Time Nick Message 00:02 sapier guess this way round we get all of it 00:03 sapier but to much of a change for this time to trust ... need to check that tomorrow 01:34 OldCoder Hi; question. ShadowNinja says to mention it here. Do worldmods load last or not? 04:38 VanessaE um, wut? 04:38 VanessaE 04:36:49: INFO[EmergeThread2]: WARNING: Could not bind block position for load: 04:38 VanessaE unknown error 04:38 VanessaE 04:36:49: INFO[EmergeThread1]: WARNING: Could not bind block position for load: 04:38 VanessaE unknown errorWARNING: Could not bind block position for load: library routine ca 04:38 VanessaE lled out of sequence 04:39 VanessaE all right g*d damn it someone wake up 04:39 VanessaE ALL FIVE of my servers are doing this now 04:39 VanessaE you guys busted something bad 04:43 VanessaE I just git pull'd. rolling back to end of day Nov 30. pretty sure it worked fine yesterday 04:46 hmmmm actually that sounds like it could be some problem related to the new sqlite rollback 04:47 VanessaE sorry to get upset... 04:49 hmmmm or not 04:49 hmmmm hmmm 04:49 hmmmm i dunno 04:49 VanessaE seems to be okay at its current older state. 05:04 VanessaE ok, they all seem stable now. 05:05 VanessaE so either it was a glitch or it was one of today's (er, Dec. 1st's) commits 05:06 VanessaE also, as a side note: why does the client now stall for a second or so when I click "connect"? (well, I hit 'enter') 05:15 OldCoder Hi; question. ShadowNinja says to mention it here. Do worldmods load last or not? 05:21 * VanessaE pokes hmmmm 05:24 * OldCoder is asleep deep 05:25 OldCoder Zzz 05:50 kahrl VanessaE: the stall on connect is because the async engine does a sleep(1) on exit 05:50 VanessaE ok 12:12 sapier Oldcoder i guess there's nothing special about world mods but I didn't verify this assumption 13:16 OldCoder sapier, ShadowNinja thought there might be. Thank you. 17:55 rubenwardy lol 17:55 sapier :-) maybe some skrewd up indices :-) 17:56 sapier 0.4.8 or current master? 17:56 sapier current master? 17:57 sapier I'm reworking search tab atm ... one of the reasons is broken indices ... did only try search :-) 17:59 sapier even downloading? ... chances I'm responsible for broken master too aren't that small too ;-) 17:59 Calinou both freeminer and Minetest segfault on launch 18:00 sapier debug or release build? 18:00 sapier no did you tell cmake to do a debug or release build? 18:01 PilzAdam is it some kind of universal rule that code by sapier segfaults for at least one person? :-p 18:02 sapier this thing seems to become even more critical sokomine told something similar yesterday ... rebuilt with debug information and error was gone ... those bugs are nasty 18:03 sapier yes index calculation is wrong for those search tab and downloading relies on index of mod ... I'm about to change that 18:04 sapier cmake -DCMAKE_BUILD_TYPE=Debug 18:08 sapier I still hope you'll have same error with debug build 18:08 sapier finding this one in release build is a nightmare 18:10 sapier great 18:11 thexyz when does this happen? 18:11 sapier I assume something with closing threadss 18:11 sapier either hidden by debug symbols or by reduced optimization level in debug build 18:12 sapier wait 18:12 sapier that's not closing threads that's data transfer 18:13 thexyz clang to the rescue 18:14 sapier nothing that can seriously be a reason 18:15 sapier I assume it's a random error my code does a little bit different things so a slightly differen timing 18:17 thexyz segfault right after start https://gist.github.com/xyzz/f9c5276a7ca512f0e409/raw/8709b42a78080308027ec1a0af003028b41c9d03/gistfile1.txt 18:17 thexyz reproducible: always 18:18 thexyz https://gist.github.com/xyzz/afbf89f01bfdbe463b78/raw/ac6305882273e8bf2b87288b714b71aa4081867f/gistfile1.txt 18:18 thexyz yes 18:18 thexyz (always means always with address sanitizer which terminates the program) 18:18 thexyz so it's either segfault or it terminates it 18:19 thexyz hm, maybe 18:21 thexyz sapier: are my logs helpful or shall I provide something else? 18:21 PilzAdam sapier, I can reproduce the crash when clicking on "Online mod store" in release, but not in debug with -O3 18:22 PilzAdam sapier, do you want a bracktrace in release mode? 18:22 sapier not quite sure thexyz It seems to complain when I erase a element in list 18:23 sapier that element is taken from list 2 lines above that line ... complete list handling is done within lock 18:23 sapier yes PA 18:24 PilzAdam https://gist.github.com/PilzAdam/7753958 18:24 sapier a struct is copied by copying all elements in c++ isn't it? 18:26 thexyz unless you redefine = it is 18:27 sapier I didn't do and the struct contains basic types and strings only 18:27 sapier this way a should be a copy of b after I did a = b 18:28 thexyz that's a strange question to ask.. 18:28 sapier :-) I know ... I learned to question even obvious things 18:30 sapier what the hell? do I read this completely wrong or do all threads call erase same time in this stack? 