Minetest logo

IRC log for #minetest-dev, 2013-12-06

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

All times shown according to UTC.

Time Nick Message
00:02 djdduty joined #minetest-dev
00:04 john_minetest left #minetest-dev
00:05 BrianT In what respect?  I know how to code if that's what is in question.  I've refactored fairly large programs if that is a concern.
00:06 BrianT But I can also find something else to do, no problem either way.
00:10 sapier Usually people start with small changes to learn how code is connected .. minetests code isn't exactly staight forward and refactoring main most likely will break a lot of things
00:11 sapier It's not you can't possible do it right ... it's I just don't know if you can do it ;-) There's no way I can know this
00:12 BrianT That's fair enough
00:13 sapier but if you wanna take the risk you can refactor main too ... just don't be upset if the way you do it isn't the way to be merged
00:13 BrianT I'm not easily upset or offended
00:13 BrianT all part of the gig
00:14 sapier noone is happy if hours of work are sent to trash ;-)
00:14 BrianT agreed
00:15 BrianT Wasn't saying I would refactor main, just letting you know that I'm not easily offended
00:15 proller BrianT, can you make mac port?
00:16 sapier that's good ... and actually a major requirement for contributing to minetest ... sometimes discussions/decisions aren't that friendly :-)
00:16 BrianT Probably not, I'm under linux, and don't have the VM working well enough to do mac
00:17 sapier android port is waiting for completion too ;-)
00:17 BrianT I haven't done anything with android to be honest
00:17 sapier was worth a try :-)
00:18 BrianT Ya, I've been asked to do android before, even downloaded the sdk
00:18 BrianT but you know how life is
00:21 sapier if you want to do bigger things tell ppl in this channel this way you can avoid more ppl working on same thing ... and have a look at pull request list lots of improvements are already done and waiting there
00:21 BrianT okay, will do then
00:22 VanessaE joined #minetest-dev
00:34 PandemoniuM left #minetest-dev
00:48 OldCoder joined #minetest-dev
00:52 EvergreenTree joined #minetest-dev
01:37 celeron55 joined #minetest-dev
01:48 VanessaE um, guys...  what happened to the ultra-fast responses of the sqlite rollback check?  it's dog slow now
01:49 VanessaE I mean slow like molasses, as in a minute or more to do a simple 1-node-radius check.  ShadowNinja mentioned it being equally slow on his server also.
02:10 jojoa1997|PC joined #minetest-dev
02:10 jojoa1997 joined #minetest-dev
02:13 Miner_48er joined #minetest-dev
02:29 zat1 joined #minetest-dev
02:49 ShadowNinja zat1: ^ SQLite rollback isn't quite as fast as it was when it was first implemented it seems. Could I have changed something to cause this? (TIMESTAMP resolves to NUMERIC, is this somehow slower than INTEGER?)
02:55 zat1 ShadowNinja: It should, IIRC NUMERIC is actually TEXT for the engine.
02:55 zat1 I can be totally wrong though...
02:55 zat1 try reverting that
03:03 ShadowNinja Of course SQLite is dynamically typed... /me starts testing.
03:04 werwerwer_ joined #minetest-dev
03:27 ShadowNinja A column that uses INTEGER affinity behaves the same as a column with NUMERIC affinity. The difference between INTEGER and NUMERIC affinity is only evident in a CAST expression.
03:28 ShadowNinja ^ http://www.sqlite.org/datatype3.html Section 2
03:30 ShadowNinja So assuming our version is up-to-date and the docs are acurate, this can't be the issue.
03:35 BrianT left #minetest-dev
04:16 ShadowNinja Our packaged version is two years old... Is there a good reason that SQLite is packaged?
04:40 ShadowNinja sqlite> SELECT typeof(`timestamp`) FROM `action` LIMIT 1; --> integer
04:43 ShadowNinja There have been a lot of performance improvements and bug fixes added to SQLite since our version.
04:43 hmmmm i agree sqlite should be updated
04:55 ShadowNinja hmmmm: Do you know why it is bundled?
