Time Nick Message 01:19 hmmmm so people are all of a sudden getting intermittent LUA_ERRERRs 01:19 hmmmm this started happening a few days ago. it can't be a mod error, the internal lua state would somehow need to get corrupted for this to happen. 01:20 hmmmm my guess is that the lua stack got messed up in an API that everybody uses 01:37 hmmmm where is core.luaentities defined? I can't find it in builtin 01:38 hmmmm nevermind, it's created by the core 02:20 VanessaE hmmmm: speaking of corruption, here's today's crash ;) http://pastebin.com/7y8JnASy 02:21 hmmmm always in connection 02:22 VanessaE MichaelEh reports that if he gets too close to the bottom of the world (he says 30800), it'll crash also. not sure of the specifics. 02:23 hmmmm i think it does happen inside the connection send thread. it's way too consistent to be a coincidence. i think glibc uses thread-local arenas for malloc() 02:23 hmmmm michaeleh should make an issue on github :) 02:24 hmmmm in any case I'm pretty sure the error you're seeing there will get fixed when all of the items found in Address Sanitizer are addressed 02:25 hmmmm for the error in error handling I'm thinking it might be prudent to start stashing the mod name somewhere when registering a callback of any sort 02:36 VanessaE sounds like a reasonable precaution 02:38 VanessaE and actually, I did suggest Michael raise an issue 02:41 VanessaE oh, did anyone resolve that "n toggles camera updates in release builds" glitch? 02:42 VanessaE #2762, looks like no. :P 02:42 ShadowBot https://github.com/minetest/minetest/issues/2762 -- “camera update” toggles when pressing N on release builds 02:43 hmmmm i haven't looked at it 02:43 VanessaE ok. you'll probably want to fix that before release. it's easier to accidentally hit that key than one might think :P 03:45 kahrl VanessaE: Wuzzy's guess is probably right 03:45 kahrl judging by this comment in keycode.cpp: // it's not a (known) key, just take the first char and use that 03:48 hmmmm would you like to whip up a patch :)? 03:48 kahrl sure 03:48 kahrl I'm trying what happens if you set the setting to the empty string 03:49 kahrl it doesn't like that :/ https://gist.github.com/kahrl/b3358ae390c438df75db 03:50 VanessaE cheat: set it to some value that's impossible to produce from a keyboard :) 03:50 hmmmm ohhh that's the same crash as the russian keyboard problem 03:53 hmmmm how can you intercept a Lua call to some function? 03:54 hmmmm i'm not really interested in the parameters, but i want to execute something before *every* on_eventhere function 03:54 VanessaE redefine the function 03:55 hmmmm ? 03:56 VanessaE local foo = functionname(...) ; function functionname(...) [ youe code here ] ; foo(...) ; end 03:56 VanessaE (I'm not 100% sure about the ... but that's the basic form) 03:57 hmmmm the ... is the part i'm interested in 03:57 hmmmm could you pass through args using this "..." syntax? 03:57 VanessaE I think so. 03:57 hmmmm hmm i never saw that 03:57 hmmmm what is that called 03:57 VanessaE I see others doing that, but I've never experimented with itr 03:57 VanessaE http://www.lua.org/pil/5.2.html 03:58 hmmmm yea reading that right now 03:58 VanessaE (works in 5.1 too apparently) 03:58 hmmmm good 03:59 hmmmm okay 03:59 hmmmm so you were pretty close 03:59 kahrl this works: https://gist.github.com/kahrl/405c00ec63dd7e293ab1 03:59 hmmmm function foobar(...) ; original_fn(unpack(arg)); end 04:00 VanessaE ah yes, right 04:00 hmmmm kahrl: looks good to me :) 04:00 kahrl shall I push it? 04:00 hmmmm yeah 04:02 VanessaE /home/minetest/minetest_core/src/server.cpp:511: void Server::step(float): A fatal error occurred: Lua: OOM error in on_playerReceiveFields(): not enough memory 04:02 VanessaE wait, what? 04:02 hmmmm VanessaE: what are you doing in