Time Nick Message 00:00 VanessaE I'm just suggesting something similar to this, http://pastebin.ubuntu.com/6747702/ 00:01 ShadowNinja No, we're preventing them from redefining the internal name, which would really make a new, almost identical node. And override the old one too. 00:02 VanessaE hm, ok 00:02 ShadowNinja The (name, redefinition) syntax is shorter and more consistent. 00:12 ShadowNinja Made redefinition of a non-existent node non-fatal and added doc: http://ix.io/9QW Comments? 00:13 VanessaE ok, I can get behind that. 00:27 kahrl ShadowNinja: I think it should be an error to redefine a non-existent item 00:28 kahrl instead of silently ignoring the call 00:30 ShadowNinja kahrl: It makes it easier to only redefine nodes if they exist, o you can support other mods without wrapping everything in "if minetest.registered_items[name] then ... end". 00:31 ShadowNinja so* 00:32 ShadowNinja And it's not nice when someone else removes the node from their mod and your mod fails. 00:33 kahrl but if you override an item, you expect that your changes have some kind of effect 00:34 kahrl what's wrong with the call in an if? 00:34 kahrl with wrapping* 00:36 kahrl by default it should fail fast, e.g. to let the mod programmer know if they have a typo 00:49 VanessaE bbl 03:17 iqualfragile overide will be quite a nice thing 04:00 ShadowNinja Thoughts on #1096? Seems like it should be done in a mod(mesecons), but maybe this would be usefull. 04:00 ShadowBot https://github.com/minetest/minetest/issues/1096 -- Add circuit simulator. by Selat 04:08 hmmmm well I think the point is to make circuits more efficient 04:09 hmmmm from what I've seen so far the code itself looks good but the code style is completely his own thing, at the very least I would hope that the spaces would get changed to tabs, but we could do that 04:09 hmmmm if we were to merge that we'd have to have one of us take a look at what it actually does, verify that it works the correct way, do some QA on the code itself, a line-by-line peer review would be good enough I suppose 07:40 celeron55 i'm surprised how well that is done 07:40 celeron55 but it's still controversial 07:42 celeron55 it adds std::list ::iterator to every MapNode 07:43 celeron55 i'm sure that contains at least one pointer or even more, immediately doubling or even tripling the size of each MapNode, independently of whether the whole map uses the functionality at all 07:44 celeron55 it's an utter waste of memory 07:56 VanessaE celeron55: oh btw... I forgot if I mentioned, I tried out those ultra-aggressive settings you gave a while back, http://paste.ubuntu.com/6698208/ ... big fat barrel of NOPE :-) Lag city. The last two lines that set the emergequeue limits are good, though. 07:57 celeron55 why is everyone testing those; i already said they don't work 07:57 celeron55 try something like 16 instead of 100 07:57 VanessaE at present, I'm at the defaults on all of those, save for the last two. seems to work fine that way 07:58 celeron55 also 20 is probably too much for a server with multiple users (it won't make anything slower but the server will use a lot of bandwidth and disk) 07:58 celeron55 VanessaE: by default it's 4 07:58 VanessaE whats the size of a mapblock again? 07:58 celeron55 size? 07:58 VanessaE some 128k isn't it? 07:59 celeron55 what size 07:59 celeron55 volume? bits? compressed or not? weight? 07:59 VanessaE the data count/size in bytes of a chunk or block as measured in that particular field. 07:59 celeron55 on disk or in memory, and including or not including meshes on the client? 07:59 VanessaE (I presume it's a mapblock = 16x16x16 nodes) 08:00 VanessaE as sent over the network. 08:00 celeron55 i guess they average at some hundreds of bytes 08:00 VanessaE oh ok, not sure where I got 128k then. 08:02 celeron55 i don't think anything related to minetest is 128k 08:02 VanessaE eh, disregard that number. I don't know where that even came from 08:05 VanessaE I'm just trying to figure in my head why even 100 was too much - the server's dedicated and has a full 100 Mbps pipe, so even if "hundreds of bytes" could equate to 1kB, it would take ~100 clients to flood the network out. 08:05 VanessaE but it only took like 5 or 6 clients to make it lag out with that setting. 