Time  Nick        Message
00:00 paramat     the niggest issue i see with servers is the automatic use of hundreds of mods, many badly coded and over intensive, then complaints about performance and MT devs getting the blame. but, it's expected from kids who don't know better yet
00:02 paramat     tumeninodes just add new registration functions, copypaste them from the others, and change the nodebox def as i showed you on github
00:03 paramat     oops, *biggest heh
00:03 sfan5       <VanessaE> (to me, the solution is simple: perform the compositing one time only, and "re-render" only the completed, static image)
00:03 sfan5       fyi that's what MT actually does
00:04 sfan5       i'm not sure if it does a lookup of textures every frame or how that works, but the tile.cpp code will cache any applied modifiers
00:05 paramat     we are always working on improving performance, just a case of dev time, not a case of neglect as some believe
00:07 paramat     server owners just have to work within current limits instead of overloading MT
00:07 Shara       paramat: we know, but there definitely are issues which it is not easy for server owners to overcome
00:08 Shara       It's about placing blame (at least for me) anyway
00:08 Shara       It's not*
00:08 Shara       Heh, bad typo
00:09 Shara       The issue for me is how difficult it becomes to maintain a server over time when the auth file gets bigger and bigger (that practically broke my server at one point due to the insane amount of checks being done on such a large file)... then the map becoming increasingly larger and more and more difficult to work with
00:10 Shara       The solutions are not always easy though.
00:10 Shara       And sometimes it's just about adjusting expectations instead
00:12 paramat     yeah, there are issues
00:13 paramat     stuff that simply isn't acceptable
00:13 paramat     ugh this 4w / linuxdirk guy ^_^
00:14 Shara       I suspect server owners start "blaming" devs so often for things because features get introduced that we don't want or actively dislike (sometimes very strongly), and from the outside it can seem that developer time is much better spent on fixing the issues and not adding new headaches.
00:14 paramat     i agree with you there :]
00:14 Shara       REality doesn't go that way of course, because volunteers.
00:14 paramat     (CSM)
00:14 Shara       (Exactly)
00:16 Shara       CSM is all about servers... but server owners do not want it... yet we're stuck with it.
00:16 Shara       But let's not start that pointless conversation again :)
00:21 Fixer       Shara: bump nerzhul to introduce auth in sqlite, if map is too big to your, you can limit and trim it to smth like +/- 5000, it is still huge and will take decades to be full
00:21 Shara       Fixer: no good. Need the PRs paramat has been working on for a real solution.
00:22 Fixer       Shara: also, bump some guys to introduce smart mapblock saving
00:22 Shara       Also I already moved my auth to a db on the server that had that problem. Not waiting who knows how long for devs to do it
00:22 Shara       Some things modders can handle, even if there shouldn't be a need :)
00:23 Shara       Yes well, I'd love very much to see better handling of how maps are saved... but you can't force a dev to work on something that they don't want to
00:23 Shara       And I doubt that would be a small thing... and would it even help on existing maps?
00:24 paramat     rubenwardy 4w seems to be trying to stir up paramat hate, at some point i'll consider this personal attacks which if continued means a possible ban from github, like mr. k
00:24 Fixer       Shara: nerzhul has the code, he needs to backport it
00:24 Shara       Fixer: code for which thing?
00:24 Fixer       nerzhul: ^^^vvv
00:24 Shara       We've covered a few different points there
00:24 red-001     paramat, colour me surprised
00:24 Fixer       Shara: db auth
00:25 paramat     k is halfway to a github ban currently
00:25 Shara       paramat: link? I do love to thumb down stupid pointless insulters
00:25 red-001     it's 4w he is basically a well known troll
00:25 red-001     paramat, who is k?
00:25 Shara       Fixer: I don't need db auth. I have db auth
00:25 paramat     obvious
00:25 red-001     no its not
00:25 Shara       Is to me
00:25 Fixer       Shara: and it helped you, right?
00:25 red-001     oh french graphics coder?
