| Time |
Nick |
Message |
| 00:11 |
|
imi joined #minetest |
| 00:12 |
imi |
hi |
| 00:12 |
imi |
sometimes my minetest (client) hangs or crashes and I see this in dmesg: [24734.774011] i915 0000:00:02.0: [drm] GPU HANG: ecode 12:1:84dffffb, in minetest [8353] |
| 00:14 |
imi |
this is a hand compiled MT: 587f6656a - (4 months ago) Bump version to 5.6.1 - sfan5 (tag: 5.6.1, origin/stable-5 |
| 00:15 |
imi |
how and where should I report this |
| 00:23 |
FavoritoHJS |
i suspect gpu driver progrms |
| 00:23 |
FavoritoHJS |
problems* |
| 00:28 |
|
AliasAlreadyTake joined #minetest |
| 00:30 |
|
YuGiOhJCJ joined #minetest |
| 00:59 |
|
smk joined #minetest |
| 01:05 |
|
FavoritoHJS joined #minetest |
| 01:10 |
|
Lesha_Vel joined #minetest |
| 01:23 |
MTDiscord |
<Awkanimus> Was able to retrieve a stack trace from my periodic segfaults. Is it normal for there to be 739-ish seemingly recursive calls to PathFinder::updateAllCosts? Backtrace is here: https://gist.github.com/Awkanimus/e5110dc107041622ff0f9d688017a6d9 |
| 01:25 |
|
amfl2 joined #minetest |
| 01:32 |
MTDiscord |
<Awkanimus> I'm going to rebuild with CMAKE_BUILD_TYPE=Debug and see if I can get the line numbers out of it. |
| 01:34 |
|
lsabino joined #minetest |
| 01:35 |
lsabino |
is it possible to use mapgen v7 but with no flying islands? |
| 01:35 |
MTDiscord |
<Awkanimus> There's a checkbox when creating the map |
| 01:36 |
lsabino |
I unchecked that but it create a lot |
| 01:37 |
lsabino |
I'm using the mod Wilhelmines Natural Biomes, maybe it cause that? |
| 01:37 |
MTDiscord |
<Awkanimus> Odd. Sorry, I've no idea |
| 01:37 |
FavoritoHJS |
you mean floating bits of land, especially around tall, rough terrain? Those are probably a feature of the mapgen. |
| 01:39 |
lsabino |
FavoritoHJS yes, there's some conf to reduce the incidence? |
| 01:39 |
FavoritoHJS |
I don't think there's one |
| 01:40 |
FavoritoHJS |
but no-one said you can't use other mapgens |
| 01:40 |
lsabino |
I'll try with v6 |
| 01:40 |
FavoritoHJS |
well, except that one |
| 01:40 |
FavoritoHJS |
v6 is very janky |
| 01:43 |
lsabino |
oh I always to use v7, but now it is creating a lot of bits of land |
| 01:43 |
lsabino |
is v5 better? |
| 01:44 |
FavoritoHJS |
in terms of keeping floating islands in check, no |
| 01:44 |
FavoritoHJS |
from what i recall v7 is more of a tamed v5 |
| 01:44 |
FavoritoHJS |
i was thinking more of valleys or carpathian |
| 01:45 |
lsabino |
I didn't try valleys, I'll do a city, that's sound a good idea |
| 01:46 |
FavoritoHJS |
am i talking to chatgpt or what, as I don't think there's a city mapgen |
| 01:47 |
lsabino |
oh no, sorry. I just commented on my building plans |
| 01:57 |
imi |
FavoritoHJS: ok, how do I figure out? where do I report it? |
| 01:58 |
imi |
might be possible though, sometimes the whole machine freezes which suggests kernel level hiccup |
| 02:01 |
FavoritoHJS |
have you overclocked your hardware or something like that? |
| 02:02 |
FavoritoHJS |
if not, consider updating your intel graphics drivers |
| 02:14 |
imi |
FavoritoHJS: no |
| 02:15 |
imi |
I've just reported a bug to ubuntu maybe that'll help |
| 02:43 |
FavoritoHJS |
very stupid question: is there a way to ensure that a function passes some arguments in, for example `rax`, the 3'rd element of the stack, stuff like that? As I'm unsure how working around abi issues is even possible without that, at least not without writing large amounts of assembler |
