Minetest logo

IRC log for #minetest-dev, 2014-04-19

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

All times shown according to UTC.

Time Nick Message
00:04 Zeno` joined #minetest-dev
00:50 hmmmm [10:06 AM] <proller> it not use occlusion and not skip air blocks
00:50 hmmmm have you not seen how small an air block is represented as?
00:51 hmmmm iirc it's like 17 bytes.
00:51 hmmmm come on.
00:51 hmmmm adding in all this complex logic to optimize something as specific as an "air block" is a bad, bad, bad idea
00:52 hmmmm spillz, it's a bitmap because you only need a *single* yes or no per block location (existence)
00:54 hmmmm the reason why i'd like to go with an octree structure is because... damn, now that I just woke up, I think my intent was to prevent blocks being created in a certain pattern using up way too much cpu
00:54 hmmmm err, memory
00:54 hmmmm yeah, it can be a hashtable as well.  the shame is that we'd use std::map which is a tree structure, though :/
00:55 Zeno` std::map is a tree?
00:55 hmmmm Zeno`, yes
00:56 Zeno` hmm ok
00:56 hmmmm it needs to be since the data is ordered by key
00:56 Zeno` ah, of course
00:56 hmmmm as of C++11 however there are "true" hashtables with std::unordered_map, and O(1) access time as per the spec
00:57 hmmmm so I was wondering if maybe, for optimization purposes, it'd be a good idea to have a CMake script that determines if the compiler supports C++11, and if so, #define HASHTABLE std::unordered_map, otherwise fall back to std::map
00:57 hmmmm and then change all the uses of std::map where we don't need ordering at all to that
01:28 damiel joined #minetest-dev
01:28 damiel joined #minetest-dev
04:34 us`0gb joined #minetest-dev
04:43 proller joined #minetest-dev
05:07 spillz joined #minetest-dev
05:15 blaise joined #minetest-dev
05:28 spillz Any thoughts on how to best go about adding a death animation to mobs? At most simple I want mob to fall down, lie there for a bit, then disappear. (Assuming this isn't already implemented somewhere)
05:28 spillz left #minetest-dev
05:58 ShadowNinja Minetest's position hashing seems to be black magic after I struggeld with it for a few days.  I propose reverting my patches, which somehow worked, although they shouldn't have.  And then adding a big "Black magic!" header to that section of code.
05:58 ShadowNinja The bit patterns it generates are non-sensical, but it works.
06:00 ShadowNinja sfan5: Finally got my minetestmapper patch working: http://sprunge.us/acHA?diff
06:05 ShadowNinja !tell sapier I didn't add that "Some exception: " catch-all.  It's only active in release builds.  "Access violation" sounds like Windows's version of SIGSEGV.
06:06 ShadowNinja ~config channel reply.whenaddressedby.chars
06:06 ShadowBot ShadowNinja: O.K.
06:06 ShadowBot ShadowNinja: !~
06:06 ShadowNinja Huh.
06:06 sfan5 "Access violation" sounds like Windows's version of SIGSEGV.
06:06 sfan5 it is
06:07 ShadowNinja O.K.  Good to know.
06:07 sfan5 hm
06:07 sfan5 with the minetestmapper patch leveldb uses individual block fetching and sqlite3 uses multiple, right?
06:09 sfan5 ^ ShadowNinja
06:10 ShadowNinja sfan5: Yep.
06:10 sfan5 k, then you can merge it
06:10 ShadowNinja sfan5: They both currently fetch ALL blocks on that Z axis though.  But that's an optimization for later.
06:15 ShadowNinja Pushed.
06:16 ShadowNinja I'd like to convert to just a getBlock(), as the fact that you can efficiently fetch blocks along the Z axis is really a artifact of how the hashing works.
06:42 PenguinDad joined #minetest-dev
07:10 sapier joined #minetest-dev
07:15 grrk-bzzt joined #minetest-dev
07:21 hmmmm ShadowNinja, how is it 'black magic'?
07:23 hmmmm seems pretty straightforward to me
07:24 hmmmm I'll agree that trying to use natural number math on what's clearly packing bits into an integer was a horrible choice though
07:27 hmmmm one thing I completely don't understand is why the resulting integer is signed, though.
07:28 hmmmm this causes nothing but problems.  in Database::getIntegerAsBlock, when values are shifted to the right, their signs are going to be extended (but masking the bits you want fixes this)
07:29 hmmmm so in that operation you're taking a signed value, doing a signed bitshift right, and then implicitly converting it to an unsigned 64 bit type when you operate on it with the bitwise AND
07:29 hmmmm bit operations, aside from bitshift, always implicitly promote the operand to unsigned.
07:34 sapier bitmaps aren't very usefull if you wanna do things in parallel for that reason. You always have to put a lock around it in order to get it done correct
07:35 hmmmm true yes
07:35 hmmmm i don't think loadBlock is going to be called in ServerThread very often though, the DiskEmergeThread would be the one to hold it mostly
07:35 sapier wtf ... why does current master have a 1s delay if you click buttons in menu?