on_playerReceiveFields 04:02 hmmmm :p 04:02 VanessaE hmmmm: server crash just a minute or two ago 04:02 hmmmm I read about a LuaJIT limitation that doesn't allow it to allocate more than 2gb of memory 04:02 VanessaE how the fuck does a formspec crash with OOM? :) 04:02 hmmmm but that was in 2010 04:03 hmmmm again, it might not be the formspec exactly, it could be code that goes crazy and tries to allocate an undefined number of things 04:03 hmmmm although I am a bit more hesitant to accept this theory for mods since Lua initializes variables to nil, not an undefined value 04:04 hmmmm so when a variable is equal to some numerical value, there's intent behind it in almost all conceivable cases 04:05 hmmmm VanessaE: FWIW what I'm doing right now will narrow down the problem code even further, so hold on tight 04:05 VanessaE no worries. 04:05 VanessaE meanwhile, here's the rest of the backtrace, just in case: http://pastebin.com/wue5Ge0W 04:05 hmmmm there is never any useful information in the backtraces for lua errors 04:06 hmmmm they're just whatever happened to be the state of the application at the time the ServerThread processes that async fatal error 04:07 VanessaE ok 04:07 VanessaE you know me, I try not to assume too much 04:07 hmmmm that 04:08 hmmmm that's a good trait* 04:09 hmmmm hmm, VanessaE, it seems you can pass ... along literally in the exact manner as you first said, too 04:12 kahrl hmmmm: that's reassuring, since that's what the minetest:env handler does :) 04:18 VanessaE bbl, bed 06:25 hmmmm from my own testing, luajit shits out after roughly 67100000 insertions of integers into a table 06:26 hmmmm i also forced an OOM error this way and it tried checking the memory usage using the offically sanctioned lua_gc() which turned out to be ~ 513 MB 06:27 hmmmm I wonder if this is an mmap + FreeBSD x64 problem 06:27 hmmmm and I wonder if it works better with Linux 06:29 hmmmm coincidentally, 538025170 / 67100000 = 8.01826, so I think each insertion of an integer into a table takes up 8 bytes 06:34 hmmmm in any case I suspect the OOM is coming from inside an lua api call 08:43 cornernote i know this isnt a Q for #minetest-dev, but i asked several times in #minetest with no response... 08:43 cornernote i'd like to show the inventory formsepc when a player clicks a node 08:44 cornernote i can use minetest.show_formspec() to show a copy of the inventory formspec, but when i change the formspec it doesnt change (eg, when i click a button in unified_inventory) 08:44 cornernote any way to do that ? 10:51 kilbith test 10:53 Krock test 10:53 est31 mine-test! 11:35 Zeno` Greetings 11:36 Zeno` #3014 11:36 ShadowBot https://github.com/minetest/minetest/issues/3014 -- Add LuaSecureRandom by est31 11:36 Zeno` ^-- how to "verify that it actually works"? 11:36 Zeno` it looks good from what I saw 11:37 Zeno` spy! 11:38 est31 I haven't tried it yet, perhaps it should be tested to generate some random sequences 11:38 est31 and print them as hex 11:38 est31 to verify they are actually random looking 11:38 Zeno` hmmmm has an application that rates randomness 11:39 Zeno` I wish I bookmarked it 11:39 Zeno` maybe I did... lemme look 11:40 Zeno` hmm, I don't seem to have it... but he used it to assess his new PNRG implementation 11:41 Zeno` basically you spat the numbers to stdout and piped them to whatever the silly thing is called and it'd rate it 12:13 paramat sfan5 any comments approval for game#607 ? it will be split into 2 commits soon 12:13 ShadowBot https://github.com/minetest/minetest_game/issues/607 -- Biomes: Improve biome system and v6 mushroom spawning WIP by paramat 12:15 kilbith yes, shorten those descriptions patterns with : `name:gsub("%l", string.upper, 1)` 12:31 paramat aha okay 12:31 VanessaE don't forget to put in a better, darker jungle wood texture :) 12:31 VanessaE also, good morning. 