00:26 Shara       :)
00:26 paramat     yes
00:26 paramat     https://github.com/minetest/minetest/pull/7030#issuecomment-364566051
00:26 Shara       Fixer: Yup. Not saying it wouldn't be good to add, just too late for me really, and could even give me trouble on one server
00:26 Shara       And well, there's a mod that does it quite nicely now
00:27 red-001     unlike k graphics 4w also doesn't contribute in any way
00:28 paramat     i can cope with 4w as long as it doesn't get personal
00:28 red-001     paramat, is that qoute basically the gist of it?
00:37 Shara       red-001: follow the link paramat gave there and there is more
00:41 red-001     Shara, got a script that hides 4w comments and it's kinda a pain to enable and disable
00:44 Shara       heh
00:46 red-001     also it's a massive hack since I didn't figure out how to get the code to run after that whole update only loading thing the js or something does, so I kinda just detect the url change a set a bunch of timers to have my code run hoping at least one of them will run after the new content loads
00:47 red-001     s/a set/ and set
00:52 paramat     yes the quote is essentially it, except he wrote the same 4 days ago in the issue
01:02 Fixer       on of the minecraft oldest servers (8 years old) has map limit just smaller than mt one (radius of about 30km)
01:02 Fixer       one*
01:30 VanessaE    sfan5, RBA may have been referring to texture lookups.  idk for sure.
01:31 VanessaE    it's been way too long and we can't ask him now :-/
01:32 VanessaE    paramat: why can't *static, immovable* entities be lightweight?  can they not be turned into ordinary scenery?
01:32 VanessaE    s/scenery/geometry/
02:00 paramat     they can probably be more lightweight, just not as lighweight as a node
02:00 paramat     the experts think entities are too intensive currently
02:49 tumeninodes paramat: thank you
03:20 Megaf       it possible to remove world size limit in Minetest?
03:20 Megaf       is it*
03:25 VanessaE    no.
03:25 VanessaE    not without compatibility problems
03:26 Megaf       well, what if I dont mind about compatibility?
03:27 VanessaE    *shrug*
03:28 * Megaf     browsers the code
03:28 VanessaE    it's been requested before but the devs don't think it's needed.
03:28 tumeninodes give it a whirl and see.. :D
03:51 Megaf       In accordance with your request, the nsxxx210.ip-5-196-88.eu server has now been deleted.
03:51 Megaf       This operation is irreversible.
03:51 Megaf       I think my server is gone forever ^
03:51 Megaf       rubenwardy: I just canceled my Kimsufi
03:51 Megaf       tomorrow I will be removing my Linode as well
04:22 paramat     "it possible to remove world size limit in Minetest?" no it's hardcoded by the co-ordinate integer type s16
04:23 Megaf       I shall make that a gloat
04:23 Megaf       float*
04:23 paramat     heh
04:24 Megaf       paramat: so basically it's possible to increase the size without much code change tho it would be harder to remove the limit
04:24 benrob0329  If you go much further out, floating point becomes too imprecise as well
04:24 Megaf       I like the idea of trully ulimited map size
04:24 paramat     it's a big job to alter every s16 to a s32, but possible
04:25 benrob0329  You'd have to move the world around the player, rather than the player around the world for truly unlimited iirc
04:25 Megaf       benrob0329: that is VERY pylosophical
04:25 Megaf       whatever you spell that
04:25 paramat     it wouldn't be unlimited, just very very big
04:26 benrob0329  True
04:26 benrob0329  We'd need to use 64bit coords as well
04:26 Megaf       not a problem
04:26 Megaf       why not 128 bit?
04:27 Megaf       soon we gonna see 128 bit CPUs
04:27 benrob0329  Because 64 is the largest currently usable by most computers
04:27 Megaf       They are out there already, for very specific stuff
04:27 benrob0329  Without major performance problems
04:28 VanessaE    because s64 would be such a huge map size you couldn't store it all on a consumer hard drive anyway.