05:03 hmmmm hmm I guess a reason could be that updating sqlite3 could potentially be a pain
05:03 hmmmm it's nice to not have an extra dependency like that
05:03 hmmmm or maybe it uses the system sqlite3 if available
05:07 ShadowNinja hmmmm: Um, isn't keeping it bundled more a pain than adding a small dependancy?
05:09 VanessaE to be fair, it already requires libsqlite3-dev to be built in the first place
05:09 VanessaE so does that really count as "adding" a dep?
05:09 hmmmm erm
05:10 ShadowNinja Lua should also be removed IMO, as soon as we have Lua5.2 support. (Or just depend on LuaJIT)
05:10 hmmmm I just checked and the system sqlite3 is used unless you explicitly specify to use the builtin version
05:10 VanessaE ShadowNinja: never gonna happen
05:10 hmmmm the idea is that we want minetest to work even if the people don't have all the proper libraries installed
05:10 ShadowNinja Hmmm, is there any reason to keep the builtin version?
05:10 VanessaE the last time Lua 5.2 was proposed, that idea was shot down in flames.
05:11 kaeza ShadowNinja, Lua community *recommends* embedding Lua sources in your project
05:11 hmmmm why don't you just update the builtin version
05:11 hmmmm it's not used all the time like you're implying it is
05:12 ShadowNinja hmmmm: I wasn't, but I don't think it should be included, as we have to keep it updated and so far I don't see a reason to keep it.
05:13 hmmmm see what other people say
05:13 hmmmm maybe they have a better reason why than i do
05:16 ShadowNinja kaeza: That seems odd. Why so? And LuaJIT recommends the oposite.
05:17 hmmmm the latter is because LuaJIT has a very specific build setup that most people screw up
05:18 kaeza ShadowNinja, because it's less effort to keep a version yourself than having to deal with different naming conventions on distros (lua vs lua-5.1 vs lua5.1 va ...)
05:18 kaeza and Lua does not change that often anyway
05:19 ShadowNinja kaeza: We have that with every other package that MT depends on, packagers manage it for their distribution.
05:20 VanessaE hmmmm: define "a very specific build setup".  All I ever need to do with it is export CXXFLAGS="-O3"; make; make install.
05:21 hmmmm it talks about the reasons why on their website
05:21 VanessaE I've seen the reasons, just don't recall what they were now
05:21 hmmmm It's strongly suggested to build LuaJIT separately using the supplied build system. Please do not attempt to integrate the individual source files into your build tree. You'll most likely get the internal build dependencies wrong or mess up the compiler flags. Treat LuaJIT like any other external library and link your application with either the dynamic or static library, depending on your needs.
05:21 VanessaE though I seem to recall there was an air of "I want to be able to push updates to you" in there too
05:22 VanessaE oh well sure, the "individual source files"
05:22 VanessaE I don't think he means that in the sense of not including the whole package as a bundle
05:24 VanessaE if you grab a current git of the luajit-2.x branch and just throw it in there with a couple of lines in the main makefile that builds it, that'll be no different than the user building it separately and linking to it via minetest's manual cmake flags
05:24 VanessaE s/the main makefile that builds/minetest's makefile to build/
05:25 ShadowNinja VanessaE: So, why would we ever want to do this?
05:26 VanessaE ShadowNinja: I can see an argument for bundling it - a guarantee that the user will never be saddled with the old interpreted lua
05:27 VanessaE but the opposite side of the coin is that it could become a choice for the user to make, like yacc vs. bison or something
05:27 VanessaE (inb4 vi vs. emacs)
05:27 ShadowNinja VanessaE: Removing support for non-JIT Lua would also do that...
05:27 VanessaE of course.
05:27 VanessaE but then the user has to track down luajit
05:27 VanessaE and I know for sure that is isn't available on at least debian 7
05:28 VanessaE (I had to build it from sources on my VPS, which runs same)
05:28 ShadowNinja VanessaE: Hmmm, poke emptty to add it then.
05:29 VanessaE anyway, for my own needs, bundling is a waste of time and effort - BUT for the average minetest user, it will be needed, for sure.
05:30 VanessaE because, may the logs forgive me, the average user is too unreliable (perhaps too stupid) to safely track down and install the mode esoteric dependencies like that one
05:31 VanessaE more*
05:31 OldCoder joined #minetest-dev
05:31 ShadowNinja Many distros have LuaJIT packages, and most users likely install from a package.