07:36 sapier wait
07:36 hmmmm the time it takes to acquire the lock is no doubt much, much less than it is to prepare and execute a SQL query and perform disk access though
07:37 hmmmm a userspace lock at that
07:37 sapier ok on comparing it to sql times that's a whole different story :-)
07:41 specing joined #minetest-dev
07:43 sapier it's definitively something in current master
07:46 darkrose joined #minetest-dev
07:46 darkrose joined #minetest-dev
07:49 sapier hmm maybe it's broken for quite some time and I did only notice it because of accidentaly fixing it in refactored menu
07:56 sapier I'm gonna push #1130 in about half an hour if noone complains
07:56 ShadowBot https://github.com/minetest/minetest/issues/1130 -- Jthread: remove locks that arent absolutely required by sapier
08:01 hmmmm the code itself looks fine, but you have some typos like a double semicolon ;; when calling nanosleep
08:01 hmmmm and same for IsRunning()'s definition
08:02 sapier I wonder why compiler doesn't give a warning about things like that
08:03 hmmmm ??  why should it
08:03 hmmmm it's not like there's any potentially dangerous behavior
08:03 hmmmm it's just a statement that does nothing
08:03 hmmmm ;;;;;;;;;;;;;;;;;;;
08:03 sapier but it's most likely unintended and user possibly wanted to do something else
08:04 hmmmm well think about loops without a body
08:04 hmmmm err nevermind bad example
08:04 sapier :-)
08:05 hmmmm yea you're right
08:05 hmmmm if (foo);
08:05 hmmmm \tDoSomething();
08:05 hmmmm the potential is there i suppose, if you don't pay much attention to what you're coding...
08:06 sapier that's a quite common mistake ... no matter I fixed it
08:06 hmmmm heh, proller removed finite liquid and weather
08:06 hmmmm woohoo
08:06 hmmmm that was some very ugly code
08:06 sapier I'd still have prefered if he did cleanup and fix it
08:07 hmmmm yeah, but making code good isn't fun
08:07 hmmmm ...i guess.
08:07 sapier who do you tell ... ;-)
08:07 ImQ009 joined #minetest-dev
08:08 sapier I've just cleant up the mess I made on (re) adding all thos special (old menu) features to mainmenu after initial rewrite
08:08 Calinou joined #minetest-dev
08:08 hmmmm i can't believe you rewrote the entire mainmenu with that gross formspec syntax
08:09 hmmmm if it were up to me, i'd fix the formspec language before attempting to do any significant amount of work using it
08:09 sapier https://github.com/sapier/minetest/tree/fstk_addon this one is supposed to get mainmenu to a maintainable state
08:10 hmmmm wow, great stuff, so you're making an entire UI toolkit now
08:10 hmmmm in lua too
08:10 sapier imho fixing formspec in a incompatible way is useless either we're gonna replace it by some standard toolkit or we do just as much as necessary to keep it sufficient
08:10 hmmmm not in an incompatible way
08:10 sapier there's no way to really fix it in a compatible way
08:11 hmmmm just add a variety of language frontends so it can be done nicely and reverse compatible as well
08:12 sapier that's almost impossible, too much context relation between formspec elements
08:12 sapier it's gonna get a very very tough task to maintain those converters
08:12 hmmmm what do you mean
08:13 sapier e.g. a button and a label having same text created on same y pos won't show their text on same y pos
08:13 sapier buttons do have some padding based uppon current window size
08:14 hmmmm oh those are just the layout rules
08:14 hmmmm not the syntax
08:14 hmmmm hmm
08:14 sapier ok then the famous "size[]" tag used to add a "continue" button
08:15 sapier to be exact ... lack of "size" causes a continue button to be added
08:15 hmmmm minetest formspec comes in a close second in "worst layout language" to CSS
08:15 hmmmm who came up with this
08:15 sapier it's grown ... like all things beeing that broken
08:16 sapier but it's doing (almost) everything we need ... a full sane replacement will be about a month of work ... which I'm not willing to spend on it
08:17 sapier and after that month you won't have a mainmenu nor converters
08:29 sapier ok pushing 1130
08:47 jin_xi joined #minetest-dev
09:15 rsiska joined #minetest-dev
09:26 sapier #1178 any complains?
09:26 ShadowBot sapier: Error: Delemiter not found.
09:26 sapier Replace deathscreen by formspec variant       #1178
09:27 ShadowBot sapier: Error: Delemiter not found.
09:37 sapier I'm gonna push #1204 in about half an hour if noone complains
09:37 ShadowBot sapier: Error: Delemiter not found.
09:38 sapier as well as #1201
09:38 ShadowBot sapier: Error: Delemiter not found.