12:31 paramat O/ 12:32 paramat i see my mistake now, i saw the 'l' as a '1' 12:32 kilbith yes, i guessed that 12:44 paramat i was watching minetest 'lets play' videos and really noticed how bad our grass texture looks. it's dull, grim and depressing, a sickly colour, i will try editing or find a new one 12:45 paramat for junglewood open an issue and gain support, i quite like the light green wood but am not too bothered either way 12:45 est31 the colour is deep healthy green 12:45 kilbith ^ 12:46 kilbith RBA likes it too 12:46 VanessaE why not base your texture off of HDX's grass? (don't copy & scale, I mean the colors and general "feel" of the texture) 12:46 est31 that grass is a photo of real grass, no? 12:46 est31 or is it from your other texture pack? 12:47 VanessaE it is a photo, yes. 12:47 paramat the dark green is popular so i'll try editing to retain that, but with more contrast and saturation, and a less yellow hue 12:47 VanessaE hence why I said not to just use it directly. 12:48 est31 well, i like the "abstracting" part of minetest_game 12:48 est31 once you try to mimick reality, you are in direct competition for realism 12:48 est31 IMO 12:49 VanessaE that's why I said to borrow the colors, not the whole texture :) 12:51 paramat i think i will go ahead and change pinetree -> pine_tree, pinewood -> pine_wood, and add aliases. game#603 12:51 ShadowBot https://github.com/minetest/minetest_game/issues/603 -- Acacia tree/wood : wrong naming convention 12:51 kilbith be consistent and go ahead for the acaccia as well 12:51 kilbith ah no, nvm 12:52 est31 btw whats the state of acacia trees, can they already grow from saplings, and are they generated by mapen? 12:52 est31 as in "are they finished" 12:53 paramat yes 12:53 VanessaE paramat: https://github.com/minetest/minetest_game/issues/611 12:53 paramat generated in mgv5/v7 12:53 paramat okay 12:55 paramat the other 3 trees have 2 versions, a sapling will grow into the classic design in mgv6, and grow into the new schematic version in mgv5/v7 12:56 est31 ok 12:56 paramat also to do: the 4 sapling growing abms can be combined into a sinkle abm on 'group:sapling', to reduce abm load 12:57 est31 VanessaE, bout game#611, what exactly was the "breaking commit"? 12:57 ShadowBot https://github.com/minetest/minetest_game/issues/611 -- Junglewood texture too light, boring 12:57 paramat i will be busy.. 12:58 VanessaE est31: https://github.com/minetest/minetest_game/commit/13dfc2c0e90b9e5f5e7380a6b9a80ce3ebdcd8f3 13:00 paramat note the old tree-top texture wasn't dark brown, that could be corrected 13:00 VanessaE yeah I know 13:00 paramat this would be a chance for a slightly darker, richer hardwood colour 13:00 paramat (if wanted) 13:03 sfan5 paramat: "Remove water below glacier rivers" are those right above the dirt? 13:04 paramat i noticed the icesheet was not as deep as the rivers, so 2 nodes of water was there above 'seabed' gravel 13:05 paramat now ice is 2 nodes deeper so glacier biome 'rivers' are pure ice 13:05 sfan5 i see 13:06 sfan5 paramat: game#607 looks good 13:06 ShadowBot https://github.com/minetest/minetest_game/issues/607 -- Biomes: Improve biome system and v6 mushroom spawning WIP by paramat 13:06 paramat okay will split into 2 commits later 13:08 paramat sfan5 how about game603? 