05:31 VanessaE what about on Windows?
05:32 VanessaE much as I hate to have to consider it, most of our userbase uses Windows
05:33 ShadowNinja VanessaE: Windows users use packages.
05:34 VanessaE sure, but they have to use search engines to find them.
05:36 VanessaE at any rate, if you think it's safe not to bundle things like sqlite, lua/luajit, etc. then so be it.
05:37 kaeza ShadowNinja, I had this in my logs from 4 days ago: http://pastebin.com/bFWhyfQM
05:37 kaeza anyway, if you think you can do this without breaking stuff, do it
06:04 darkrose joined #minetest-dev
06:04 darkrose joined #minetest-dev
06:12 damiel joined #minetest-dev
06:26 thexyz having everything bundled is better because it's easier to port minetest
08:28 werwerwer joined #minetest-dev
08:29 init joined #minetest-dev
08:30 proller joined #minetest-dev
08:30 khonkhortisan_ joined #minetest-dev
08:32 Ritchie_ joined #minetest-dev
08:33 us{0gb joined #minetest-dev
08:40 VanessaE joined #minetest-dev
09:56 proller joined #minetest-dev
10:47 darkrose joined #minetest-dev
11:04 AllegedlyDead joined #minetest-dev
11:59 sapier joined #minetest-dev
12:03 sapier imho as long as luajit errorhandling isn't as good as lua error handling lujit is not an option for default
12:38 Gethiox joined #minetest-dev
13:00 Ritchie joined #minetest-dev
13:07 VanessaE joined #minetest-dev
13:13 [PavelS] joined #minetest-dev
13:26 hmmmm joined #minetest-dev
13:26 thexyz current minetest/minetest master hangs completely when starting the game (single or multiplater) under windows, msvc build
13:26 thexyz reproducible in wine too
13:27 sapier most of jthread fixes have been done and tested on linux only by now
13:28 thexyz but win-only code was changed too, right?
13:28 sapier not to that extent but yes
13:29 sapier sadly I can't really debug on windows ... msvc requires licese for debugging and mingw build is broken
13:29 thexyz I see
13:30 thexyz commiting code without even compiling it is a bit meh if you ask me
13:31 thexyz and I don't see any problems with having a MSVC license, express version is free
13:31 sapier I don't care about windows any longer as there's no way to test those 1000 possible lib and compiler combinations ppl seem to require to work there ... if there's a defined set we support I'll test again but not until this is done
13:31 thexyz alright
13:31 thexyz so we no support windows now?
13:31 thexyz that's okay, I guess
13:31 thexyz everything's fine
13:31 sapier I can't tell what "we" do ... but I don't
13:32 ImQ009 joined #minetest-dev
13:32 sapier I can't do 30 different compile variants and tests for each single commit
13:33 thexyz alright
13:33 sapier if we want windows as quality gate for commit we need to define what windows is
13:33 thexyz alright
13:33 thexyz I don't ask you to test everything
13:34 thexyz but when modifying code which will only be used by Windows builds please at least check that some basic functionality works
13:34 thexyz in at least the most common cases
13:34 thexyz (like official build setup)
13:34 sapier I can't even compile with official setup
13:34 thexyz but I guess I'm not the one to decide those things
13:34 thexyz then you should file an issue
13:35 thexyz https://github.com/minetest​/minetest/issues?state=open
13:35 sapier it's a wontfix ... on windows everyone has his own setup which is as fragile noone want to touch anything
13:35 thexyz did you file an issue?
13:35 thexyz if not then please do
13:36 sapier I had msvc 2012 and current mingw running 3 weeks ago ... current master doesn't compile on mingw again
13:36 thexyz did you file an issue?
13:36 thexyz if not then please do it when you have time for it
13:36 sapier my commit for mingw was dropped due to breaking buildbot ...
13:37 sapier and I can't fix buildbot because it's broken by design ... spending 2 weeks for fixing it isn't an option to me
13:37 thexyz did you file an issue?
13:37 sapier I filed a pull request for the mingw thing
13:38 thexyz which fixes the build for you?