| 03:38 |
muurkha |
use a compiler that implements the calling convention you want to use; GCC implements several which you can choose with __attribute__ |
| 03:55 |
|
Menchers_ joined #minetest |
| 03:57 |
|
sometalgoo joined #minetest |
| 04:09 |
|
izzyb joined #minetest |
| 04:18 |
|
gh00p joined #minetest |
| 04:46 |
|
roadbull joined #minetest |
| 04:52 |
lissobone |
Greetings. |
| 04:52 |
lissobone |
I might consider using arena_lib. |
| 04:52 |
lissobone |
The problem with hard dependencies is that they have to stay very stable, otherwise all dependent mods and games crash. |
| 04:53 |
lissobone |
Just some thoughts. |
| 04:53 |
lissobone |
Developing a library is a great responsibility. |
| 04:57 |
MTDiscord |
<MisterE> Well, arena_lib is actively maintained, and used and tested on our minigame server. Zughy takes code correctness very seriously |
| 04:59 |
MTDiscord |
<MisterE> There are occasional breaking changes, that require mods to update, though those are going to be documented. The actually breaking changes occur at milestones in the planning. So far, it seems Zughy has made sure to give advance warning and compatability code (which is removed in another update) for breaking changes |
| 05:00 |
|
MTDiscord joined #minetest |
| 05:01 |
MTDiscord |
<MisterE> Such as engine warnings that a particular feature is no longer supported, and the mod should update ... |
| 05:11 |
lissobone |
[Best prices in the Tri-State Area!] |
| 05:23 |
|
lemonzest joined #minetest |
| 05:32 |
|
est31 joined #minetest |
| 05:46 |
|
sys4 joined #minetest |
| 06:28 |
|
kamdard_ joined #minetest |
| 06:53 |
|
hmmmm joined #minetest |
| 06:53 |
|
hmmmm joined #minetest |
| 06:58 |
lissobone |
I used statistical mechanics to compute the best configuration of the player spreader. |
| 06:59 |
lissobone |
Maybe spawn players in special boxes at the start? |
| 07:02 |
lissobone |
[[Specil]]. |
| 07:33 |
|
definitelya joined #minetest |
| 07:38 |
lissobone |
I'll first make an API with numerous functions, then build my game on top of them. Sounds like a gamer move. |
| 07:41 |
|
mrkubax10 joined #minetest |
| 07:43 |
|
Verticen joined #minetest |
| 07:49 |
|
calcul0n_ joined #minetest |
| 08:10 |
|
proller joined #minetest |
| 08:21 |
|
hmmmm joined #minetest |
| 08:21 |
|
hmmmm joined #minetest |
| 08:24 |
luk3yx |
kilbith: Where did i3_cannot.ogg come from? |
| 08:27 |
|
hmmmm joined #minetest |
| 08:27 |
|
hmmmm joined #minetest |
| 08:28 |
|
Talkless joined #minetest |
| 09:04 |
|
Sven_vB joined #minetest |
| 09:09 |
|
sys4 joined #minetest |
| 09:12 |
|
___nick___ joined #minetest |
| 09:48 |
|
Sven_vB joined #minetest |
| 09:55 |
|
___nick___ joined #minetest |
| 09:57 |
|
___nick___ joined #minetest |
| 10:24 |
|
Fixer joined #minetest |
| 10:25 |
|
erstazi joined #minetest |
| 10:27 |
|
__nick__ joined #minetest |
| 10:32 |
|
Sven_vB joined #minetest |
| 10:34 |
|
proller joined #minetest |
| 10:43 |
|
appguru joined #minetest |
| 10:43 |
lissobone |
Greetings. |
| 11:05 |
|
mrkubax10 left #minetest |
| 11:31 |
|
mrkubax10 joined #minetest |
| 11:31 |
ROllerozxa |
mfw neither one of the minetest wikis have functioning syntax highlighting |