08:06 VanessaE (and the network bandwidth was barely a blip) 08:22 celeron55 VanessaE: it's generally limited by the client's mesh generation speed 08:22 VanessaE ah 08:22 celeron55 so with a fast connection you effectively have 100 meshes waiting to be generated, which generally results in many seconds of meshgen lag 08:22 VanessaE users were reporting 2-3 minutes' worth of lag at the time. 08:23 celeron55 (not with my new laptop though...) 08:23 VanessaE that would tend to explain it :) 08:23 VanessaE I was flooding the poor sods out :D 08:23 celeron55 it probably flooded the connection in some way too; dunno how exactly 08:23 celeron55 at least if they were using old clients 08:24 VanessaE could be. from the server's end, it was barely a blip. 08:24 VanessaE http://digitalaudioconcepts.com/vanessa/hobbies/minetest/stats.html 08:25 VanessaE round about 18:00 on that chart 08:25 VanessaE only had those settings in place for perhaps 10 minutes or so 08:27 VanessaE (I should adjust the CPU collection code to also report *only* that used by the minetest servers in addition to the total) 10:49 Megaf hi all 15:21 BlockMen umm..guys? current master (engine and game) doesnt render blocks anymore (win7 64bit opengl 2.1.0) 15:22 rubenwardy lol 15:23 celeron55 oh that's such a minor bug 15:23 celeron55 just live with it 15:24 BlockMen yeah, indeed. i have now a perfect drawtime :P 15:24 BrandonReese probably broke with ShadowNinja's commit "Fix Spelling of 'attempt'" if he didn't spell attempt right there is no telling what else he broke 15:24 BlockMen celeron55, can i push this meanwhile? https://github.com/BlockMen/minetest/commit/84bbe1ec6e5f76ab8b9e11ebf247f82dcad17330 15:25 celeron55 BlockMen: anyway; you should try to bisect a commit which broke it for you 15:25 VanessaE BlockMen: actually, I just pulled from git about 20 minutes ago, at least the server side is fine. 15:25 BlockMen celeron55, will do 15:26 VanessaE (I'm at 5434d87d on the server, anyway) 15:27 VanessaE currently at a4c5f10e on the client side 15:27 celeron55 BlockMen: i guess you can push it, but kahrl is the one who has most authority on it according to the subsystems 15:28 celeron55 or maybe thexyz; whatever 15:29 rubenwardy late SOPA protest? 15:29 BrandonReese I just pulled from git and recompiled and everything seems to be in working order. Kubuntu OpenGL 15:30 BlockMen celeron55, i thought since its just a MSVC specific thing and just a header it wouldnt be that subsystem special 15:30 BlockMen but if i should wait for the maintainer im fine too 15:31 xyz sure why not 15:31 xyz it can't make things worse, so 15:31 xyz any reason for not using cassert though? 15:32 xyz BlockMen: ^ 15:34 BlockMen xyz, whats the difference between cassert and assert? 15:34 VanessaE side topic: one of my players just got another hit on issue #1101 15:35 ShadowBot https://github.com/minetest/minetest/issues/1101 -- 0.4.9-36-g574ed6a Minetest Client /mainmenu.lua:562: in function handle_multiplayer_buttons 15:44 xyz okay, I read about it and figured out I was wrong 15:44 xyz anyway, use whatever works for you 15:45 xyz the difference is what namespace functions will be included to 15:45 xyz but usually you'll get both I guess 15:45 xyz global and std namespaces used 15:49 BlockMen xyz, ic, thx 15:49 BlockMen so whos part is it now, yours or kahrls (since c55 didnt know either)? 16:00 xyz who cares, just push it 16:00 xyz change is too simple 16:04 BlockMen done 16:07 BlockMen and to the rendering problem: 10fdbf7375 works fine while e21b29f422 does not. so it must be one of the 3 commits between i guess 16:08 BlockMen but im afk for now 16:08 Tesseract git bisect start && git bisect good 10fdbf7375 && git bisect bad e21b29f422 16:11 xyz I think it's 8b0b857eaaa50c6ec217a46c0577395c78ec04c7 16:11 xyz I mean, sapier as usual wrote some stuff for windows without even compiling it 16:12 celeron55 seems very likely 17:21 BlockMen cant say whether its 8b0b857eaa or 9edb91da57. with 8b0b857eaa i cant connect to server. 