09:39 sapier ShadowBot: shut down
09:44 sapier xyz I'd use ctype instead of ctype.h but my compiler doesn't even find ctype without .h
09:45 Calinou joined #minetest-dev
09:45 sapier never mind ... found out it's cctype
09:46 sfan5 c headers in c++ are always "#include <c" + c_header.replace(".h", "") + ">"
09:47 sapier it's not often you se a change that useless in a standard
09:47 sapier that's a "we wanna make it different" change ... ;-/
09:48 sapier well wasted time to discuss about ... sfan5 any comments to the dethscreen replacement?
09:49 emptty joined #minetest-dev
09:49 sfan5 looks good
09:50 sapier anyone interested in creating a api for mods to replace those built in formspecs? ;-)
10:04 BlockMen joined #minetest-dev
10:18 Taoki[laptop] joined #minetest-dev
10:22 sapier merging #1178 in about half an hour
10:22 ShadowBot sapier: Error: Delemiter not found.
10:26 BlockMen sapier, before switching more to formspec its would be wise to fix #1223
10:26 ShadowBot https://github.com/minetest/minetest/issues/1223 -- MT (sometimes) crashes when leaving pause menu under Windows
10:26 BlockMen or at least find the reason for the crashes
10:37 Taoki[laptop] joined #minetest-dev
10:38 proller joined #minetest-dev
10:39 kattsmisk joined #minetest-dev
10:41 Taoki[laptop]_1 joined #minetest-dev
10:42 Taoki[laptop]_1 joined #minetest-dev
10:46 Taoki[laptop] joined #minetest-dev
10:57 sapier Blockmen can you provide logs for this one?
11:04 sapier I tried with sfan's build 674be38, no crash for me
11:07 tomreyn joined #minetest-dev
11:07 Calinou joined #minetest-dev
11:08 sfan5 FYI I updated to mingw-gcc 4.8 and recompiled irrlicht (1.8.1) from <build where problem was reported> to <latest build>
11:09 sapier ok so the bug is gone?
11:10 sfan5 did you say that it does not happend for you?
11:10 sfan5 I don't use windows, idk
11:10 sapier we need to remove that "something bad happened" error message it's useless to us. There should be some additional message like "If you can reproduce this please start minetest on console and provide console as well as debug.txt file to minetest developers"
11:11 sapier it doesn't happen to me neither on linux nor on windows
11:11 sfan5 Minetest should have a debug mode in which it fremap()'s stderr to stderr.txt to collect additional info (e.g.) stack traces that do not end up in debug.txt
11:12 sapier that wouldn't help if users only provide screenshots of a crapy error message ;-)
11:13 sfan5 driver->createScreenshot (or similar)
11:13 sfan5 we should have automatic bug reporting
11:14 sapier no ... I hate applications calling home without asking ... but there should be a easy way e.g. assisted guide to provide all possible information to developers
11:14 sfan5 automatic bug reporting does not imply not asking the user
11:15 sapier true ... but we'd need to write another server application to get those data to github
11:16 sapier ok pushing deathscreen now
11:17 sfan5 writing server apps isn't hard
11:17 sfan5 and we probably don't need them in github automatically
11:17 sfan5 someone could do it manually
11:17 sapier but merging identical bugs on different os/versions to a single issue could be hard
11:18 nore joined #minetest-dev
11:19 sapier we don't have enough manpower to merge pull requests how are we supposed to handle hundreds and thousands of automaticaly provided bug reports
11:19 sapier any comments on #1206?
11:19 ShadowBot https://github.com/minetest/minetest/issues/1206 -- Remove a lot of superflous ifndef USE_CURL checks by sapier
11:22 sapier sfan are you aware that tab key for playerlist can't be changed in settings? (even if you did implement this missing part)
11:25 Jordach joined #minetest-dev
11:26 sfan5 sapier: cannot be changed?
11:26 sfan5 why?
11:26 sapier I don't know but you can't use key setting to set a key to tab ... tried yesterday to reproduce a bug pilzadam had
11:27 sapier the only way to do it is changing minetest.conf with text editor
11:28 sfan5 yeah but it can be changed (even if not with the GUI)
11:29 sapier If you wanna use it as default users have to change it and be able to get the original setting back
11:29 sapier from gui
11:29 sfan5 add a dropdown
11:30 sfan5 oh.. also add some info text when selecting leveldb as world type
11:30 sapier I don't care how you do it ;-)
11:30 sfan5 aka "this world will not be compatible with all minetest builds"
11:30 sapier can't do this
11:31 sfan5 no?
11:31 sapier no room left on settings tab ;-) and we don't support tooltips for dropdowns (yet)
11:31 Taoki[mobile] joined #minetest-dev
11:31 sfan5 >and we don't support tooltips for dropdowns (yet)
11:31 sapier and no I wont add thisjust to add a dropdown forget about it
11:31 sfan5 do whatever is needed to remove the "yet" :-)
11:32 Taoki[mobile] joined #minetest-dev
11:32 sapier some features have to be left for others to implement
11:36 BlockMen joined #minetest-dev
11:36 BlockMen sapier, wtf? why have you merged it now?