18:33 sapier seems like locking is completely broken ... again 18:34 sapier that's perfectly normal in multithreaded applications pa has a hell of a machine with 8 cores 18:35 PilzAdam I have 4 multithreaded cores, to be precise 18:36 sapier ok then it should be comparable ... but still different background threads os version can make a lot of difference 18:38 sapier yet I don't have any idea how four threads can be within same mutexed area at same time 18:42 sapier could you guys try my wait jthread branch? 18:42 sapier I found some posts via google telling mutex and detached threads may be buggy on some libc versions 18:45 Exio4 i have six threads and 3 modules, what could i try? ;P 18:46 sapier I guess this one needs to be found in theory debugger doesn't help that much 19:58 sapier I just updated the search tab it should now install correct mod and pagecount should be correct too 19:58 VanessaE hope you are keeping your fork of shaders rebased each time you break it :) 19:59 VanessaE (if RBA ever shows back up to finish it up) 19:59 sapier lol I try to but fixing core crashes is priorized 20:11 PilzAdam hmmmm, https://gist.github.com/PilzAdam/7757735 after setting the mapgen from v6 to v7 via https://github.com/PilzAdam/nodetopia/blob/master/mods/base/mapgen.lua#L23 at first start after creating the map 20:25 proller bad noise params 21:21 sapier https://github.com/minetest/minetest/pull/1031 john_minetest this should fix crash in release mode 21:49 daswort is this default in 0.4.8? → https://forum.minetest.net/viewtopic.php?id=7838 21:56 iqualfragile awsome, my minetest server craches when i try to join 21:57 sapier current master? 21:57 iqualfragile yes 21:57 sapier release build? 21:57 iqualfragile yes 21:57 iqualfragile just segmentation fault, nothing else 21:57 sapier switch to 1031 21:57 iqualfragile kahrl: merge 1031 please 21:57 sapier try it first ;-) 22:07 sapier I could merge it myself but it's too much change to merge without proper review 22:08 iqualfragile works 22:08 iqualfragile sapier: fix mobfspam plz 22:09 sapier wonder why noone did report this prior release ;-) 22:12 sapier yes but doesn't contain the crash fixes 22:12 iqualfragile yes, that v6/v7 thing allready confused some users 22:16 iqualfragile v7 is not experimental afaik, but you would need to define some biomes 22:18 sapier I'll rebase the search branch once the fixes are added 22:39 daswort when i click on "online mod repo" i the error "luaerror: c++ exception" 23:03 ShadowNinja Yay, LuaJIT exception mangling. daswort: If you can compile with non-JIT Lua you might get a better error message. 23:40 ShadowNinja sapier: #1031 fixed the segfaults? 23:42 VanessaE what about the issue I reported last night? 23:42 VanessaE I'd say that was rather critical 23:43 VanessaE http://irc.minetest.ru/minetest-dev/2013-12-02#i_3467471 23:45 sapier fixed by 1031 as well as those nasty crashes on release build 23:45 VanessaE h good 23:45 VanessaE oh* 23:46 ShadowNinja kahrl and hmmmm might be abl to help with that. 23:46 ShadowNinja able* 23:47 sapier wait vanessae have you been talking about the delay or the emerge thread things? 23:47 VanessaE neither, see the log link. 23:47 VanessaE backing down to end-of-day Nov. 30 stabilized the servers. 23:48 sapier I strongly recommend merging 1031 soon those jthread changes for log threadname broke any mutex on release build 23:48 VanessaE as soon as someone signs on, ^^^^^ that crash happened - on all five servers. 23:48 VanessaE and yes I run release builds now 23:48 sapier don't do that with current master 23:48 VanessaE got tired of the slower speeds of debug builds :) 23:49 sapier as mutexes don't work at all in current state 23:49 VanessaE eek 23:49 VanessaE well right now I'm in a stable build so I'll leave it there until you guys give me the "okay" to push forward 23:50 VanessaE what's the status of RBA's shaders? did anyone figure out what broke? 23:50 sapier noone realized that assert(pthread_mutex_init() == 0); is replaced by nop on release build 23:51 VanessaE eh.. threads, semaphores, mutexes... back in MY day we had ONE thread and maybe two interrupt priorities, and WE WERE HAPPY! :) 23:52 VanessaE (seriously, that stuff goes a bit over my head) 23:52 sapier those times multicore was thought to be a fruit 23:52 VanessaE lol 23:53 ShadowNinja sapier: "assert(X);" --> "" via a macro? Shouldn't it do "assert(X);" -> "X;"? 23:53 sapier obviously it doesn't do this way 23:54 sapier as I said it's replaced by nop in release resulting in no locking at all 23:54 sapier it's not difficult to imagine what's happening 23:55 VanessaE ouch. 23:55 sapier someone should verify 1031 on windows so this can be merged soon 23:55 sapier for the moment hands of release build of master ... there's no safe way to run it 23:56 VanessaE ok. 23:56 sapier +f 23:56 * VanessaE resets her builds back to debug :-/ 23:56 ShadowNinja sapier: Well I'll agree to it, is there any reason that this would fail on Windows? 23:56 sapier unless you want to use your world backups you should do that immediatly 23:57 sapier none I know of but I didn't realize the assert() error too ;-) so I might not be best one to ask 23:57 sapier -of +about 23:59 sapier I'd like kahrl or hmmmm to check 1031 prior merge but I don't know if they have time to do 23:59 VanessaE ok, all servers and my client are back to debug mode.