04:28 benrob0329  Also that
04:28 benrob0329  But that would allow for space to be truly, well spacious
04:29 benrob0329  Also the main world could be a single planet
04:29 paramat     all that world would never become generated anyway
04:29 VanessaE    Megaf: at s64, the map could go to +/- 4'294'967'296m
04:30 benrob0329  That would be awesome, but also horridly difficult to implement afaik
04:30 paramat     s32: 2 billion nodes in every direction
04:31 Megaf       hold your horses
04:31 VanessaE    or that, too.
04:31 Megaf       so
04:31 Megaf       can we make theworld larger using s32 then?
04:31 Megaf       I mean
04:31 Megaf       wait
04:32 Megaf       we use s16 currently
04:32 Megaf       so, drawbacks in going s32?
04:32 VanessaE    if you go beyond s16, it could be whatever, the lack of compatibility would be the same
04:32 paramat     so the vector co-ord would need to be <= 64 bits? so a limit of 21 bits per axis?
04:32 benrob0329  And potentially end up with the mapgen algorithms spiraling out of control
04:32 VanessaE    mainly functions that hash node or block positions
04:33 Megaf       it's 4:32 AM and I'm here talking about 32 bit integers and mapgen algorithms
04:33 benrob0329  Sounds like a normal day
04:33 Megaf       That's the scariest part
04:34 benrob0329  paramat: mapblock relative coords might need to be a thing
04:34 VanessaE    paramat: well, that would work too.  2M nodes per axis then
04:34 Megaf       I like the sound of that
04:34 benrob0329  Or even better, octree coords :^)
04:34 VanessaE    (that'd be about half the size of the US)
04:34 VanessaE    (I think)
04:34 Megaf       Two Million Nodes https://i.ytimg.com/vi/cmQLnBtDjbw/maxresdefault.jpg
04:35 Megaf       http://i0.kym-cdn.com/entries/icons/original/000/002/252/NoMeGusta.jpg
04:36 benrob0329  I mean, I'd probably donate towards such a change
04:36 Megaf       any change we can have the same area as Earth?
04:36 benrob0329  0.6 anyone?
04:36 Megaf       benrob0329: oh boy
04:36 Megaf       benrob0329: if you open a github issue for that I will support it
04:37 benrob0329  We need bounties
04:37 benrob0329  I'd donate to a few issues if I could
04:38 VanessaE    lesse.  earth is 40074 km in circumference...  so you'd need 22 bits per axis
04:38 VanessaE    so minimum s128.
04:39 VanessaE    well, 22 bits on X/Y
04:40 Megaf       Earth is big
04:40 VanessaE    so s64 would do, if you don't mind limiting Z to 20 bits (1048 km)
04:41 Shara       Random test of sea corals using rooted drawtype: http://gingercat.tk/dg/hw_coral_test.png
04:41 VanessaE    so 64 bits would be enough to model the earth down to a depth of ~948 km
04:41 Shara       No more horrible flickering/air mess!
04:41 VanessaE    nice, Shara
04:42 paramat     good stuff
04:43 Megaf       Shara: Now I know how you wanted to draw the corals...
04:43 Megaf       :P
04:43 Shara       :P
04:43 benrob0329  Shara: noice
04:43 Megaf       the same colours I was thinking about
04:43 Shara       Needs a lot of work still, but first step toward coral ocean biome anyway
04:44 VanessaE    Megaf: so.  gonna convert us to 22/20/22 bit coords? :)
04:45 Megaf       eh
04:45 Megaf       we could use something GPS like
04:45 Megaf       based on degrees
04:45 VanessaE    no.
04:46 Megaf       Shara: some of the corals could hurt if touched
04:46 VanessaE    the result would be even less compatible than it is now
04:46 Megaf       precisely
04:46 paramat     degrees is for spheres
04:47 Megaf       how do you navigate in a cube?
04:47 Megaf       cuboid*
04:48 Megaf       maybe current coord system is optimal
04:53 Megaf       [04:44:36] <VanessaE> Megaf: so.  gonna convert us to 22/20/22 bit coords? :)
04:53 Megaf       how hard can that be?