11:37 sapier because noone can reproduce it and even if there is a bug it's a single one
11:38 sapier so if you can provide the information I requested I'm quite sure it can be fixed in a couple of minutes
11:38 BlockMen 1st) how else uses windows and can test it?
11:38 sapier I did test on windows
11:38 BlockMen 2nd) just the fact you cant doesnt mean there is
11:38 BlockMen 3) what about the 2 forum posts?
11:39 BlockMen 4th) there is no need to merge it now, so why that hurry?!
11:39 sapier Bugs without information can't be fixed, according to that information it's not even clear it's formspec related
11:40 BlockMen well, when it appears just after the switch to formspecs and is crashing after the fields are send, what else should it be related to?
11:40 sapier instead of discussing why don't you just provide the information requested so we can fix it?
11:40 BlockMen 5th) when the game crashes by using pause menu it IS a blocker, dont remove tags senseles
11:40 sapier timing?
11:41 sapier sometimes for someone isn't a blocker to me
11:41 sapier everytime for (almost) everyone is a blocker
11:41 BlockMen at least for you
11:42 sapier I'm not gonna continue a discussion stopping us from finding it, can you reproduce it?
11:42 BlockMen yeah, i can. and we already talked about that...
11:42 sapier In case you can please post console and debug.txt
11:43 BlockMen and this discussion is not senseless like you want to say. its just odd to push something that can break even more
11:45 sapier if you can't provide information about this bug I'm gonna consider it to be a phantom only happening for you
11:45 BlockMen so you ignore user posting...great
11:45 PenguinDad joined #minetest-dev
11:46 sapier you're experienced enough to find debug.txt on your machine and capable of starting a console, if you don't do this to get the bug fixed I consider your behaviour to be childish
11:46 BlockMen my behaviour is childish? because i take of bug reports?
11:46 BlockMen ok, fine. than i will follow your style, close the issue
11:46 sapier I ignore user postings just ignoring any request to provide information required to help them
11:47 BlockMen fine, idc then anymore
11:47 BlockMen do you sapier-thing
11:47 BlockMen *+r
11:47 proller 8)
11:47 sapier hurray ... another half an hour wastet just because someone doesn't want to provide debug.txt
11:48 BlockMen sapier, ppl have other things to do than just provide instantly information of bugs ypu produced
11:48 BlockMen i said there is a problem, i never said i woant try to rpovide more information
11:48 BlockMen *wont
11:49 sapier well I asked how many times now to provide that information? ... without any reaction ?
11:49 BlockMen i was OFFLINE
11:49 Taoki[mobile] joined #minetest-dev
11:49 BlockMen ans that was 1 hour! ago
11:49 BlockMen my god, and im the childish??
11:49 sapier 13:40:20) sapier: instead of discussing why don't you just provide the information requested so we can fix it?  ... you've been offline since that time?
11:50 sapier ok please stop accusing each other, I guess you wanna tell you don't have time to provide that information now, ok no problem, once you did I'm gonna fix this bug next
11:51 sapier we both want bugs to be fixed as soon as possible. And if this one is real it's most likely quite easy to fix once propper information is available.
11:55 werwerwer joined #minetest-dev
11:55 PilzAdam joined #minetest-dev
11:56 kaeza joined #minetest-dev
12:00 BlockMen here little diva, more information are not possible to get
12:00 BlockMen http://pastebin.com/LJtZWXmV have fun
12:01 sapier thanks
12:08 emptty1 joined #minetest-dev
12:08 sapier1 joined #minetest-dev
12:09 Taoki[mobile]_1 joined #minetest-dev
12:13 Taoki[mobile]_1 joined #minetest-dev
12:20 celeron55 oh hell
12:20 celeron55 https://forum.minetest.net/viewtopic.php?id=9103
12:20 OldCoder joined #minetest-dev
12:20 celeron55 people are just shouting random games that they have no involvement with whatsoever
12:21 celeron55 ...
12:21 celeron55 i hate this
12:21 Gethiox joined #minetest-dev
12:21 celeron55 i already deleted two such replies yesterday; do i have to delete like most replies
12:23 BlockMen celeron55 make 1) in red and bold ;)
12:32 BlockMen sapier, if you cant fix right now re-add the "blocker" label
12:36 celeron55 now there is a separate topic for voting
12:47 PilzAdam joined #minetest-dev
12:49 celeron55 so, then one thing:
12:52 celeron55 should there be some kind of an official and recommended place for writing down and voting for feature requests for the engine, so that it is only meant for people building mods and subgames?
12:53 celeron55 currently it seems to be very hard to figure out what features of that form are actually needed by how many people
12:55 celeron55 each person mostly just knows what they personally want, and as the modders and engine developers largely aren't the same people, that could help a lot
12:55 PilzAdam sapier1, please dont add a GUI option for db backends
12:56 celeron55 it might be even possible to catch some of the random C++ developers coming here to develop some of that stuff
12:56 PilzAdam it would simply result in incompatible formats all over the place
12:56 PilzAdam most users should stick to sqlite
12:57 proller return falling nodes to c++ !