9edb91da57 fixes that but which one causes the rendering problem is not clear. 17:24 xyz wtf 17:24 xyz https://github.com/minetest/minetest/blob/master/src/gettext.cpp#L27 17:25 xyz apparently this works because MSVC just ignores the part starting with "and" 17:25 xyz but something tells me you're supposed to &&, not and 17:30 celeron55 how the hell do these end up in the codebase 17:30 celeron55 is there a compiler option for disabling those keywords? 17:31 celeron55 maybe -fno-operator-names 17:31 celeron55 (for gcc) 17:33 xyz -O2 -Wall -Werror -Wformat-security -Wignored-qualifiers -Winit-self -Wswitch-default -Wfloat-equal -Wpointer-arith -Wtype-limits -Wempty-body -Wno-logical-op -Wstrict-prototypes -Wold-style-declaration -Wold-style-definition -Wmissing-parameter-type -Wmissing-field-initializers -Wnested-externs -Wno-pointer-sign -std=gnu99 17:33 celeron55 ? 17:34 xyz this could improve quality of the code greatly 17:35 xyz (although I copied this from C course page) 17:35 celeron55 i think we shouldn't use the gnu C++ "standards" if at all possible 17:35 celeron55 so -std=something-actually-standard 17:36 celeron55 actually what even are the options 17:37 celeron55 i guess it would be c++03 then until we drop msvc2010, in which case c++11 (or c++0x (understood by a bit earlier compilers)) 17:39 celeron55 (but the and/or/whatever keywords actually are part of the standard, so this doesn't fix that) 17:57 sapier hmm that "and" is in there for quite some time ;-) ... of course it's wrong ... guess I did a lot of lua programming by that time 17:59 sapier btw I don't do a freebsd build too ;-) 18:00 BlockMen sapier, http://irc.minetest.ru/minetest-dev/2014-01-14#i_3547116 and http://irc.minetest.ru/minetest-dev/2014-01-14#i_3547238 18:01 sapier last one seems to be fixed isn't it? 18:01 sapier first one is windows only too? 18:02 xyz but you don't write freebsd-only code as well 18:03 BlockMen how should i know? i have win only. and no, i wrote current master, so it isnt fixed 18:03 sapier who can tell freebsd is working same as linux? 18:03 xyz and we know that it's more-or-less compatible 18:03 xyz unlike windows 18:03 xyz my point is not that you should test everything everywhere when you commit code 18:03 sapier remembering the threading issues proller had on jthread switch there are major differences 18:04 xyz this is impossible and would be quite stupid to ask 18:04 xyz my point is that when you modify platform-specific code better make sure that it's correct 18:04 sapier I asked multiple times about those fixes to be tested 18:04 xyz so we don't get shit like ";e" 18:04 xyz which doesn't even compile 18:05 sapier as we don't have a usefull working test platform for windows I consider this os as low priority, if we manage to get a sane windows environment not relying on various different binarys and non available compilers I can test windows too 18:06 xyz so you think it's okay to write code without even compiling it? 18:06 sapier I told this multiple times and have no reason to change my opinion by now 18:06 sapier if you're that upset about it I'm gonna not add any windows version in future 18:07 xyz great attitude we have there 18:07 VanessaE kahrl: regarding issue 1011, 8966c16a is the cause. 18:07 xyz "i can't agree with you so just fuck you, right" 18:07 VanessaE er, 1101 18:07 xyz and you didn't answer the question 18:07 sapier no in a perfect world I'd fully agree with you ... but I don't have time to test 200 different setups 18:08 sapier and as long as there ain't a reference platform to check against I wont waste time testing "some" platform 18:08 xyz > so you think it's okay to write code without even compiling it? 18:08 xyz oh, that's fine 18:08 xyz let's say our platform is MSVC 2010 on Windows 18:09 xyz mingw is not officially supported 18:09 xyz objections? 