| 11:31 |
ROllerozxa |
I thought only the dev wiki was broken but the regular wiki can't do syntax highlighting either |
| 11:32 |
ROllerozxa |
could just throw prism.js in and be done with it |
| 11:59 |
MinetestBot |
[git] Panquesito7 -> minetest/minetest: Replace minetest_game with "Minetest Game" where appropriate d82d18b https://github.com/minetest/minetest/commit/d82d18bfb1a51e0f2f0a89999474e98890addf7e (2023-01-07T11:58:16Z) |
| 12:38 |
|
proller joined #minetest |
| 12:41 |
|
Leopold joined #minetest |
| 13:12 |
|
Talkless joined #minetest |
| 13:27 |
sfan5 |
ROllerozxa: I'm sure there are working plugins but someone needs to care for the wiki(s) |
| 13:38 |
|
appguru joined #minetest |
| 13:48 |
|
appguru joined #minetest |
| 13:49 |
|
rubenwardy joined #minetest |
| 14:11 |
|
mrkubax10 left #minetest |
| 14:27 |
kabou |
when I do a `minetest.read_schematic(filename,options)` minetest seems to halt completely |
| 14:27 |
kabou |
it crashes with "segmentation fault" and dumps core |
| 14:28 |
kabou |
even `minetest --trace` shows no hints as to what went wrong |
| 14:29 |
sfan5 |
attach a debugger to get more details |
| 14:33 |
|
jwmhjwmh joined #minetest |
| 14:35 |
jwmhjwmh |
kabou: What's your Minetest version? A bug was fixed with that function recently. Are you calling it at load time, or after? |
| 14:36 |
kabou |
after |
| 14:36 |
Krock |
gdb minetest |
| 14:36 |
Krock |
r |
| 14:36 |
Krock |
bt |
| 14:37 |
kabou |
i'm writing a mod that checks schematics from a chat callback |
| 14:37 |
Krock |
^ that's the thing you can do to find out what's going wrong |
| 14:37 |
kabou |
do I need a minetest binary compiled with debug support |
| 14:37 |
Krock |
would be nice, but more important is to know which MT verison you're using |
| 14:37 |
lissobone |
Just compile one. |
| 14:38 |
kabou |
Minetest 5.6.1 (Linux) |
| 14:38 |
kabou |
Using Irrlicht 1.9.0mt8 |
| 14:38 |
kabou |
Using LuaJIT 2.1.0-beta3 |
| 14:38 |
kabou |
from debian testing |
| 14:38 |
Krock |
alright... recent enough. release build should give you some symbols |
| 14:39 |
kabou |
Thread 16 "Server" received signal SIGSEGV, Segmentation fault. |
| 14:39 |
kabou |
[Switching to Thread 0x7fffc8e226c0 (LWP 183480)] |
| 14:39 |
kabou |
0x00007ffff7174dd9 in ?? () from /lib/x86_64-linux-gnu/libc.so.6 |
| 14:39 |
Krock |
pastebin.com is your friend |
| 14:39 |
Krock |
or other paste services |
| 14:40 |
kabou |
those three lines is about the info it gives |
| 14:40 |
Krock |
there must be more, e.g. when you hit Enter to continue the backtrace output |
| 14:41 |
kabou |
i'm sorry, not very familiar with the debugging process with gdb |
| 14:41 |
kabou |
I run `gdb minetest` |
| 14:41 |
kabou |
then `run` |
| 14:41 |
Krock |
then you have this crash and type "bt" |
| 14:41 |
sfan5 |
related: I had the idea a while ago that we should add a --debugger option to Minetest that automatically runs it under gdb with all the right options to get useful output |
| 14:41 |
kabou |
it starts minetest, I start a world that has the mod enabled |
| 14:41 |
kabou |
then I run the offending chat command |
| 14:42 |
kabou |
and apart from the last `minetest.log` before it calls `minetest.read_schematic`, those three lines is all I get |
| 14:42 |
jwmhjwmh |
See if the persists with the latest Minetest version from git. |
