Time Nick Message 01:23 Megaf rubenwardy: I want to trace whats causing a memory leak when players join 01:23 Megaf will gdb help me with that? 01:23 rubenwardy no 01:23 rubenwardy valgrind will help you with actual memory leaks 01:24 rubenwardy but it won't tell you if it's correctly allocated stuff, say chunks 01:24 rubenwardy *mapblocks 01:24 rubenwardy only if it's an actual memory leak 01:24 Megaf without a single mod, and I mean 0 mods, not even player and hearts loaded, it still crashes 01:24 Megaf ERROR[Main]: ServerError: AsyncErr: ServerThread::run Lua: OOM error from mod '*builtin*' in callback on_joinplayer(): not enough memory 01:24 rubenwardy oh right 01:24 Megaf 2017-12-13 01:18:35: ERROR[Main]: Current Lua memory usage: 907 MB 01:24 rubenwardy that's a Lua issue 01:25 Megaf I'm using LuaJit 01:25 rubenwardy that's a lot of memory for no mods 01:25 rubenwardy you sure? :O 01:25 Megaf now, I want to know what to blame, LuaJIT, my compiler flags or a bug in minetest 01:25 rubenwardy that is to high a value for having no mods 01:25 Megaf rubenwardy: I did rm -rf * in the mods folder 01:25 rubenwardy I'd expect that with a lua mapgen 01:25 rubenwardy are you using MTG? 01:26 Megaf And that's what I said memory leak 01:28 Megaf rubenwardy: valgrind makes the thing run very slow 01:28 rubenwardy yup 01:28 rubenwardy that's to be expected 01:28 rubenwardy it's counting every single allocation you make 01:29 rubenwardy and every deallocation you make 01:30 rubenwardy that's probably an gross oversimplification 01:30 Megaf waiting for the server to get killed now 01:31 Megaf rubenwardy: memory use is fluctuating between 300 MB and 900 MB 01:31 * twoelk suspects murder 01:31 Megaf very interesting 01:31 rubenwardy lol 01:31 rubenwardy hmmm 01:32 Megaf Not sure kernel will OOM lua with that behaviour under valgrind 01:32 Megaf anyway. So far this is what I have http://paste.debian.net/plain/1000477 01:33 Megaf total used free shared buff/cache available 01:33 Megaf Mem: 3938 1327 131 241 2479 2088 01:33 Megaf normal mem use in this system is 100 MB without minetestserver, it's my new dedicated 01:34 Megaf rubenwardy: memory fluctuation http://paste.debian.net/plain/1000478 01:36 Megaf will rebuild without luajit 01:36 Megaf after I manage to shut it down 01:37 Megaf ==24131== HEAP SUMMARY: 01:37 Megaf ==24131== in use at exit: 1,735,842 bytes in 6,996 blocks 01:37 Megaf ==24131== total heap usage: 132,091 allocs, 125,095 frees, 48,539,666 bytes allocated 01:37 Megaf ==24131== possibly lost: 1,600 bytes in 5 blocks 01:37 Megaf ==24131== ERROR SUMMARY: 43886 errors from 18 contexts (suppressed: 0 from 0) 01:38 Megaf rubenwardy: any word before I recompile the server? 01:40 rubenwardy I doubt it's a compilation issue 01:40 Megaf I will use built in lua now 01:40 rubenwardy ah 01:40 rubenwardy there's LuaJit GC64 or w/e 01:40 rubenwardy which allows a higher limit 01:40 rubenwardy but using 900MB with MTG is very suspicious 01:41 Megaf on my Linode VPS server actually manages to recover from the leak, as it only happens when people join 01:41 Megaf after that memory drops a lot, 01:42 Megaf Anyway, rebuilding using builtin luajit 01:42 Megaf it will just take a while because I rebuild lots of things 01:42 rubenwardy I don't see anything in MTG which could get anywhere close 01:42 Megaf rubenwardy: not even once I blamed minetest/subgame 01:42 rubenwardy could do print(dump()), and stream to a file 01:42 rubenwardy will be BIG 01:42 Megaf I'm just investigating 01:44 Megaf That's quite interesting http://paste.debian.net/plain/1000480 02:01 Megaf rubenwardy: memory leak persists with built in lua 02:02 Megaf let me try with a clean git clone from backport-0.4 subgame 02:03 Megaf persists 02:08 Megaf rubenwardy: so, no mods at all = memory leak, built in lua = memory leak, standard clean MTG = memory leak 02:08 Megaf ideas? 