18:09 sapier ok if you define it this way I wont provide any windows code in future as I can't get that platform work 18:09 xyz but that means we won't have compatibility with windows 18:09 xyz which is too unacceptable 18:10 PilzAdam we cant expect sapier to test his pull requests on all platforms 18:10 sapier ok then I suggest those people using windows on regular base test code once they're asked 18:10 PilzAdam thats why we call master "unstable" 18:10 xyz PilzAdam: that's not what I mean, please try reading my point 18:11 PilzAdam sorry 18:11 sapier xyz I fully agree with you that this would be great but It's just not possible to test any single change on all platforms 18:11 sapier it's more then doubling development time 18:12 xyz again, my point — writing platform-specific code without testing it (not even compiling, I'm not talking about testing every use case) is unacceptable 18:12 xyz sapier: as I said already, I'm not talking about "every single change" 18:13 sapier ok the quick solution is I won't write platform specific code anymore and just leave windows code to someone who knows what he's doing ... result will be same windows doesn't build 18:13 PilzAdam writing broken platform-specific code is better than writing nothing 18:14 xyz if you don't write platform specific (windows) code then your code won't work on windows and therefore won't be merged and as the result we get working (somehow) windows builds 18:14 xyz I don't really know what to do about it so I'll just ask celeron55 to solve this MAINTAINER CONFLICT 18:15 sapier I wont discuss this any longer, I don't have time to check any code for all platforms, so if you can't live with some shorttimes broken code you'll have to live with no code at all, those are the options to decide 18:16 sapier and btw if you think it's tooo difficult to add a simple missing include we've got more problems 18:17 xyz you again are saying "any code for all platforms" which is not what I mean 18:17 xyz maybe you should try reading what I say or something 18:18 xyz whoa, and how did adding a single include solve BlockMen's problem? 18:18 sapier It didn't so compiling wouldn't have shown the real problem at all 18:19 sapier And chances to check all minetest features at all platforms are even less ;-) 18:20 xyz yes, the word about "compiling" was to show you how extremely wrong this is 18:21 xyz if you write code which doesn't even compile then what else can we talk about 18:21 sapier it's quite simple that pull request was there for 3 weeks to test, if noone else cared to test it it's just not important 18:21 xyz ah so it's okay to merge something untested if it's open for too long, I see 18:22 sapier if it's only a platform issue while main platform works yes 18:22 xyz so what's the main platform? 18:23 xyz I can't find any definition of that on the wiki 18:23 sapier os/2 warp for sure 18:24 xyz so..? 18:24 sapier btw httpfetch doesn't build on mingw too 18:24 xyz I really want to know what platform do you define as "main" and why 18:25 sapier obviously linux is main platform for developers if I have to tell you about this you must have missed something 18:25 xyz yes, now you added "for developers", I guess I missed this the previous time you said something about main platforms 18:26 sapier development branch is ... according to its name primary for developers 18:27 xyz well, fine 18:27 xyz let's say you win this time 18:27 sapier and while I fully understand your point, no matter how long you complain I still wont have time to test everything at any platform 18:28 sapier of course I try not to break window 18:28 sapier s 18:28 sapier and if it happens I will fix it 18:28 xyz I've never said you should test everything everywhere 18:28 xyz my point was that if you write platform-specific code then please make sure it works 18:28 xyz got it now? 18:29 xyz that doesn't mean you should do it, that's why it's called "my point" 18:29 sapier if I have time I do so but that wont be always case ... especially if master is broken on that platform anyway 18:30 xyz then how do we solve this issue? 18:30 sapier we need more regular windows testers ... or define one open available clean windows test environment 18:31 xyz open and windows aren't friends 18:31 sapier I can't build current master neither on mingw nor on msvc2012 18:32 xyz moreover, this won't help us because official builds are done with VS 18:32 sapier we are the ones doing the official builds 18:32 xyz I'm the one, to be precise 18:32 celeron55 i figure an automated windows build, even with mingw, would help 18:32 xyz now what? 18:32 sapier as far as I know msvc2010 was used because of directx not beeing available in mingw by that time is that correct? 