13:38 sapier which did by that time
13:38 sapier I'll try if it still works
13:39 thexyz can you give me the link?
13:40 sapier https://github.com/sapier/minetest/commit​/1355a3480e913b8c85aab8819821715659df2100 should be this one
13:40 sapier but as I told I don't know if it's still working
13:41 thexyz I see; I guess only one commit was merged
13:41 thexyz from that pull request
13:41 sapier as far as I know the first one seemed to break buildbot for someone
13:42 sapier no idea why msys buildbot needs to use msvc libs but it seems to do
13:42 thexyz I don't see any messages in the github issue which leads us to another problem with Minetest development process
13:43 sapier true it's burried in irc logs somewhere
13:46 sapier ok that fix works for current master again
13:46 sapier but mingw doesn't hang for me on enter singleplayer world too
13:49 sapier wait ... what I built doesn't seem to be latest master
13:50 thexyz https://github.com/minetest/minetest/issues/1039
13:50 thexyz reported by several people
13:51 sapier I try to get mingw build work again
13:53 sapier but still if we want to prevent things like that to happen in future we need a sane automated win32 build
13:53 Zeitgeist_ joined #minetest-dev
13:54 sapier travis doesn't support windows platforms right?
13:56 celeron55 windows needs a thing called "line" so that happy linux devs wouldn't need to maintain some clunky windows stuff
13:56 celeron55 8)
13:56 celeron55 it would just run regular ubuntu packages or something
13:56 celeron55 how hard can it be!
13:57 sapier :-) virtualbox/vmware in seamless mode?
13:58 thexyz sapier: i think it's possible to setup mingw build in travis
13:58 thexyz but what's the point? it builds fine, it just doesn't work
13:58 sapier imho first step would be make our build bot really build required libs
13:59 sapier there are only two libs that are a little bit more difficult ... openal ... and directx ... guess we don't even have to worry about later one
14:00 sapier another question why do we set windows version defines within minetest code?
14:00 john_minetest joined #minetest-dev
14:00 john_minetest Hello
14:03 thexyz sapier: will you fix it?
14:03 sapier I broke it so I'm gonna fix it ... at least the hang but I won't do anything about win32 buildability
14:04 thexyz alright
14:04 sapier thats one of those issues if you start it everyone requires it to be a full blown 100% perfect golden edge solution ... not realizing current one is pure crap
14:05 sapier anyone to check 1030 while i fix the windows threading?
14:08 john_minetest sapier: sorry, I am stuck on CLI without X. Can't test it
14:09 sapier didn't change anything john but I need at least a second core dev to test and agree to it
14:10 john_minetest oh, okay
14:11 zat joined #minetest-dev
14:12 zat joined #minetest-dev
14:27 iqualfragile joined #minetest-dev
14:30 Warr1024 joined #minetest-dev
14:32 OldCoder joined #minetest-dev
14:53 salamanderrake joined #minetest-dev
15:02 PilzAdam joined #minetest-dev
15:04 ShadowNinja sapier: You should use lua_rawseti instead of lua_pushnumber and lua_setteble.
15:05 sapier what's the difference?
15:05 ShadowNinja PilzAdam: In one world I spawned under a tree's leaves. Is that better than before?
15:05 PilzAdam dunno
15:05 PilzAdam why havent you merged it yet?
15:06 ShadowNinja sapier: It should be faster and involves only one step.
15:06 ShadowNinja PilzAdam: Because I was checking that.
15:09 sapier I'll have a look first I find out why master doesn't work on windows
15:13 sapier argh ... thats a damn stupid bug
15:19 ShadowNinja What about #608? (Or mine/PilzAdam's tweaks of it)
15:19 ShadowNinja (Shaded chat)
15:23 ShadowNinja Should #489 be closed or merged?
15:24 ShadowNinja And is #954 good?
15:25 PilzAdam why do we even bundle json?
15:26 ShadowNinja Because we bundle ALL the things. ;-P
15:27 ShadowNinja Also, I fixed #862. Does that look good?