'pine_wood / pine_tree' (and adding aliases) to be consistent with acacia nodes 13:09 paramat i remember shadowninja requesting this 13:10 sfan5 i suggest renaming every one of them to be consistent with acacia_tree and adding aliases 13:11 paramat aha okay, tree and jungletree too, fine by me 13:11 VanessaE agreed; will an alias of an alias still work btw? 13:11 VanessaE (node used to be A, got renamed + aliased to B, meanwhile some mod aliases its old nodes to A) 13:11 paramat (erm i mean jungletree too) 13:13 kilbith we could drop the aliases support for 1 year if necessary 13:14 est31 no 13:14 est31 that would break any structure with pines 13:14 est31 that has been built before 13:15 est31 aliasses dont hurt 13:15 est31 if we have an automated "migration service", I agree 13:15 est31 but removing functionality is completely useless 13:16 VanessaE aliases *are* automated migration, 13:16 VanessaE problem is, it only changes the nodes in parts of the map that people have visited. 13:17 kilbith i said if necessary like if relayed aliases are uneffective 13:17 est31 VanessaE, do aliases actually change the nodes? 13:17 est31 I've thought you need an exta abm for that? 13:17 est31 e.g. like with the stairs abm? 13:18 VanessaE the stairs ABM handles rotating upside down stairs --> 6dfacedir 13:18 kilbith no, that ABM cha,gs the facedir 13:18 kilbith oops 13:18 est31 why is there no signs abm too? 13:18 est31 signs:sign -> signs:sign_new 13:18 est31 or so 13:19 VanessaE last time I added an alias (less then a week ago), I had to keep a backup of the test world to make sure the nodes under test were reverted between changes 13:19 VanessaE why would signs need an ABM? 13:19 paramat aliases don't replace nodes no, the stairs abm does though 13:19 VanessaE paramat: you sure about that? 13:19 est31 VanessaE, havent you changed how signs are displayed in the past, requiring everybody to screwdriver them? 13:20 kilbith the signs_lib yes 13:20 VanessaE est31: that only affected signs_lib because it used 6dfacedir for the wooden wall signs, while default code uses wallmounted. 13:20 VanessaE an ABM wouldn't be useful in that case because I can't safely rename the node, and an alias can't rotate a node 13:21 est31 why cant you rename the node?+ 13:21 kilbith neither an entity 13:21 est31 also, considered changing the param2type? 13:22 VanessaE can't rename the node because then that would break any world you install it in, should you decide to remove it. 13:25 paramat no not sure =) 14:37 paramat hmmmm this seems related to your recent luajit OOM tests https://github.com/paramat/riverdev/issues/1 15:16 hmmmm hrmmm 15:17 hmmmm does riverdev use the memory-efficient API variations? 15:17 paramat not yet 15:18 hmmmm it would be a real shame if this is a luajit problem at its core, but at the same time a mapgen shouldn't require more than 512 MB 15:18 hmmmm I just never imagined that people would use 3d noise to such a heavy extent.. 15:18 paramat lol 15:20 paramat 10 3D noises plus 7 2D noises 15:20 hmmmm but there have been OOM errors in mundane callbacks such as player_setfields or something 15:20 hmmmm there could be an internal problem 15:21 hmmmm also LuaJIT might have problems telling the difference between a runtime error inside an error handler, or a runtime error inside of a lua function call that happened inside of another lua function call 15:22 hmmmm I saw what HybridDog posted to reproduce the double fault 15:24 paramat the riverdev issue could perhaps just be the excessive mesh data buildup caused by exploring 15:24 hmmmm i personally don't believe so... the luajit memory limitation is real 15:25 hmmmm and it's orders of magnitude less than how much memory the user actually has available 15:25 hmmmm having too many meshes wouldn't cause lua to hit some kind of artificial usage limit 15:26 hmmmm well i suppose it mostly depends on /where/ the mesh memory was allocated 15:26 hmmmm yeah, your theory may be plausible i suppose. 