| 14:42 |
kabou |
ah okay, just did bt |
| 14:44 |
kabou |
https://pastebin.com/ur1zFQJJ |
| 14:44 |
Krock |
sfan5: yes. Wine also has an automatic debugger attach on crash to give a backtrace. it would be quite helpful to have that for minetest. perhaps we can read out the core dump file? |
| 14:45 |
Krock |
kabou: okay that backtrace is pretty useless. I can only tell that it went through the Lua API (obviously) and a few Minetest functions... |
| 14:45 |
Krock |
although .. why lua_pcall? |
| 14:45 |
kabou |
gonna check if there is a dbg build in debian |
| 14:45 |
lissobone |
chesecak:-[ |
| 14:46 |
Krock |
kabou: mind sharing a minimal version of your mod so that I could try to reproduce it? |
| 14:46 |
sfan5 |
kabou: you're better off just building from source |
| 14:47 |
jwmhjwmh |
I expect an invalid pointer is being passed to lua_pushstring. |
| 14:48 |
kabou |
https://pastebin.com/WxBgZSvq |
| 14:48 |
kabou |
that's te script |
| 14:50 |
Krock |
on it.... |
| 14:54 |
kabou |
could it be related to the fact that I got the filename from an insecure environment? (the script is a trusted mod) |
| 14:55 |
Krock |
cannot reproduce. works for me on this week's Minetest 5.7.0-dev |
| 14:56 |
Krock |
loaded 152 mts schematics from mineclone |
| 14:56 |
jwmhjwmh |
https://github.com/minetest/minetest/commit/386bfcda2b3565b38c79af76743ad3d76c452fd5 |
| 14:56 |
Krock |
so the fix here does indeed work ^ |
| 14:57 |
Krock |
I suppose it's time for a 5.7.0 release |
| 14:58 |
kabou |
well I guess my script is a bit of an uncommon case anyway |
| 15:00 |
Krock |
I remember fixing a similar issue in place_schematic (or so) a while ago - seems like that function suffered from a similar issue. well then. if you're already compiling Minetest, you could get the newest one with the fix included |
| 15:01 |
kabou |
gonna get the deb with dbgsyms first |
| 15:04 |
kabou |
paste with symbols: |
| 15:04 |
kabou |
https://pastebin.com/ivWVRySj |
| 15:04 |
Krock |
https://github.com/minetest/minetest/commit/386bfcda2#diff-be4624c7b000e30c1dfd7240605f7b0f6414b8da1b9cdaebef53922d765d8c42L1748 |
| 15:05 |
Krock |
it's precisely this line |
| 15:06 |
kabou |
allright then ,that's settled. thanks for helping me find what's going on. Time to build current git minetest then |
| 15:10 |
Krock |
alternatively you could try to get a Frankenstein Debian by using the nightly Minetest build from the Ubuntu PPA and hope that it runs |
| 15:13 |
MTDiscord |
<Benrob> building from git is probably easier (after you do it once or twice) and less likely to break your system |
| 15:14 |
kabou |
in fact I try to add the debian build stuff to minetest git and have it build a .deb |
| 15:14 |
kabou |
worked before when debian was severely lagging |
| 15:15 |
kabou |
they're reasonably up to date now |
| 15:15 |
kabou |
btw the debian packages are of a higher quality than the ubuntu ppa ones I think |
| 15:15 |
kabou |
especially if you want to run a server only |
| 15:38 |
lissobone |
Ok, so. |
| 15:38 |
lissobone |
I am back. |
| 15:38 |
lissobone |
I've got good news. |
| 15:39 |
lissobone |
I have implemented the smooth radius and coordinates transition. |
| 15:39 |
lissobone |
While figuring out the algorithm I found out something interesting. |
| 15:40 |
lissobone |