15:28 PilzAdam Id like to hear what kahrl_ thinks about #862
15:29 sapier "(16:30:15) ShadowNinja: Because we bundle ALL the things. ;-P" why don't we bundle irrlicht too? :-)
15:30 ShadowNinja sapier: That's what I was thinking of asking before. ;-)
15:31 ShadowNinja And maybe we should bundle glibc. ;-D
15:31 PilzAdam also bundle a Linux kernel
15:32 sapier a whole mintetest os? :-)
15:48 celeron55 we bundle jsoncpp and others to make building on any and all library-crippled platforms easy
15:49 celeron55 (lua has the additional reason of API stability)
15:53 bas080 joined #minetest-dev
15:54 Anchakor_ joined #minetest-dev
15:55 sapier does anyone know how windows thread handling is supposed to work?
16:00 Jordach joined #minetest-dev
16:01 sapier ok guess I've got a solution for windows .... have been 3 bugs at once
16:05 SpeedProg joined #minetest-dev
16:05 ShadowNinja celeron55: Wouldn't to be better to provide links to things than to package them?
16:14 sapier https://github.com/minetest/minetest/pull/1040 anyone using windows please test
16:24 ShadowNinja sapier: Should JThread use stdout/stderr?
16:24 sapier jthread can't use errorstream
16:25 sapier I could add an assert in there but I'm not exactly sure about windows threading behaviour
16:25 Gethiox2 joined #minetest-dev
16:26 ShadowNinja sapier: It uses std::cerr/stderr. It should either assert or eturn an error code for the application to handle.
16:26 ShadowNinja return*
16:27 sapier then It'll be an assert but don't complain about windows beeing unstable
16:27 sapier by the way is minetest really supposed to crash on any lua error?
16:30 ShadowNinja sapier: Yes, because if a error occurs you can't grarantee that anything will work. Eg, a function errors out and returns nil, then that nil is serialized and written to disk losing all of the data that the table originally contained.
16:31 sapier crash --> complete terminate of application
16:31 sapier but it's not really any error ... I just realized sometimes an empty dialog box is shown too
16:32 sapier guess that error thingy needs still some work
16:34 thexyz sapier: I'll test that pull in a moment
16:37 ShadowNinja sapier: Also, s/}\nelse {/} else {/ (Unless JThread uses the first style everywhere)
16:43 sapier can someone plz start checking functionality instead of style only?
16:43 sapier some of those bugs in are quite silly it's hard to believe 3 different ppl don't see em if they do a carefull review
16:45 thexyz sapier: it looks like running a singleplayer world and connecting to a server works fine; anything else to test?
16:46 sapier guess that already tests everything involved by that patch
16:47 sapier I'm gonna push the cleaned version
16:49 sapier ok pushing now
17:07 ImQ009 joined #minetest-dev
17:08 Akien joined #minetest-dev
17:11 salamanderrake joined #minetest-dev
17:12 PilzAdam sapier, https://github.com/minetest/minetest/commit​/09a50d0458f46c6129b4bea94502908241b3aed3#d​iff-1305560bd8befb32862f0feeefabd02eL1071
17:13 sapier whats up with that?
17:13 PilzAdam do you notice anything wrong in there?
17:14 VanessaE other than "escapes" should be "escape" and "can not" should be "cannot"?
17:15 kahrl_ ShadowNinja: what did you change to fix #862?
17:15 thexyz no one reads docs anyway
17:15 PilzAdam hint: https://forum.minetest.net/viewtopic.php?id=7936
17:15 sapier I haven't had a look at that one for some time
17:15 thexyz so let's just remove a method every day
17:15 VanessaE ah present tense verb, not noun.  way to fuck up a sentence :)
17:16 sapier Oops :-)
17:16 ShadowNinja kahrl_: I had it bind to 0.0.0.0/:: when connecting to a server instead of binding to the server's address (which wasn't available on the machine).
17:16 kahrl_ I see
17:16 sapier the formspec escape line was lost
17:17 sapier sorry
17:18 ShadowNinja kahrl: So, does it look good now?
17:18 kahrl I'm testing
17:20 kahrl Syscall param socketcall.bind(my_addr.sin6_flowinfo) points to uninitialised byte(s)
17:20 kahrl by 0x6C5513: UDPSocket::Bind(Address) (socket.cpp:360)
17:22 ShadowNinja kahrl: You didn't get that error before?