04:53 VanessaE    famous last words. ;)
04:55 Megaf       eh
04:58 Megaf       meh, Im having a look at the code
04:58 Megaf       every single mapgen and 100s of other files redefine s16 all the time
04:58 Megaf       cant we just define that once in a .h file?
04:58 Megaf       https://github.com/minetest/minetest/search?p=2&q=s16&type=&utf8=%E2%9C%93
05:00 Megaf       (I dont know what Im talking about)
05:06 Sokomine    rather than enlarging the world an option to remove map blocks that where generated once and never changed by players might be helpful. disk space is cheap but not unlimited
05:08 Sokomine    Shara: nice corals. are they flowerlike?
05:09 Shara       Sokomine: plantlike_rooted
05:09 Megaf       hum, mapblocks after certain range from buildings could timeout and be degenerated
05:13 Sokomine    Megaf: currently can't be done. we have no idea if players did anything in the mapblocks. or when they where last loaded or created
11:35 CWz         happy 200+ server, minetest!
11:48 Krock       CWz, that's nothing new actually
11:48 Krock       > peak 504
11:49 sfan5       the peak value actually isn't accurate
11:49 sfan5       but yeah we've had >200 servers before
13:49 Jordach     992 tris for an automatically extruded entity
13:55 Jordach     !tell benrob0329 https://i.imgur.com/3jQ58MG.png
13:55 MinetestBot Jordach: yeah, yeah
14:01 Jordach     https://i.imgur.com/mpCQufk.png it mostly likes complex shapes
14:13 Jordach     https://i.imgur.com/IMnFZnz.png
14:46 Megaf       Hello everyone
14:47 Krock       hi Megaf
14:47 Krock       Jordach, I see potential $.$
14:48 Megaf       Blender is so underestimated by many people
15:41 Calinou     death to Source FilmMa--
15:41 Calinou     oh wait…
16:12 Jordach     SFM is better than blender for animating
16:20 red-001     mods can put thier own settings in the secure namespace right?
16:21 IhrFussel   I found the source of my protection problem
16:22 IhrFussel   2018-02-10 17:20:21: ERROR[Main]: AREAS FROM FILE ARE NOT A TABLE
16:22 red-001     that just means the data is corrupted
16:23 IhrFussel   It's weird how it always works on 2nd try without changing anything
16:23 red-001     IhrFussel, did the file exist when you got that error, if yes did it look like some data got cut off while writing it?
16:23 red-001     yes the areas mod overwrites the file if it fails to load it
16:23 IhrFussel   Now the question is why doesn it fail to load sometimes
16:23 IhrFussel   does*
16:24 IhrFussel   io.open() gives no error
16:24 red-001     IhrFussel, does the data get wiped on second try?
16:24 IhrFussel   No
16:24 red-001     huh
16:24 IhrFussel   It's the exact same file
16:24 red-001     oh it must have been destroyed when someone modified something relating to protection after it failed to load
16:25 red-001     IhrFussel, copy the deserilize function to a lua sandbox and run it a few hundard times on that data and see if it works?
16:26 shivajiva   is it like deserialize fails because the data appears not to be a table even though it is?
16:26 shivajiva   because the error points to that
16:27 red-001     shivajiva, could be data not being fully loaded
16:27 IhrFussel   if type(self.areas) ~= "table" then minetest.log("error","AREAS FROM FILE ARE NOT A TABLE") self.areas = {}
16:27 red-001     nil after all is not a table
16:27 IhrFussel   The weird part is that it only started happening recently
16:27 shivajiva   and a missing final brace isn't either
16:27 red-001     IhrFussel, could you also do a  minetest.log("error", dump(self.areas)) before discarding the data?