18:33 xyz "by that time"? 18:33 xyz do you mean it's available now? 18:33 sapier of course 18:33 sapier I managed to build all of minetests depencencys except of openal and directx itself, but I was able to link direct3d support with mingw 18:34 sapier about 3 months ago 18:34 sapier right now neither mingw nor msvc2012 work 18:34 BlockMen i doubt that this issue is msvc only 18:34 BlockMen and i can build current master on msvc 18:34 sapier what version of msvc? 18:34 xyz well, msvc is believed to have better optimizations 18:35 BlockMen 2010 18:35 xyz I really don't know, if everybody's okay with supporting only MinGW then why not 18:35 xyz but you'll have to find someone to make builds 18:35 * BlockMen isnt 18:35 xyz BlockMen: why not? 18:36 BlockMen see it positive. i you can get it run under msvc it runs everywhere 18:36 sapier ok so we have right now msvc2010 builds but doesn't work ... mingw doesn't link msvc2012 doesn't link 18:36 xyz BlockMen: no, that's not true 18:36 xyz moreover, MSVC is quite "different" in some aspects 18:36 sapier obviously msvc2010 doesn't even ensure it's buildable with 2012 ;-) 18:36 BlockMen :( 18:36 xyz sapier: why didn't you build openal? 18:37 sapier because I didn't find buildable source code 18:37 celeron55 http://repo.or.cz/w/openal-soft.git 18:37 celeron55 that probably is what is used as the open source one 18:38 xyz http://kcat.strangesoft.net/openal.html 18:38 xyz yup 18:38 BlockMen but shouldnt we just drop msvc when we find a problem that seems to be msvc specific and not easily solvable? 18:38 celeron55 that probably builds ridiculously easily on anything 18:39 sapier hope that's not same one I already tried without succes but I'll give it a try 18:39 celeron55 i think it's what linux distros package 18:39 xyz BlockMen: well, as I see the problem, it's that people write code for windows without testing it on windows 18:39 sapier there's a even more severe problem with msvc ... gettext 18:40 xyz what about it? it works fine 18:40 sapier it's not supported as of 2012 for licensing issues 18:40 sapier and never will be 18:40 xyz and the problem is..? 18:40 celeron55 i guess that probably means we drop msvc support completely when we drop msvc2010 support 18:41 celeron55 or alternatively switch from gettext to something else 18:41 sapier locale handling code supplied with msvc2012 is incompatible to gpl and is required to be linked to lib 18:41 sapier license incompatible 18:42 sapier that's why there's no gettext dll for msvc2012 out there 18:42 xyz sapier: can you elaborate? 18:43 celeron55 i neither don't really get how it can be more incompatible than it was before 18:43 sapier you'd have to ask a lawyer for details but that's what you find on looking for a gettext dll for msvc2012 18:44 sapier some change in end user license of msvc2012 18:44 celeron55 why do these cancers have to exist 18:45 celeron55 both pretend they own the world and then make the life of everyone harder than it should be 18:45 xyz sapier: can you provide a link or something? I tried to google "msvc2010 gettext dll" and nothing relevant comes up 18:46 sapier it's quite some time since I did look for that dll I haven't saved links 18:46 sapier I guess looking for gettext msvc 2012 official could be a good starting point to find it 18:47 xyz celeron55: they're very autistic; they call "win32" "woe32" because it's no "win" 18:48 xyz > Microsoft uses the term “Win32” to denote these; this is a psychological trick in order to make everyone believe that these OSes are a “win” for the user. However, for most users and developers, they are a source of woes, which is why I call them “Woe32”. 18:48 xyz anyway, getting slightly off-topic here 18:48 celeron55 oh god 18:50 xyz sapier: gettext runtime library is under LGPL, gettext tools (which we don't use) are under GPL; I can't see the problem here 18:51 sapier I only tell what those ppl providing the old versions write 18:51 sapier if you find a msvc2012 dll I'd be glad, I didn't manage to find one 18:53 celeron55 hmm, isn't LGPL really like GPL at the dependency side of things (i mean, when it depends on something) 18:53 xyz sapier: compile it yourself? 