17:23 kahrl I did, but I'd feel better if it was fixed
17:23 kahrl this happens during TestSocket btw
17:24 ShadowNinja kahrl: What should that be initialized to?
17:24 kahrl 0
17:24 kahrl sin6_scope_id should be initialized to 0 as well
17:26 kahrl I wonder why the memset in Address::Address() doesn't count as initializing
17:27 ShadowNinja kahrl: It only sets the actuall address, not the whole structure.
17:27 kahrl oh right, it calls the other constructor
17:28 kahrl which lacks the memset
17:29 kahrl maybe just copy everything from Address::Address() to every other constructor?
17:29 kahrl actually the memset would be enough
17:31 ShadowNinja Yes, that seems like a good way to do it.
17:34 sfan5 trivial fix for lua_api.txt ( https://forum.minetest.net/viewtopic.php?id=7936 ) incoming in 5 mins
17:36 ShadowNinja sfan5: Already noted. sapier may have a fix ready that will conflict.
17:36 sfan5 really?
17:36 ShadowNinja sfan5: Yep, check the backlog.
17:36 sfan5 when will sapier's fix be merged?
17:37 sapier nope didn't do anything for this
17:37 ShadowNinja Alright, then push it sfan5.
17:37 sfan5 done
17:38 sfan5 the github web editing feature is helpful
17:40 ShadowNinja Yes, but only for editing docs. Other things need to be tested, making it mostly useless. :-|
17:48 sapier ShadowNinja plz create a testsuite for lua errors and run them ... current state is terribly broken
17:49 sapier I see all variants of mixup, crash, no error at all, even wrong error is show sometimes
17:52 sapier and in some rare ocasions it even seems to do what it's intended to do
17:56 ShadowNinja sapier: What do you mean?
17:56 sapier it's NOT working
17:56 ShadowNinja Also, I agree with this https://forum.minetest.net/vie​wtopic.php?pid=120953#p120953 the quote should be changed/removed.
17:57 ShadowNinja sapier: What isn't?
17:57 proller ahaha. ufo is buggiest mod, in third place after mobs and mobf
17:57 VanessaE awww, I like the "random bunch of lunatics" :(
17:58 sapier e.g. if I cause an error in on_generated I end up in a follow up assert() done in on_step ... I don't even see the original error
17:58 sapier most time minetest does exit at once instead of showing the error dialog
17:59 sapier and sometimes I see an empty error dialog
17:59 sapier except of that one assert problem in all cases the backtrace was shown on console
18:01 bas080 joined #minetest-dev
18:02 ShadowNinja VanessaE: That can stay, but the quote should change.
18:03 ShadowNinja sapier: You cause a error to happen that doesn't show up until on_step?
18:03 ShadowNinja (Eg, minetest.get_connected_players = nil in on_generated)
18:04 sapier I cause an error within on_generated that one doesn't show up but instead another error is caused which is direct result of first one
18:04 ShadowNinja Well, I can't really do anything about that...
18:05 sapier so you tell this was wrong prior your fixes too?
18:06 VanessaE ShadowNinja: oh right, the quote, derp.  because of the 'shitload' remark?
18:08 ShadowNinja VanessaE: Yes.
18:08 VanessaE ShadowNinja: so redact it?  "with a [ton] of fuel" or some such?
18:09 ShadowNinja VanessaE: Well, if we can't find a better quote.
18:09 VanessaE ShadowNinja: wait...
18:10 VanessaE hm, no, that won't work.
18:10 sapier ok at least 0.4.6 is broken too .. different ... but still broken ... yet ShadowNinja "I can't really do anything about that..." isn't a usefull answer ... I count that as "I don't want to do anything about it"
18:11 sapier I wonder how anyone wants to use luajit for mod development :-)
18:11 VanessaE [12-04 22:57] <VanessaE> https://forum.minetest.net/vie​wtopic.php?pid=120769#p120769
18:11 VanessaE [12-04 23:02] <NakedFury> you win
18:11 VanessaE [12-04 23:02] <NakedFury> we wont ever find anything cooler as a product of a glitch or bug
18:12 VanessaE I think there's something useful there :)
18:13 ShadowNinja VanessaE: I thought of that, but it involves a screenshot, we don't wand to have users follow a link.
