Minetest logo

IRC log for #minetest-hub, 2021-04-23

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

All times shown according to UTC.

Time Nick Message
01:35 T4im joined #minetest-hub
01:38 T4im joined #minetest-hub
04:00 MTDiscord joined #minetest-hub
06:54 MTDiscord joined #minetest-hub
07:39 calcul0n_ joined #minetest-hub
08:00 ShadowNinja joined #minetest-hub
09:06 calcul0n__ joined #minetest-hub
09:21 Fixer joined #minetest-hub
09:47 hlqkj joined #minetest-hub
15:45 Fixer joined #minetest-hub
17:17 scr267 joined #minetest-hub
17:27 scr267 joined #minetest-hub
18:16 hlqkj_ joined #minetest-hub
19:10 hlqkj joined #minetest-hub
19:33 sofar joined #minetest-hub
20:42 MTDiscord <VanessaE> another person PM'd me on the forum with that same "not enough memory" when loading biome_lib, as discussed before.  He's using 5.4.1, win64 build.
20:42 MTDiscord <VanessaE> wtf?
20:44 MTDiscord <VanessaE> where are these errors even coming from?  the whole mod is only 152 kB or something, and uses less than 1 MB of memory for its tables et al., and it works perfectly for me in singleplayer and on like four of my servers
20:45 MTDiscord <VanessaE> is there some kind of LuaJIT- or 64-bit-related bug in the Windows 5.4.1 build?
20:48 sfan5 if that's the error where it barely says anything and just "out of memory" it's coming from C++ and not Lua(JIT)
20:50 MTDiscord <VanessaE> well
20:50 MTDiscord <VanessaE> it's "not enough memory"
20:50 MTDiscord <VanessaE> not "out of memory"
20:50 MTDiscord <VanessaE> that's a message thrown by MT (I checked the src), but the question is why?
20:51 MTDiscord <VanessaE> I'm wondering if there's something wrong with the package on contentdb maybe?
20:53 MTDiscord <VanessaE> hm, doesn't look like it.  only 144k unzipped.
20:54 MTDiscord <VanessaE> this is the error btw,   2021-04-22 18:36:36: ERROR[Main]: ModError: Failed to load and run script from C:\Users\Tommaso\minetest-5.4.1-win64\bin\..\mods\biome_lib\init.lua: 2021-04-22 18:36:36: ERROR[Main]: not enough memory
20:55 sfan5 thrown by MT where?
20:55 MTDiscord <VanessaE> lib/lua/src/lmem.h:#define MEMERRMSG    "not enough memory"
20:56 MTDiscord <VanessaE> lib/lua/src/ldo.c:      setsvalue2s(L, oldtop, luaS_newliteral(L, MEMERRMSG)); lib/lua/src/lstate.c:  luaS_fix(luaS_newliteral(L, MEMERRMSG)); lib/lua/src/lmem.h:#define MEMERRMSG    "not enough memory"
20:56 MTDiscord <VanessaE> those are where the value is actually used
21:00 MTDiscord <VanessaE> I have only one theory into this:  I've been transitioning the mod away from biome_lib:foo() method-like notation to the more usual biome_lib.bar()  (period) notation, and there may be a couple of functions that are still under the old name (but with . instead of :).  But those didn't give me any problems, apart from stack overflows in one case (which I already fixed, but which is NOT yet in public use)
21:01 MTDiscord <VanessaE> I guess I do need to push a new release to get that engine decoration code into peoples' hands, but I would not expect it to help the two users who reported this "not enough memory" thing.
21:02 MTDiscord <VanessaE> (fwiw, the other user is "Big3Force" on the forum)
21:07 sfan5 ok so it seems like it's from Lua
21:07 MTDiscord <VanessaE> (new release pushed)
21:07 sfan5 but it's internally out of memory so it can't report anything useful
21:07 sfan5 so mod code causing a stack overflow or similar doesn't sound out of question
21:08 MTDiscord <VanessaE> perhaps, but if that were the case, it would explicitly report a stack overflow