18:54 xyz ah, found something related http://lists.gnu.org/archive/html/bug-gnu-utils/2004-01/msg00049.html 18:55 sapier ok that seems to be the issue I remember 18:56 celeron55 so it's all about interpretation of the license 18:56 sapier it's allways about interpretation :-) 18:58 xyz http://www.gnu.org/licenses/old-licenses/gpl-2.0-faq.html#WindowsRuntimeAndGPL 18:58 xyz even if it were under GPL I still don't see the problem 18:59 sapier ok then someone should build a gettext dll and provide it, I found noone doing this 18:59 sapier if you're sure about it you can do it, I'm not willing to take the risk 19:00 xyz why can't you do that? 19:00 xyz what's the "risk" you're talking about? 19:01 BlockMen just to peek in, do we need that at all? is there a good reason to switch to msvc2012? 19:03 ShadowNinja I'm getting those map errors that VanessaE got... http://pastebin.ubuntu.com/6751991/ 19:03 VanessaE oh my 19:04 sapier ok any idea particles are there and have texture but nodes don't have 19:05 xyz no, that situation is too fucked up 19:05 celeron55 BlockMen: none at the moment, except maybe c++11 19:05 ShadowNinja And when teleporting there: http://pastebin.ubuntu.com/6752008/ 19:05 xyz c++11 would be quite nice to have, at least some parts of it 19:05 celeron55 BlockMen: for some reason sapier insists on using it 19:05 VanessaE ShadowNinja: I'm curious - if you ignore world load errors, what does the resultant block look like when it's loaded? 19:05 VanessaE is it gibberish, or empty? or something else? 19:06 ShadowNinja It will probably be regenerated. 19:07 celeron55 i once found some very small i18n library that supported loading non-compiled gettext language files 19:07 celeron55 but i can't find it now; i've tried for over 20 minutes already 19:07 xyz well, there seems to be no problem with gettext 19:07 xyz it's under lgpl, so everything is well 19:07 celeron55 no real problems, yes 19:07 xyz the problem is with GPL'd software 19:08 celeron55 i like to dig up all the alternatives for reference though 19:08 xyz what if you link it to MSVC runtime and then want to distribute, are you allowed to distribute runtime .dll with your software or not? 19:08 BlockMen celeron55, i would like to keep 2010 as long as possible since a have a full licence for ultimate 19:09 xyz lol 19:09 celeron55 BlockMen: that's a stupid reason 19:09 * BlockMen knows that :P 19:09 xyz did you buy it? 19:09 celeron55 just download 2012 illegally; if someone asks whether you're allowed to do that, tell them i gave you permission 19:09 BlockMen nope, from my university 19:10 xyz then you can have 2012 too, can't you? 19:10 BlockMen celeron55, i bet that will work 19:10 BlockMen xyz, no, only 2010 19:10 ShadowNinja VanessaE: Nope, it does this: http://pastebin.ubuntu.com/6752024/ 19:10 celeron55 well actually, i probably wouldn't do that myself 19:10 celeron55 the MSVC builds that i have made of minetest have been all made with the express edition 19:10 VanessaE ShadowNinja: eep! 19:10 VanessaE ShadowNinja: what's the actual world look like in that area? 19:11 VanessaE (I mean, a screenshot thereof) 19:11 celeron55 BlockMen: tell them it's a bad university if they don't allow you to learn C++11 19:11 ShadowNinja VanessaE: Nothing loads. 19:11 VanessaE ShadowNinja: hm, interesting. 