12:57 sapier1 but it's crap to have a "Ifdef android" around database format
12:57 celeron55 proller: ?
12:57 sapier1 I'm gonna try if sqlite will work on android too maybe the issue vanishes
12:58 proller celeron55, just random feature needed in core ;)
12:58 celeron55 proller: that kind of shout votes are exactly what makes things hard
12:58 celeron55 proller: so thank you for the example
12:58 celeron55 proller: now, maybe you could vote also for the thing i suggested
12:58 celeron55 or against it
12:59 sapier I'm positive about the voting system
13:00 proller best modders - peoples who can change core
13:00 celeron55 proller is making some kind of a world record in uselessness of communication at the moment
13:01 proller who understand how core works, they can describe what they want from core
13:03 celeron55 make a table of default database formats per platform, then it's clean
13:04 celeron55 ^ to the issue above
13:04 PilzAdam http://wiki.minetest.net/Database_backends
13:05 celeron55 we should take into use sqlite_synchronous = 0 so that it isn't slower than leveldb
13:05 celeron55 and also enable sqlite's locking so that it can't be opened by two instances at the same time
13:05 celeron55 than nobody has any reason to use leveldb
13:06 PilzAdam except that sqlite gets buggy when bigger than 4 GiB
13:07 celeron55 well except that; nobody has yet figured out for sure whether it is a bug in minetest or sqlite though
13:09 celeron55 if you try to search for the issue in google, you can only find statements that it supports something way more than 4GB
13:36 Zeitgeist_ joined #minetest-dev
13:36 Zeitgeist_ joined #minetest-dev
13:40 kaeza anyone builds with MinGW/MSYS?
13:40 sapier <<
13:40 kaeza I'm getting a weird error and not sure how to fix
13:41 kaeza https://gist.github.com/kaeza/defa19f73e6e123b3b61
13:42 kaeza apparently MS recommends using GetProcAddress to call this function: http://msdn.microsoft.com/en-us/library​/windows/desktop/ms684139(v=vs.85).aspx
13:42 sapier you do a 32 bit build but want to use a 64 bit api call that's not possible
13:43 sapier thought this was fixed quite some time ago
13:44 kaeza soo... my MGW is b0rked?
13:44 sapier no minetest is broken
13:44 sapier using 64bit api in 32bit build is stupid
13:44 kaeza when was that fixed?
13:49 kaeza1 joined #minetest-dev
13:49 sapier I don't know we discussed this about 2 months ago if I remember correct
13:50 kaeza1 will try pulling again to see if it fixes
13:51 EvergreenTree joined #minetest-dev
14:09 BlockMen joined #minetest-dev
14:30 BlockMen Just for those who are interested. sapier cant fix that issue as easy as promised (http://irc.minetest.ru/minet​est-dev/2014-04-19#i_3657171) and he is able to reproduce the issue on MSVC2012.
14:30 BlockMen I re-add the block label to #1223 since sapier does not (for what reason ever).
14:30 BlockMen Here a protocoll of private conversation between him and me (german) http://pastie.org/private/nl8ypgjx4wwt2o7mdnygow
14:30 ShadowBot https://github.com/minetest/minetest/issues/1223 -- MT (sometimes) crashes when leaving pause menu under Windows
14:40 hmmmm joined #minetest-dev
15:15 EvergreenTree joined #minetest-dev
15:30 rdococ joined #minetest-dev
15:35 Garmine joined #minetest-dev
16:22 zat joined #minetest-dev
16:26 EvergreenTree joined #minetest-dev
16:38 jin_xi joined #minetest-dev
17:54 ShadowNinja hmmmm: Becuase of the weird way it packs positions.  (0, -1, -1) becomes fffffffffefff000, rather than the more logical choice of ffffff000.
17:55 ShadowNinja <hmmmm> if (foo); <hmmmm> \tDoSomething();  Remember? That happened in the shaders code.
17:55 sapier #1238 is supposed to fix #1223 ... irrlicht isn't capable of handling it's gui elements correctly, deleting a gui element from within a run step doesn't seem to be a good idea on msvc. If someone has an idea why gcc code does handle this way better I'd be curious.
17:55 ShadowBot https://github.com/minetest/minetest/issues/1238 -- Fix menumanager to handle irrlicht gui elements more correct by sapier
17:55 ShadowBot https://github.com/minetest/minetest/issues/1223 -- MT (sometimes) crashes when leaving pause menu under Windows
17:57 ShadowNinja !web title <issue 1178>
17:58 ShadowBot Replace deathscreen by formspec variant by sapier · Pull Request #1178 · minetest/minetest · GitHub
17:58 ShadowNinja Hmmm.  Not sure what was happening before.
17:58 sapier happening before what?
17:58 ShadowNinja [ShadowBot] sapier: Error: Delemiter not found.