21:09 MTDiscord <VanessaE> (I mean, it does do that already)
21:22 MTDiscord <VanessaE> In any case I pushed the new release to cdb and in Dreambuilder.  If there was a stack overflow that isn't being reported properly, there won't be anymore.
21:30 Fuchs_ joined #minetest-hub
21:41 MTDiscord <VanessaE> sfan5:  could you give c4151a07 a try in 5.4.1-win64, and see how e8afe017 (HEAD) compares?
21:42 MTDiscord <VanessaE> the first is where those two users would have been provided they were up-to-date prior to half an hour ago
21:42 sfan5 !mod biome_lib
21:42 MinetestBot sfan5: Biome library [biome_lib] by VanessaE - https://forum.minetest.net/viewtopic.php?t=12999 - https://gitlab.com/VanessaE/biome_lib
21:43 sfan5 what would I install to compare? biome_lib itself doesn't do anything
21:43 MTDiscord <VanessaE> eh, moretrees or any one mod in plantlife would be enough
21:44 MTDiscord <VanessaE> though I'm not clear on just how much of the latter they were using
21:46 MTDiscord <VanessaE> (you can ignore all the warnings biome_lib throws with HEAD, they're harmless)
21:48 sfan5 !mod moretrees
21:48 MinetestBot sfan5: More Trees! [moretrees] by VanessaE - https://forum.minetest.net/viewtopic.php?t=4394 - https://gitlab.com/VanessaE/moretrees
21:51 sfan5 it instantly crashes
21:51 sfan5 2021-04-23 23:51:10: ERROR[Main]: ServerError: AsyncErr: environment_Step: OOM error from mod 'biome_liby
21:51 sfan5 2021-04-23 23:51:10: ERROR[Main]: Current Lua memory usage: 2052 MB
21:51 MTDiscord <VanessaE> wow
21:52 MTDiscord <VanessaE> and HEAD?
21:52 sfan5 that's HEAD
21:52 MTDiscord <VanessaE> wat
21:52 sfan5 the other commit crashes with "ERROR[Main]: not enough memory"
21:53 MTDiscord <VanessaE> where the heck is this all coming from?
21:59 sfan5 I've verified it three times now
21:59 sfan5 the faulty commit is https://gitlab.com/VanessaE/biome_lib/-/commit/c4151a0
22:00 MTDiscord <VanessaE> wat
22:15 Fuchs joined #minetest-hub
22:23 MTDiscord <VanessaE> what does os.date() even do on windows then?
22:23 MTDiscord <VanessaE> (PiL and lua-users say that it should just work as on POSIX)
22:27 sfan5 you're not gonna believe what it does
22:27 sfan5 https://i.imgur.com/G0SYGbQ.png
22:27 MTDiscord <VanessaE> wtf
22:29 sfan5 https://sourceforge.net/p/mingw-w64/bugs/793/
22:29 sfan5 though luajit shouldn't do this when strftime fails
22:32 MTDiscord <VanessaE> so I guess MT needs to provide its own equivalent API call then
22:32 MTDiscord <VanessaE> (or intercept os.date())
22:33 MTDiscord <VanessaE> or, is there a way for my mod to detect [windows + luajit] ?
22:36 MTDiscord <VanessaE> could I trap this error with assert()
22:38 Fuchs joined #minetest-hub
22:39 sfan5 just replace it with os.date("%Y-%m-%d %H:%M:%S")
22:40 MTDiscord <VanessaE> ah! so the %F is what breaks it
22:40 MTDiscord <VanessaE> I didn't quite understand that mingw link; that explains why the ETA timer at shutdown doesn't crash then
22:40 MTDiscord <VanessaE> thanks 🙂
22:41 sfan5 https://github.com/LuaJIT/LuaJIT/commit/fc63c938b522e147ea728b75f385728bf4a8fc35
22:41 sfan5 turns out luajit has already fixed this
22:41 sfan5 but there's no released version with the fix
22:46 MTDiscord <VanessaE> https://gitlab.com/VanessaE/biome_lib/-/commit/4e3493a9812c690052ee319bf94983e95f57f0b6
22:46 MTDiscord <VanessaE> pretty sad when the commit msg is longer than the actual code fix  🙂
23:25 Fuchs joined #minetest-hub

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