19:11 xyz dreamspark usually gives access to latest versions of microsoft software 19:12 VanessaE guess I should have expected that. 19:12 BlockMen celeron55, as they would care 19:12 sapier I use msvc2012 because my windows build box isn't big enough for mingw msvc2012 AND msvc2010 ... and I see no reason to waste another half a week to get msvc2010 build to work 19:12 sapier well mingwbuild is broken atm too 19:13 BlockMen well, whatever. my version is not valid reason for switching or not, so... 19:13 sapier C:\MinGW\msys\1.0\build\minetest\src\httpfetch.cpp:29:25: fatal error: sys/utsname.h: No such file or directory 19:13 ShadowNinja kahrl: ^ 19:14 kahrl well, that was apparently proller 19:14 kahrl just change it to #ifndef _WIN32 or something 19:15 xyz we still haven't decided on "official" windows build env 19:15 * BlockMen prefers MSVC2010 19:15 xyz I think we should do that 19:16 sapier if we do so this environment should be able to build all dependencys too 19:16 sapier at least all non os or compiler suite contained dependencys 19:16 xyz of course 19:17 sapier msvc2012/2010? can do almost all too maybe gettext is a problem but if you find a way to build it to it'd be fine too 19:17 xyz well I can give it a try 19:19 sapier still I'd prefere mingw because for mingw we have ide as well as debugger available without registration 19:20 xyz any other opinions? 19:21 sapier we could define two windows envs too, at least if we manage to get both in a stable state 19:21 ShadowNinja kahrl: Can you make that change? sapier: Any other errors? 19:21 Calinou MSVC is proprietary, requires EULA + registration 19:22 sapier yea serialization.cpp is still broken ... but that's due to someone insisting on using msvc zlib with mingw build 19:23 sapier those are two changes in httpfetch 19:23 xyz sapier: we can't get one now 19:23 xyz Calinou: windows too 19:23 sapier maybe two are more easy then one? 19:23 xyz yeah suure 19:26 kahrl ShadowNinja: I think it's a build issue so that's for xyz to commit 19:26 xyz hey, I'm dealing with it right now by determining what builds envs we support! 19:26 kahrl by the way, the way minetest builds the useragent does not conform to RFC 2616 19:28 kahrl Calinou: I didn't have to register to use the express version... 19:30 sapier they changed for all newer versions kahrl 19:31 sapier if you don't register you can use gui only for 30? days 19:31 celeron55 fucking finally! 19:31 celeron55 https://github.com/j-jorge/libintl-lite 19:31 celeron55 i found it 19:31 kahrl sapier: well I didn't need the gui 19:32 kahrl of course there's still the eula 19:32 Calinou if we're talking about standards, there shouldn't be .minetest in homedir 19:32 celeron55 libintl-lite looks very approachable if it ever turns out gettext isn't worth the gnufagging 19:32 Calinou but ~/.config/minetest :P 19:34 celeron55 Calinou: that has been discussed 19:35 kahrl actually there has been a response on one of the issues about that 19:35 kahrl https://github.com/minetest/minetest/issues/864 19:35 Calinou is the answer "fuck the standards" as usual? 19:35 Calinou :/ 19:35 Calinou I fear such a behaviour from you guys 19:36 kahrl Calinou: how is "." not a standard? 19:37 Calinou it's bad practice 19:37 Calinou and discouraged by FreeDesktop as far as I remember 19:37 Calinou it clutters your home dir 19:37 xyz sapier: can you explain why do we need to compile every dependency from source? 19:38 Calinou more and more programs have been using ~/.config/ 19:38 kahrl freedesktop encourages bad practices such as installing systemd and pulseaudio too 19:38 Calinou ... 19:39 sapier because binaries usually don't match your environment 19:39 celeron55 but everyone loves poettering, right? 19:40 sapier e.g. msvc libs for mingw ... whoever did this just wanted to have "something" working 19:40 xyz sapier: ah, I see; but if we're talking about choosing one single environment 19:40 celeron55 i did it o/ and it always has worked 19:41 sapier yes because you had didn't have to get gettest freetype ... working by that time ;-) 19:42 celeron55 oh yeah, it doesn't in that case 8) 19:44 sapier debugging is quite difficult if you can't get debug symbols to your binary :-/ 20:38 ShadowNinja kahrl: Better? http://ix.io/9Ry 20:43 kahrl ShadowNinja: yeah 20:53 sapier ok I think I found blockmens error 20:53 sapier can someone explain to me why all semaphore implementations have some feature missing? 20:53 sapier posix is missing relative sleep time 20:53 sapier windows is missing read function for current semaphore count 20:54 kahrl sapier: what purpose would the read function serve? 