02:09 rubenwardy nope 02:09 rubenwardy not enough info, and I can't reproduce 02:09 rubenwardy also, how are you measuring memory? 02:09 Megaf [01:34:43] rubenwardy: memory fluctuation http://paste.debian.net/plain/1000478 02:09 Megaf and watching htop 02:09 Megaf basically one of minetestserver threads keeps eating 100% of cpu 02:10 Megaf and the process eats up to 1.2 GB of RAM 02:10 Megaf before dropping to arround 300 MB and going up again in a loop 02:12 rubenwardy need to measure lua memory 02:13 rubenwardy collectgarbage().count 02:13 rubenwardy collectgarbage().count() 02:13 Megaf running clearobjects now 02:13 Megaf then I will vacuum the db 12:40 pipo hey folks, what going on on minetest server list? is that an attack for phishing passwords? 12:40 pipo a lot of servers got two entries in the list - many of them got a second entry like "minetest.fun:" - many of them have the same user online! 12:41 Megaf sounds like a bug 12:42 Megaf http://servers.minetest.net/ looks good though 13:04 Megaf maybe not a bug 13:06 pipo now they removed all "minetest.fun:" ant we have "minetest.gams:" entries 13:06 pipo [14:02] <> i still think someone tries to get user passwords by duplicating known servers with a differnt ip-adress:port 13:06 pipo [14:03] <> if players type in thier password on one of the faked servers it will be sent to that wrong address 13:06 pipo [14:03] <> on the new servers its only a hash, that is send ofc 13:07 pipo look, most of the "minetest.games:" entries have the same users online and all are using the same mapgen 13:10 Megaf It is an attack, yes 13:11 pipo any way we could warn players NOT to sent thier passwords to the faked server addresses? 13:22 crazyR might be a safe option to shutdown the list untill they have all been filtered out 13:23 crazyR maybe its time to make server owners register there servers before they go on the list. so that phishing can be tracked 13:24 ThomasMonroe thats a good idea 13:25 crazyR @sfan5 whats your view on this 13:26 Megaf [13:23:50] maybe its time to make server owners register there servers before they go on the list. so that phishing can be tracked 13:26 pipo sfan5 is marked away 13:26 Megaf agreed 13:26 crazyR pipo i know but he will see this when he comes back. 13:27 pipo ok 13:28 crazyR an even easier way would be to link minetest forum accounts to server urls.. 15:43 jas_ game#1977 15:43 ShadowBot https://github.com/minetest/minetest_game/issues/1977 -- Update dirt footstep sounds by jastevenson303 18:51 Wuzzy hey. does anyone think whether #4484 is important enough to be added to 0.5.0 milestone? 18:51 ShadowBot https://github.com/minetest/minetest/issues/4484 -- Player immortal while attached 19:58 Fixer oh yes 20:21 Krock Wuzzy, also the selection boxes seem to be on the wrong position when they're attached. will have to fix the position before that can be changed 20:21 Krock > working on it < 20:22 Wuzzy thx 22:43 Fixer damn lighting after WE https://i.imgur.com/wqdS9z6.png 22:44 sfan5 /fixlight, i don't even know whether this qualifies as a minetest bug 22:54 Jordach isn't that the smooth lighting on mapblock bug? 23:13 Fixer fixlight fixed it after second attempt, first one made it much worse 23:28 Fixer anyone knows, is this intended? https://i.imgur.com/SPeG9er.png 23:38 sofar sadly, I think it is 23:39 Fixer issued a bugreport