18:13 VanessaE put a thumb on the page next to the quote?
18:13 VanessaE if you want
18:14 ShadowNinja sapier: Fix your mod. I don't know of any way that I could fix that. You shouldn't be breaking on_step.
18:15 ShadowNinja VanessaE: Of course we don't want to advertize map corruption...
18:15 john_minetest joined #minetest-dev
18:15 VanessaE ok :)
18:15 john_minetest oops, forgot to rejoin here.
18:16 naxthesurvivor joined #minetest-dev
18:23 john_minetest joined #minetest-dev
18:24 Zeitgeist_ joined #minetest-dev
18:25 sapier ShadowNinja if you don't think lua errors should be shown in a usefull way why did you even work at it?
18:27 ShadowNinja sapier: Um, I can't show that kind of error. I've never even heard of anything providing a error message specificaly indicating that.
18:28 sapier I just want the first error to be shown immediatly
18:29 john_minetest sapier: Ping me when you need testing
18:29 sapier no critical pending issues fixed by me john atm ... the modstore fix is waiting for coredev agreement
18:30 ShadowNinja sapier: You want something like this? http://pastebin.ubuntu.com/6531003/
18:31 sapier yes that seems to be quite similar to what I did to cause it
18:31 sapier wait it wasn't global onstep but entity onstep
18:32 ShadowNinja ... Does anyone else think this is rediculous?
18:32 Calinou joined #minetest-dev
18:33 Gethiox2 joined #minetest-dev
18:33 sapier on it isn't once you limit cpu time per step you need to do things like that
18:34 sapier and data messed up on_generate may cause any error if it doesn't stop on_step from continuing
18:35 bas080 joined #minetest-dev
18:35 sapier shadow do you do some lua mods too?
18:36 ShadowNinja sapier: Yes, quite a few.
18:36 ShadowNinja kahrl: Is bind_address good now?
18:36 sapier :-) wonder why you didn't run into those issues yet ;-)
18:37 ShadowNinja sapier: I don't understand how this could possibly be a issue with minetest.
18:37 sapier not showing a crash error but showing another one is an issue
18:38 sapier I don't know why the original error is lost but there's not a single sign of it within traces
18:39 ShadowNinja sapier: There is NO ERROR in the code that sets crash to true, or nils out a needed value. It is unly a error when someone tries to use the niled out value.
18:42 sapier https://gist.github.com/sapier/7830051
18:44 sapier ok it's still to simplified I guess I need to post the real functions
18:45 sapier ok now there's the code
18:46 sapier quota_enter and quota_leave do all the calculations ... hiding it from caller
18:46 sapier therefore I need to ensure they're called in correct sequence ... that's what the assert is for
19:07 werwerwer joined #minetest-dev
19:08 us}0gb joined #minetest-dev
19:16 ShadowNinja How much of this is still applicable? http://dev.minetest.net/TODO
19:17 Gethiox3 joined #minetest-dev
19:21 kaeza most (all?) of the items in formspec main menu could be removed
19:22 OldCoder joined #minetest-dev
19:22 kaeza +section
19:22 kaeza most of the other items are still valid I think
19:23 kaeza "Make gold not look like American cheese " :I
19:24 ShadowNinja sapier: Is all of that fixed?
19:27 proller joined #minetest-dev
19:29 sapier 1-3 are fixed ... I don't even know what 4 is about
19:48 kaeza what is the general opinion on allowing FOV change from Lua?
19:49 sapier I don't have an opinion to this ;-)
19:49 emptty joined #minetest-dev
19:50 kaeza I have a simple implementation which allows to set FOV to both absolute (set to 40°, 50°, etc) and relative (set to 120%, 75%, etc) values ready
19:52 kaeza may need some cleanup, and a rebase https://github.com/kaeza/mt-firearms/blob/m​aster/firearmslib/util/player_setfov.patch
19:55 kaeza that mod also comes with a (bit over-engineered) demonstration (a sniper rifle) https://www.youtube.com/watch?v=CPRou-DFqgI
20:01 PilzAdam why are src/jthread/win32/jthread.cpp and jsemaphore.cpp marked as executable?