20:55 sapier it's for implementing a broken minetest implementation ;-) 20:55 sapier mintest first checks if a queue is empty 20:55 sapier I used semaphore count in new mutex queue implementation 20:56 kahrl ah, that kind of stuff 20:56 kahrl why not replace all those cases with a try-read approach? 20:57 sapier in long run this is the way to go but I didn't want to change all those code at once 20:57 sapier maybe I can ask the queue instead of semaphor on windows, I don't want to add a mutexed counter to semaphore implementation 20:58 sapier I thought ReleaseSemaphore(sem,0,&retval) would return the current count in retval while incrementing semaphore by 0 .... but I made same mistake others have done ... 0 is a invalid value 20:59 sapier and retval is not updated at all in this case 20:59 sapier but adding a os dependency to mutexqueue is quite ugly too 21:00 sapier I could ask the queue with that call beeing protected by the mutexqueues mutex 21:01 kahrl you could try NtQuerySemaphore 21:01 khonkhortisan This line will fix the numpad, but I don't know where to put it: if (event.KeyInput.Char != 0) { event.KeyInput.Key = 0 } 21:02 sapier ok better idea true 21:02 sapier I'll try that one 21:03 khonkhortisan Right now I just have case KEY_END: if (evnent.KeyInput.Char == 0) { in intlGUIEditBox 21:07 ShadowNinja I'll push that user-agent change soon if nbody objects. (What section is this in? I guess kahrl knows it so he's in charge if it) 21:07 ShadowNinja +o 21:08 kahrl good question... it affects network, masterserver maybe 21:08 ShadowNinja xyz: Commit s/_MSV_VER/_WIN32/ on that include and below in the useragent setter. 21:08 kahrl httpfetch could be counted as util too 21:09 sapier ok works for me I'm gonna create a pull request for this fix 21:09 ShadowNinja Well, you wrote it, and the whole purpose of sections is so that people who know that code can manage commits for it. So I guess you manage it. 21:11 kahrl that could be said of GUITable too though 21:11 kahrl but that's your section 21:11 kahrl in any case, I'm fine with the useragent change 21:18 sapier https://github.com/minetest/minetest/pull/1102 21:22 sapier btw it's not https://github.com/minetest/minetest/commit/8b0b857eaaa50c6ec217a46c0577395c78ec04c 21:23 sapier but https://github.com/minetest/minetest/commit/04e9a9d5410a151d232a577b46791d2edffba527 who added that bug 21:23 sapier ok I'm still responsible for that bug ;-) 21:56 kahrl sapier: can you change it so it only calls GetProcAddress once? 21:56 kahrl (once per process, thread or semaphore) 21:56 kahrl afaik GetProcAddress is expensive because it has to walk the entire DLL export directory and do a lot of string comparisons 21:58 sapier ok 22:00 sapier better this way? 22:00 kahrl yes 22:00 kahrl (I assume static initializers are run after all DLLs are loaded?) 22:00 sapier hope it works this way 22:01 sapier I hope so 22:01 sapier should be tested on different win32 versions 22:08 kahrl it might cause trouble (probably only if jthread were moved to a dll): http://blog.barthe.ph/2009/07/30/no-stdlib-in-dllmai/ 22:08 kahrl because when static initializers are run, the loader lock is held, and GetModuleHandle acquires the loader lock 22:09 kahrl according to that post 22:09 sapier ok I add a comment to that location 23:14 xyz m_env should be deleted before m_emerge 23:14 xyz in ~Server 23:15 xyz otherwise there's invalid memory access in EmergeManager::setParamsToSettings 23:17 xyz because params are deleted in ~ServerMap which is deleted in ~ServerEnvironment 23:19 xyz btw, it may be a good idea to check stuff with ASan 23:22 xyz it'll catch stuff like this 23:22 xyz https://github.com/minetest/minetest/blob/master/src/guiPauseMenu.cpp#L255 23:22 xyz "// ALWAYS return immediately after quitMenu()" lol no fug you 23:31 aracnus Hi people. 23:33 aracnus I posted a message yesterday on forum, about translating the main Minetest website to other languages (I intend to do this for Brazilian Portuguese). Is there anyone here that I could talk about it? (or, first, is it interesting?) 23:39 xyz removing drop() here https://github.com/minetest/minetest/blob/master/src/guiTextInputMenu.cpp#L118 stops ASan from complaining about heap-use-after-free, not sure why 23:56 iqualfragile are entity movement finaly sent unreliable?