17:58 sapier that one :-)
18:00 luizrpgluiz joined #minetest-dev
18:00 luizrpgluiz when going out version 1.0 of minetest?
18:00 sapier not anytime soon luizrpgluiz
18:01 sapier lemme guess you're using windows 8? ;)
18:05 luizrpgluiz Thank god no
18:06 sapier but windows is correct? or do I have to revoce my prejudices? :)
18:12 hmmmm ShadowNinja, it does??
18:12 hmmmm you're sure??
18:12 hmmmm that's screwed up
18:12 ShadowNinja hmmmm: Yep.
18:13 hmmmm the position packing seems to be a bug
18:13 ShadowNinja hmmmm: I messed with this for a few days with the mapper until I gave up.
18:13 luizrpgluiz I did not like windows 8, I like the windows 7 and linux installed on my dual core notebook
18:13 hmmmm dammit
18:13 hmmmm if only reverse compatibility wasn't an issue
18:13 ShadowNinja ^
18:14 hmmmm what do you guys feel about doing the position integer packing the correct way, making all new stuff use that, leaving the old stuff in, and then removing it after a certain length of time when we're fairly certain we can drop reverse compatibility
18:15 luizrpgluiz ?
18:16 hmmmm (by the way, i'm not 100% knowledgable on the topic, but there is a map DB version as well, correct?)
18:16 luizrpgluiz I did not like the windows 8 interface because it drove me crazy
18:18 ShadowNinja hmmmm: Nope, just a block version.
18:18 hmmmm ugh
18:18 hmmmm alright then
18:18 hmmmm this IS a problem
18:18 hmmmm we should start versioning map files themselves.
18:19 ShadowNinja insert into blocks (pos, data) values ("version", 0);
18:19 ShadowNinja ^ That will work since SQLite3 doesn't enforce column data types.
18:19 hmmmm huh
18:19 hmmmm blocks is a table type
18:20 hmmmm shouldn't there be like a blocks table, and general info table
18:20 hmmmm and the info table will contain the version key/val
18:20 hmmmm ??
18:20 ShadowNinja And it will work for LevelDB too since it uses strings for keys.
18:20 luizrpgluiz left #minetest-dev
18:20 ShadowNinja Yes, that yould be done too.
18:20 hmmmm too??
18:20 ShadowNinja -y+c
18:20 hmmmm why not do just one
18:20 hmmmm (the one that makes more sense)
18:20 ShadowNinja s/too/instead/
18:21 hmmmm I personally like having a separate table for general information
18:21 hmmmm keep blocks with blocks
18:21 hmmmm oh boy so here's another thing TODO
18:21 hmmmm it's not a pressing issue, it's just that this is a bad part of minetest that should be fixed sometime
18:21 hmmmm wait, we had good parts of minetest?  8)
18:56 Robby joined #minetest-dev
19:13 kaeza1 joined #minetest-dev
19:15 Megaf joined #minetest-dev
19:15 Megaf folks
19:15 Megaf two new and annoying serious bug
19:16 Megaf one is server side
19:16 Megaf that16:16:20: ERROR[ServerThread]: Server::ProcessData(): Cancelling: peer 18 not found
19:16 Megaf Sometimes it happens only for a couple of minutes
19:17 Megaf and sometimes it will stay happening for hours
19:17 Megaf and another bug is client side, when you press F10, minetest stop accepting any input
19:18 ShadowNinja Megaf: 1) That's not really a bug, it should be sent to infostream.
19:19 ShadowNinja Megaf: 2) Wow, wonder who broke that.  sapier: ? ^
19:20 Megaf ShadowNinja: not a bug?
19:20 sapier as you said it's not really broken ;-)
19:21 Megaf It makes the server stop responding
19:21 sapier update your client
19:21 Megaf I really hate you folks
19:21 Megaf you are fuckking ignorant
19:21 sapier this happens in case of client sending crap
19:21 rsiska joined #minetest-dev
19:21 Megaf I dont make the client
19:22 Megaf I run a server
19:22 ShadowNinja Megaf: That's an error, but it isn't an important one.  You can ignore it.
19:22 sapier it's not beeing ignorance but having checkt that particular issue about 10 times alway with same result
19:22 Megaf That's a new bug
19:22 Megaf it didnt happen before
19:22 sapier it did you just didn't see it because of client just silently hang
19:22 Megaf and server stops reposing
19:23 sapier completely or for this client only?
19:23 Megaf well, no one can join my server now and play because of that
19:23 sapier ok that's completely different to what you first said
19:24 Megaf I said that the server will stop responding while that is happening
19:25 sapier stop responding to everyone? or that client?
19:25 Megaf everyone
19:25 Megaf you can even login/join/enter the server
19:26 sapier you've got current master?
19:26 GhostDoge sapier: I noticed this too in the same way as Megaf
19:27 kaeza1 joined #minetest-dev
19:27 sapier ok lets try to find out what you've got in common, last things change there have been tested on vanessae's servers for about 2 months without issues
19:28 sapier megaf? os, compiler, cpu?