16:28 shivajiva   yea ^
16:28 red-001     and really if the files exists and can't be loaded the mod should error out
16:28 red-001     file*
16:28 shivajiva   lets see what it's got
16:28 red-001     and maybe dump the length of the raw data read from file
16:29 red-001     that should show if it's getting cut off
16:31 shivajiva   your check is stopping the file being overwritten thankfully
16:31 red-001     shivajiva, no?
16:31 red-001     logging an error doesn't crash the state
16:32 red-001     you're thinking of error("")
16:32 red-001     (unless there is some setting no-one told me about)
16:33 shivajiva   I thought type checking forced it to error?
16:33 red-001     no
16:34 red-001     assert(type(self.areas) ~= "table") would error out
16:34 red-001     on it's own type(self.areas) ~= "table" just returns false
16:34 IhrFussel   If I'm not quick and someone protects something the file will be a new one
16:35 IhrFussel   It doesn't happen right now...I'm restarting 10 times now
16:36 shivajiva   yea it's asserting it that would :)
16:36 shivajiva   had to check my sauth code :P
16:38 IhrFussel   But if it loads mostly fine and sometimes fails it's 99% not a write error correct?
16:39 shivajiva   yes that would be logical
16:39 red-001     I would assume so
16:40 shivajiva   it points to deserialize failing the other 1%, time for an issue?
16:40 IhrFussel   areas doesn't save the data on shutdown so if it fails to load once and I restart without anyone saving something and it *suddenly* works fine then the problem must be the loading of the file
16:41 shivajiva   but not io
16:42 shivajiva   diff error for io
16:43 IhrFussel   No io.open() error in both cases...read and write
16:43 shivajiva   it has a handle on the file or it doesn't at that stage
16:44 shivajiva   if lua had an issue reading files it would have been picked up long ago
16:45 shivajiva   so the next process is deserialize
16:46 IhrFussel   self.areas = minetest.deserialize(file:read("*a"))
16:46 shivajiva   lets sit on it till you can dump the table and it's length
16:47 shivajiva   remember xban suffered similar issues and was happy to overwrite it's file
16:53 sfan5       are you using safe_file_write yet?
16:54 IhrFussel   It's not about writing, loading fails without writing
16:55 IhrFussel   The protection file itself is fine, it contains return { DATA }
16:56 IhrFussel   I also don't believe that I shutdown the server now the 4th or 5th time in 1 week at the exact time when a player tried to protect something...very very very unlikely
16:57 IhrFussel   And debug.txt would contain "minetest.log("error", "[areas] Failed to serialize area data!")" if saving failed
16:58 IhrFussel   io.open() after that also has no error else Id see "minetest.log("error","FATAL AREAS ERROR ... COULD NOT WRITE TO FILE, ERROR CODE: "..err)"
17:00 IhrFussel   Really I added several minetest.log() lines to loading/saving and only the result of minetest.deserialize() throws an error
17:02 sfan5       is your file too large perhaps?
17:02 IhrFussel   And it appears to be random...I cannot find a pattern (yet)
17:03 IhrFussel   932 KB
17:06 IhrFussel   I added now a dump(self.areas) in case it fails again and minetest.deserialize() says it's not a table ... just a matter of time until I can tell what the log reports as output
17:07 shivajiva   excellent
17:07 IhrFussel   Can dump() handle any data type?
17:08 shivajiva   anything but user data, that's the only time it fails
17:08 IhrFussel   I can only imagine if it's not a valid tab le sometimes, it's likely a bren string
17:08 IhrFussel   broken*
17:10 IhrFussel   could a ~ 1 MB file cause problems in lua?
17:11 Shara       I don't believe so, given xban files used to get much larger before troubles started
17:12 shivajiva   it's more like >5 MB where it gets funky with xban
17:12 Shara       My areas.dat is 875KB and no issues... yet
17:12 Shara       But not that far behind yours
17:13 IhrFussel   I also don't think it can be the SSD...if it was faulty I would probably have way worse problem s than just a large file not loading sometimes
17:14 shivajiva   I've looked at areas and it has to be rewritten if it's going to utilise a db correctly
17:20 Shara       Well, away for a while now, but I am concerned areas is a potential problem my server will also hit at some point...
