoon! I'm running luantiserver and I get http://pastie.org/p/2PLd4o3yDY6r2uaSUFqKAj 18:47 lmat I built from source (master branch as of now) and I did everything the same way I build minetest maybe a year or two ago. I'm assuming "Game [] could not be found." means that I failed to install the "minetest" game into my server (which was installed by default before). What should I do differently? 18:49 MTDiscord you need a copy of https://github.com/minetest/minetest_game in your games directory 18:51 MTDiscord (assuming a run in place build now) then you can run ./bin/luanti --gameid list and that should include minetest_game, and then you can pass --gameid minetest_game when starting the world. 18:59 lmat luantic: thank you! I'll give it a go ':-) 19:07 lmat Worked like butter! I think I didn't need --gameid when starting luantiserver because minetest_game is the only game. Thank you! 19:37 MTDiscord np :) 20:09 SwissalpS /15 20:09 [MatrxMT] what does that mean? random pressing? 20:10 SwissalpS switching channels in irssi, a space snuck in first 22:04 [MatrxMT] Hi there! I was wondering, why does Luanti not use some pretty user i/o library like minecraft does? 22:04 [MatrxMT] I mean, I really like using the MC server console: it has autocomplete and allows you to scroll through the history, which is, afaik, is not possible in CURSES mode in Luanti server. Why is it like that? Is it considered a "useless dependency" or some kind of not very useful think to waste core team's time on? 22:04 [MatrxMT] If it is the second case, I might try to implement such thing... 22:05 sfan5 the immediate reason is that nobody wrote the code 22:05 sfan5 in the bigger picture it has never been a priority and IIRC the curses console was also just an one-off contribution from someone 22:06 sfan5 chat autocomplete has been suggested but has never be done because it would need to be flexible enough to support the different commands mods might want to add 22:07 sfan5 so: contributions welcome 22:07 [MatrxMT] I thought of some flexible api with which you could like describe a command and its arguments so that it could autocomplete 22:08 [MatrxMT] like there would be "classes" of arguments: player names, coordinates, etc 22:09 [MatrxMT] you could write a file that would say, for example: 22:09 [MatrxMT] ban [player] [time](s) 22:09 [MatrxMT] and it would be parsed and when you press tab after "ban", it would understand that you want a players name so it would iterate through player names 22:10 [MatrxMT] But I'm afraid that I'll come with PR and core team'll say "frick you thats a bad idea/implementation" 22:11 [MatrxMT] Huge stone just dropped from my soul (this means "I am very relieved" in case if English doesn't have this phrase) 22:11 [MatrxMT] Thanks a lot 22:11 [MatrxMT] I really wanna contribute to luanti so I guess this will be my start 22:13 sfan5 see https://github.com/luanti-org/luanti/issues/9663 22:18 sfan5 for reasons mentioned there a good balance would probably be if mods could just declare a simplified parameter order to indicate where player names can appear 22:19 [MatrxMT] Why just player names? I think the idea of referencing lua snippets is brilliant, isn't it? 22:20 sfan5 the often mentioned SSCSM does not exist and won't for a while 22:20 sfan5 besides how do you run the client mods on the server to enable autocompletion in the curses console? 22:20 sfan5 that would require lots of special handling if not an entirely different API