What I wanted to do is to make the numbers transition at a constant rate. But at first they went really fast at the beginning of the transition and really slow at the end. |
| 15:40 |
lissobone |
I might use this somewhere else. |
| 15:44 |
|
mrkubax10 joined #minetest |
| 16:09 |
|
appguru joined #minetest |
| 17:38 |
|
olliy joined #minetest |
| 19:02 |
|
Noisytoot joined #minetest |
| 19:18 |
|
appguru joined #minetest |
| 19:31 |
|
hmmmm joined #minetest |
| 19:31 |
|
hmmmm joined #minetest |
| 19:40 |
|
sparky4 joined #minetest |
| 19:40 |
kabou |
after building minetest in place, `make clean` does not remove the minetestserver binary |
| 19:41 |
kabou |
it does remove the minetest binary from `bin` |
| 19:51 |
muurkha |
doh |
| 19:55 |
|
___nick___ joined #minetest |
| 19:57 |
|
___nick___ joined #minetest |
| 20:05 |
|
mrkubax10 left #minetest |
| 20:08 |
|
FavoritoHJS joined #minetest |
| 20:12 |
|
MTDiscord joined #minetest |
| 21:00 |
|
jwmhjwmh joined #minetest |
| 21:06 |
|
Guest27 joined #minetest |
| 21:26 |
|
appguru joined #minetest |
| 21:37 |
|
___nick___ joined #minetest |
| 21:58 |
|
gabin_ joined #minetest |
| 22:03 |
|
gabin__ joined #minetest |
| 22:34 |
|
Verticen joined #minetest |
| 23:13 |
MTDiscord |
<Awkanimus> My minetest server segfault saga continues. Rebuild with RelWithDebugInfo and eventually it droped a core. Using GDB I now have a stacktrace with line numbers. Stack trace and version info are here: https://gist.github.com/Awkanimus/d9037f646d0e36b8c377773b4840842f |
| 23:27 |
|
Trifton joined #minetest |
| 23:27 |
sfan5 |
that's a textbook stack overflow |
| 23:27 |
sfan5 |
you hit a bug in the pathfinder |
| 23:28 |
sfan5 |
if you figure out how to reproduce it it'd be nice to report it on the issue tracker |
| 23:30 |
FavoritoHJS |
seems like it has to do with `PA_DIJKSTRA`, it also appears like the self pointer isn't updating so if it's using a linked list of some sort make sure it doesn't have objects linking to themselves |
| 23:33 |
sfan5 |
the `this` pointer refers to the class the method is in, it doesn't change |
| 23:33 |
MTDiscord |
<luatic> This could be a case where a recursive implementation needs to be converted into an iterative one to spare the stack |
| 23:34 |
FavoritoHJS |
hmm, that algoritm seems horrendously slow in the "no path" case. |
| 23:34 |
MTDiscord |
<luatic> e.g. it could be the case that this isn't a bug but rather a stack storage limitation, AFAIK the C stack is pretty small by default |
| 23:34 |
FavoritoHJS |
I wonder if that plus a small stack is causing the problem... |
| 23:34 |
MTDiscord |
<luatic> dammit now I'm curious how it's implemented |
| 23:34 |
FavoritoHJS |
that's... bad |
| 23:35 |
FavoritoHJS |
if you have no path and attempt to find a long path it can stall for 5 seconds |
| 23:37 |
MTDiscord |
<luatic> oof, that code is a mess |
| 23:44 |
MTDiscord |
<Awkanimus> A bug in pathfinder you say. I'll try to keep figuring out how to reproduce it and will report if I figure it out. It tends to occured randomly, but approximate twice a day. |
| 23:46 |
FavoritoHJS |
hmm... Devtest has a pathfinding tool, that could help replication |
| 23:48 |
|
Desour joined #minetest |
| 23:58 |
MTDiscord |
<Awkanimus> Thanks, that's good to know |
| 23:59 |
MTDiscord |
<Awkanimus> Thanks for the insights all |