15:26 paramat well i dunno 15:27 paramat i can see this being a much needed very long feature freeze 15:29 paramat we need to fix this excessive memory use problem when flying around a map. i tried bisecting and got as far back as 0.4.8, still the same behaviour, just seems this is how MT works 15:30 hmmmm you can fix this by lowering the block unload timeout 15:30 paramat 0.5GB per minute buildup while flying 15:30 paramat i tried that and got unclear results 15:30 paramat need to try again 15:30 hmmmm server_unload_unused_data_timeout, client_unload_unused_data_timeout 15:30 paramat 'client unload unused data timeout'? 15:31 hmmmm probably more the client than the server 15:31 paramat oh i didnt try the server one 15:31 paramat will try that 15:31 hmmmm the client unloads unused blocks after 10 minutes, the server, 29 seconds 15:31 paramat ah 15:32 paramat i have tried reducing the client one to 1 minute and exploring, but saw no reduction in memory use 15:32 paramat seems it removes meshes from the client's view but doesn't remove the data from memory 15:32 paramat AFAIK 15:34 hmmmm they're removed from the client's view because they don't exist in memory anymore 15:34 paramat ah 15:35 paramat i guess the raw world data is kept in memory because it's so compact, only meshdata is deleted / recreated when needed 15:36 paramat then when i exit to menu, i get varying degrees of memory clearout, from none to almost all 15:36 paramat that's a separate problem perhaps 15:39 paramat https://github.com/minetest/minetest/issues/2653#issuecomment-120201932 and the following comment 15:41 paramat anyway i certainly think 'client unload unused data timeout' should be hugely reduced, 10 minutes can be 7GB if flying 15:43 paramat 2-3 mins perhaps 15:47 paramat perhaps it was fixed in freeminer?.. 15:50 rubenwardy A few comments about the android app: 1. The black screen on start up is confusing, there should be an indication of progress. 2. You shouldn't show fly/noclip/fast buttons if the player doesn't have those privs. 3. I expected, in the inventory, to be able to tab an item and then tab where I want it to go - you have to drag them instead. 4. Pressing the back button to open the menu is not intuitive for me, I'd have expected it to exit 15:50 rubenwardy the game. 5. closing menus with the back button breaks the flow, my back button vibrates and takes ~0.5 seconds to do anything. 15:50 rubenwardy Showing the debug info is handy, but should probably be hidden in the pause menu rather than on screen 15:54 paramat will push soon game#612 15:54 ShadowBot https://github.com/minetest/minetest_game/issues/612 -- Biomes: Tune v5/v7 biomes and v5/v6/v7 mushroom distribution by paramat 16:11 paramat now pushing game 612 16:16 paramat push complete 17:13 paramat will push game#613 soon 17:13 ShadowBot https://github.com/minetest/minetest_game/issues/613 -- Flowers: Edit mushroom code, rename spore nodes by paramat 17:13 VanessaE "edit mushroom code"? 17:13 VanessaE you may wanna be more specific there ;) 17:14 VanessaE https://xkcd.com/1296/ 17:14 paramat code cleanup 17:15 VanessaE (note the fourth commit in the comic ;) ) 17:16 paramat yeah, 'cleanup' is better, will change 17:52 wischi2 is mgv6 hardcoded? 17:53 wischi2 I mean the biome definition 17:56 paramat yes 17:56 paramat but there's a choice between classic biomes and a new system with taiga and tundra 17:56 paramat by default snowbiomes are enabled 17:57 VanessaE paramat: define "a new system" please -- is the old mgv6 terrain not compatible with this "new" one? 17:57 wischi2 are v5 and v7 more flexible (at least I found some "register_biome" calls in the v5 lua part) 17:58 paramat VAnessa terrain is unchanged, biomes are moved around and different sizes 17:58 VanessaE hm, ok 17:59 VanessaE I'd really love to see a tool that can strip a map down to just what's been modified since mapgen time 17:59 paramat (moved around a little) 17:59 VanessaE so that it's easier to add new biomes, ores, etc. 17:59 paramat yeah v5/v7 use the biome API so you can define your own biome sytems 17:59 VanessaE nono 18:00 VanessaE I don't mean in code 18:00 paramat sorry that was for wischi2 18:00 VanessaE I mean e.g. run such a tool, then add say moreores, and its new ores appear in all newly-generated blocks (i.e. the stuff that was stripped out gets re-added) 18:01 wischi2 oh my name changed ^^ (damn reconnects) 18:04 wischi wth!? github can't search in commit messages? 