17:22 Megaf       Shara: why you say that?
17:22 Megaf       I've been using areas since my first server, only had problems twice
17:22 Megaf       :)
17:22 Shara       Because everything else that was not in a db broke so far
17:22 Shara       But... away now :P
17:22 Megaf       Okidoki
17:23 IhrFussel   I experience those corruption problems since 2 weeks now ... didn't update MT recently...I did sometime in January the upgrade to 0.4.17
17:24 IhrFussel   Not sure if 0.4.17 changed minetest.deserialize() or any corresponding c++ code
17:25 IhrFussel   related c++ code*
17:28 Megaf       I wonder if it could be related to UTF-8 or non UTF-8 stuff in areas names
17:29 Megaf       Minetest has serious problems with accents and non English stuff, lile á, ç and ê
17:29 Megaf       Often it inteprets those as strings of stuff
17:33 IhrFussel   That doesn't really explain why it fails to load after 1 restart but succeeds after another
17:39 Megaf       !title https://en.wikipedia.org/wiki/CopperheadOS
17:39 MinetestBot Megaf: CopperheadOS - Wikipedia
17:39 Megaf       Android distro focused on security
18:17 Fixer       DATA PROSPECTATOR
18:18 Megaf       !server Megaf
18:18 MinetestBot Megaf: Megaf Server v4 | 163.172.154.119:30003 | Clients: 0/20, 0/2 | Version: 0.4.17-Megaf / MegafXploreNext | Ping: 12ms
18:19 Megaf       Will eventually move it to Amterdam
18:20 red-001     pretty sure minetest.(de)serialize is all Lua
18:20 rubenwardy  yes
20:05 sofar       goddammit nextcloud you piece of shit
20:07 Krock       NEEDS CAPS LOCK
20:08 Fixer       HE LOVES YOU
20:09 sofar       the upgrade from 11 to 12 to 13 leaves you stuck at 12 because 12 doesn't support php-7.2
20:09 Jordach     nice
20:10 sofar       and a completely broken install, too
20:17 Fixer       is this shaders or smth http://minecraftonline.com/w/images/c/c4/3WYfn91.jpg
20:20 Jordach     render
20:20 Jordach     GLSL shaders don't make fireflies everywhere
20:53 Calinou     sofar: don't you mean that 13 doesn't support PHP 7.2?
20:53 sofar       12
20:54 sofar       but you can't skip it when upgrading
21:13 Jordach     wonders of open source
21:18 Jordach     https://www.openhub.net/p/minetest-c55
21:21 Jordach     !tell paramat you've got 603 commits to Minetest Engine :D
21:21 MinetestBot Jordach: I'll pass that on when paramat is around
21:22 red-001     that data is really outdated last I checked
21:22 red-001     couldn't figure out how to fix it
21:27 rubenwardy  it over estimates
21:27 rubenwardy  maybe because it counts the same commit multiple times
21:27 rubenwardy  due to be on different branches
21:28 rubenwardy  paramat has 342 actual commits
21:30 red-001     also it lists paramat's first commit as being a year ago
21:31 red-001     looks like it lists every email as a seperate user
21:56 Calinou     red-001: .mailmap can be used to fix that
21:56 Calinou     but not all services read it
21:56 Calinou     ("git shortlog" will take .mailmap into account, for example)
21:56 Calinou     a .mailmap was added recently to godotengine/godot :)
22:16 Fixer       inactive: 77%, heh
22:20 Fixer       omg, mc2 has 185 issues open and 280 closed
22:33 Fixer       sofar: prune throws me error when trying to prune: Binary was compiled with 'CGO_ENABLED=0', go-sqlite3 requires cgo to work. This is a stub
22:33 Fixer       sofar: win7sp1
22:51 Krock       Does anyone have an idea why LuaJIT could panic when raising an luaL_error or using luaL_checklstring on a table?
22:53 Krock       probably not the right place to ask