19:28 sapier line bandwith
19:33 sapier come on guys it's quite annoying if you report a unreproduceable issue and run away right after it ...
19:34 PenguinDad sapier: Fedora 20, clang, AMD Phenom II X6 1100T
19:35 sapier line width, typical amount of players?
19:36 PenguinDad It was a local server with 2 players and 100 mbit/s bandwith
19:38 sapier is there a way to reproduce?
19:39 PenguinDad and it happened when the server was under high load due to water flowing down in a singlenode map, everything burning and me and my friend killing each other
19:39 sapier is it possible load was as high as making peers timeout?
19:41 PenguinDad but it might also have been my slightly broken router that crashed under the load :/
19:42 Megaf sapier: it happens on ARM and on i32
19:42 Megaf my server runs on ARM
19:42 sapier that error message is shown by server on reception of a packet from a peer that has already timed out
19:43 Megaf I know that
19:43 Megaf it happens when someone is connecting and gives up
19:43 Megaf but that should not make the server stop responding and make it send millions of messages each second
19:44 sapier well the console output may cause additional load especially on arm ... but there has to be some other reason triggering the initial overload causing peers to timeout
19:45 Megaf sapier: it's a 4 core 1,2 GHz CPU with 2 GB of RAM
19:45 Megaf I got this board only to run minetest server
19:46 PenguinDad well for me it could have been an entity that fell down the SkyBlock map
19:46 kaeza it seems updating to latest git solved the IsWow64 thingy
19:46 * kaeza wonders how old his repo is
19:48 sapier megaf what's your typical cpu load?
19:48 Megaf 13%
19:49 Megaf 11% right now
19:49 sapier any mods on it that may cause lag?
19:49 Megaf Only mesecons and pipeworks
19:50 Megaf Usually then dont add much load
19:50 sapier "only" :-)
19:50 Megaf even complex logic gate circuits
19:51 sapier my current asumption is something causes a temporary lag beyond client timeout causing peers to timeout and thus causing those messages causing even more load
19:52 sapier you could falsify that assumtion by trying to reproduce it on vanilla minetest game
19:55 Megaf sapier: it happens to more servers,
19:56 Megaf theres a topic about it and also an issue report
19:57 Taoki[laptop] hmmmm: Hey there. I was wondering if latest mapgen v7 has support to define your own fluid node (for lakes and oceans) per biome, just like ground nodes. Having some ideas in mind which will likely use custom liquid nodes.
19:57 hmmmm yes it has that ability
19:58 sapier what issue megaf?
20:00 hmmmm Taoki[laptop]:  see node_water and node_dust_water
20:00 Megaf https://github.com/minetest/minetest/issues/1231
20:01 Megaf ShadowNinja: And this one is about the F10 thing. https://github.com/minetest/minetes​t/issues/1239#issuecomment-40878129
20:03 sapier great ... you commented just now that it's blocking server too ;-) which is completely different to what was reported before ;-)
20:04 VanessaE_ I can't reproduce the F10 problem.
20:04 sapier I can
20:04 VanessaE_ maybe because my irrlicht is old
20:04 sapier the console appears and then keyboard is ignored
20:05 sapier as you're here VanessaE_ did you encounter a similar issue like megaf?
20:05 VanessaE_ could be related to PA's problem with tab causing formspec close events?
20:06 Megaf sapier: I updated my client yesterday, no problems with F10, and updated again 30 minutes ago, problems with F10
20:06 VanessaE_ you mean the problem described by issue 1231?  no.  I have not seen anything nearly that severe.
20:06 Megaf I'm suing the latest stable irrlicht
20:06 VanessaE_ my client was last updated yesterday.
20:06 Megaf Krock has the same issue
20:06 Megaf I mean, the peer one
20:07 sapier "same" as F10?
20:07 Krock joined #minetest-dev
20:07 VanessaE_ speak of the devil.
20:07 markveidemanis joined #minetest-dev
20:07 markveidemanis left #minetest-dev
20:08 spillz joined #minetest-dev
20:08 Megaf sapier: sorry, #1239
20:08 ShadowBot https://github.com/minetest/minetest/issues/1239 -- Client will stop responding when pressing F10. commit 8745935a06
20:08 Megaf oops
20:08 Megaf #1231
20:08 ShadowBot https://github.com/minetest/minetest/issues/1231 -- ServerThread ServerEnv & Server::ProcessData() ERRORS
20:08 Megaf This one
20:08 Krock left #minetest-dev
20:09 VanessaE_ Also, megaf, cut out this shit:
20:09 VanessaE_ [04-19 15:21] <Megaf> I really hate you folks ~~~ [04-19 15:21] <Megaf> you are fuckking ignorant
20:09 VanessaE_ You've already proven in the past that you tend to have a slightly "odd" environment compared to the others here, and the core devs are not your personal codemonkeys.
20:10 Megaf VanessaE_: not this time, just normal minetest things here
20:10 VanessaE_ nevertheless.  My point stands.