18:04 VanessaE nope. 18:04 wischi lol 18:04 VanessaE at least not that I've found. and it's irritating as fuck 18:05 VanessaE so I just `git log |grep -A -B ...` 18:05 paramat now pushing game 613 18:10 paramat push complete! 18:11 wischi @VanessaE: thx 20:00 paramat sfan5 any comments on game#614 ? 20:00 ShadowBot https://github.com/minetest/minetest_game/issues/614 -- Default/trees: Combine sapling ABMs into one ABM by paramat 20:01 sfan5 oh yes 20:01 sfan5 that'll improve performance 20:01 sfan5 paramat: looks good 20:01 paramat cool 20:05 VanessaE paramat: game#611 20:05 ShadowBot https://github.com/minetest/minetest_game/issues/611 -- Junglewood texture too light, boring 20:07 rubenwardy lol 20:07 rubenwardy If you keep mentioning it, he'll take the hint eventually 20:07 paramat +1 20:08 kilbith i just realized that darker was better :P 20:08 VanessaE kilbith: told you it was :) 20:10 paramat looks like blockmen changed the wood colour to match the light green core of the jungle trunk. it would have been better to match the trunk core to the dark wood 20:10 VanessaE agreed. 20:48 Tesseract I've rebased and tweaked my threading pull. 20:50 VanessaE paramat: an argument could be made that a jungle tree's core turns lightly-colored when exposed to air, until it's made into processed wood planks :) 20:52 rubenwardy My bad :'( https://github.com/minetest/minetest/commit/63b12f1f048d83632069a4d0b3d2649cf9be7904 20:53 Tesseract ^ Removed 2992, which has already been merged. 20:53 rubenwardy #2992 #2613 20:53 ShadowBot https://github.com/minetest/minetest/issues/2992 -- Biome API: Make fallback biome stone and water, disable filler by paramat 20:53 ShadowBot https://github.com/minetest/minetest/issues/2613 -- Clean up threading by ShadowNinja 20:53 sfan5 >excempt until aug 4 20:53 sfan5 that doesnt apply anyway 20:53 sfan5 -c 20:53 Tesseract Oh, right. 20:54 kilbith VanessaE: a wood always gets darker when it's cut and drained of water 20:54 Tesseract So, we should probably enable security before release. 20:55 kaeza wont that break mods? 20:57 paramat sure the planks are usually a little darker, but the hue of the current jungle trunk core is wrong so we'll need a new jungle tree top texture with same bark but different core 20:57 kaeza security should be disabled by default until .14(?) 20:57 kaeza imho 20:57 VanessaE Tesseract: definitely no. that's a feature 20:57 VanessaE and as we're in freeze and that will break a couple of mods, definitely not now 20:57 Tesseract kaeza: No, it should only break old versions of a few mods. 21:10 hmmmm Tesseract: You gonna push the threading PR? 21:11 rubenwardy VanessaE, it's already merged and the aim was to enable it for 0.4.13 21:13 VanessaE I still say 'no' 21:13 VanessaE enable it *after* the release 21:13 VanessaE give modders time to figure out how to deal with it 21:48 RealBadAngel too light -> boring. black -> an orgasm :P 21:49 RealBadAngel lets make all the textures black, that would be something original ;) 21:50 VanessaE ... 21:51 RealBadAngel oh cmon, just jokin :P 21:54 sfan5 Minetest 0.4.13 21:54 sfan5 fifty shades of black edition 21:54 VanessaE hah 21:54 VanessaE wait 21:55 VanessaE didn't we already HAVE that edition a couple versions ago? :) 21:55 VanessaE (preload item visuals turning all textures black, for a while)