20:01 RealBadAngel_ joined #minetest-dev
20:09 Akien joined #minetest-dev
20:09 sapier maybe error on copying them
20:24 SpeedProg joined #minetest-dev
20:53 Calinou joined #minetest-dev
21:04 sapier1 joined #minetest-dev
21:07 emptty Do you guys also have that bug where the furnaces are blinking?
21:07 emptty It prevents me from smelting stuff properly
21:10 Calinou what fuel do you use?
21:10 emptty leaves
21:11 emptty but actually, it's currently empty (out of fuel)
21:11 ShadowNinja The furnace should use minetest.swap_node.
21:19 Calinou use another fuel that lasts longer?
21:24 emptty Actually, the furnace is empty
21:24 emptty and I cannot fill it because it's blinking like hell
21:24 emptty I've built another one, and that's the same problem
21:26 emptty I managed to fill it with coal, and it's still blinking like hell
21:31 ShadowNinja \o/ tcp_connection rebased against master works.
21:32 sapier1 tcp_connection?
21:34 ShadowNinja sapier1: A branch that adds TCP support to Minetest instead of UDP RELIABLEs.
21:36 ShadowNinja The connection is noticable faster, and old clients can still connect.
21:36 ShadowNinja +y
21:37 sapier1 guess it's a good idea to make protocol do things as we already send everything reliable
21:39 iqualfragile sapier1: wut? everything is send reliable? thats stupid
21:39 sapier1 do you know any packet that isn't sent reliable?
21:39 iqualfragile no i do not, but i know that a lot of packets could be sent unreliable
21:40 sapier1 I admit I haven't done much in that part of code but I've never seen anything else then reliable packets
21:40 iqualfragile that should get fixed
21:41 iqualfragile hmm… my server seems to have problems connecting to the masterserver
21:42 iqualfragile 22:15:54: ERROR[ServerThread]: Serverlist at url servers.minetest.net error (Timeout was reached)
21:42 sapier1 wow :-) advanced spawning prototype causes average load of 0,5% ... with worst lag added beeing 12ms
21:42 iqualfragile x100
21:42 iqualfragile sounds great
21:43 PilzAdam simplemobs spawning has an average runtime of 0ms +-10 per spawn
21:44 ShadowNinja PilzAdam: -10ms?
21:44 sapier if you use os.clock that's nothing to be proud of < 10ms isn't counted at all
21:44 PilzAdam sapier, thats why the "+-10" is there
21:45 sapier that's quite useless any single spawn of adv_spawning is less then 10ms too yet I don't do singular spawns but count all time consumed by spawners in total
21:45 PilzAdam how do you benchmark it?
21:46 sapier lua socket library
21:46 sapier calculating time between enter function and leave cumulating all parts and evaluating in global on_step
21:47 PilzAdam sapier, add it here please: http://dev.minetest.net/index.php?title=Lua​_Optimization_Tips&amp;rcid=1728#Profiling
21:47 PilzAdam eh, -&rcid=1728
21:48 sapier I guess striping the quota part and make a mod from it is even more easy
21:49 sapier current mobf already uses this mechanism to reduce typical load to < 50% even while map is generating and hundreds of mobs are fighting at same time
21:50 sapier but it's not used everywhere in mobf ... to much to change at once
21:53 sapier but first thing is finishing the advanced spawning
21:54 iqualfragile idea: use framed glass as default glass
21:56 sapier ok we need a way to find out if a position is active or not
21:57 PilzAdam get_node_or_nil()?
21:57 sapier ist silly to spawn in inactive areas where entitys get deactivated at once
21:57 Jordach iqualfragile, i get strange bugs with it
21:57 sapier active not loaded
21:58 PilzAdam calculate the distance to players and compare to the minetest.conf setting
21:59 iqualfragile Jordach: which ones?
22:00 Jordach vanessa's survival server uses them for technic, they have merged faces, textures cannot be seen
22:00 sapier hmm could work pa but doesn't really scale for servers with lots of players
22:28 djdduty joined #minetest-dev
23:50 john_minetest cya
23:50 john_minetest left #minetest-dev

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