20:10 spillz so this: http://irc.minetest.ru/minet​est-dev/2014-04-19#i_3657765 is the same issue that I reported here: http://irc.minetest.ru/minet​est-dev/2014-03-20#i_3622350
20:10 sapier ok first of all I'm gonna fix the message ... It's not an error but just something that can happen if your peers behave had
20:10 sapier bad
20:11 VanessaE_ sapier: and as you and I have experienced, that's about 75% of the minetest clients currently deployed :)
20:11 sapier next for some reason peer timeouts aren't logged that'd be way more usefull then the error message shown right now
20:11 Megaf indeed
20:12 Megaf sapier: I think you are right, much of the load is caused only by that message
20:12 Megaf because it happens several times each second
20:13 sapier well that's causing a runnaway cpu load
20:13 sapier I pushed the change
20:15 Megaf sapier: http://paste.debian.net/94599/
20:15 Megaf That's like 1/50 of all messages
20:15 sapier I wonder why that much peers time out on your machine
20:16 Megaf I believe it's a chain raction
20:16 Megaf 9 peers connected, one start that, and others begin to timeout
20:17 VanessaE_ I have never seen that happen on my servers before unless a shitload of tablets all try to connect to Survival or Creative or something (those servers are too heavy for tablets, so the tablet clients crash and time out)
20:17 VanessaE_ which is basically never
20:17 sapier if it's that close to the edge even a single mod bug may cause it
20:17 Megaf VanessaE_: yep, may be that, there are several players who play on tablets on my server
20:18 sapier ok megaf there are quite a lot of buggy tablet versions out there I'd not be surprised if there where tablet clients not working at all
20:20 Megaf Still, it should not make the server act like that
20:20 sapier it should be better now but I doubt it's gonna be a 100% solution
20:20 Megaf we should just block all thing that is not minetest...
20:20 sapier depends if those tablet clients cause a dos attack your server will break
20:20 sapier minetest is in no way hardened to dos attacks
20:22 kaeza1 joined #minetest-dev
20:27 proller hmmmm, no problem to transfer air block, but hard to updatedrawlist when you have 3000 blocks, most of them air and underground and skipped, but every block checked every 200ms
20:27 sapier but let's try fist if this change helps
20:28 Megaf ok, I just opdated the server and restarted it
20:28 Megaf pipeworks working and using only 6%
20:29 Megaf 3 clients
20:30 sapier problem is not typical load but load spikes ... if those happen for e.g. 4 seconds it's gonna cause peers to timeout
20:31 nore joined #minetest-dev
20:36 Megaf Thanks for your help sapier
20:37 sapier If someone provides information I can work with I always try to do what I can megaf
20:37 sapier especially if I'm involved in the origin of that issue
20:39 proller sapier, try to setup and run your public server
20:39 Megaf Now we need to fix #1239
20:39 ShadowBot https://github.com/minetest/minetest/issues/1239 -- Client will stop responding when pressing F10. commit 8745935a06
20:39 proller sapier, you cant make connection better if you never use it
20:39 sapier proller I'm on dsl 3000 line that'd be of no use
20:40 proller optimize it!
20:40 proller get free vps
20:40 proller ask friends
20:40 Megaf sapier: I ran my server with only 100 KB/s of upload
20:40 sapier you can't optimize a connection to work perfect on small line AND big one
20:40 sapier and this doesn't seem to be a connection issue
20:41 proller if your connection will work on slow line - it will work on fast too
20:41 sapier no it wont
20:42 proller then revert your changes and newer touch connection
20:44 kaeza joined #minetest-dev
20:49 rdococ joined #minetest-dev
21:12 Miner_48er joined #minetest-dev
21:25 iqualfragile joined #minetest-dev
21:52 grrk-bzzt joined #minetest-dev
22:03 kaeza1 joined #minetest-dev
22:07 Megaf sapier: Well done
22:07 Megaf 19:05:43: ACTION[ServerThread]: tysonlover607 joins game. List of players:
22:08 Megaf 19:07:13: ERROR[ConnectionSend]: con(5/1)RunTimeouts(): Peer 94 has timed out. (source=peer->timeout_counter)
22:09 BlockMen left #minetest-dev
22:10 sapier better?
22:13 Megaf yep
22:13 Megaf server will no longer hang
22:13 kaeza sapier, which irrlicht version do you use to build with mgw?
22:13 sapier 1.8.1
22:14 kaeza hmm...
22:15 sapier wait it's 1.8 only
22:15 kaeza any ideas? https://gist.github.com/kaeza/ec1ca2c30c456777bd9b
22:15 kaeza oh
22:17 sapier guess you're linking to another lib then the one specified in your headers
22:17 kaeza -- IRRLICHT_INCLUDE_DIR = C:/MinGW/home/Diego/irrlicht-1.8.1/include
22:18 kaeza hmm...
22:18 sapier if header is found in some path prior this one that one is used

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