Time Nick Message 02:08 hmmmm guys 02:08 hmmmm so I guess new year's release isn't happening? 02:08 hmmmm :/ 02:09 hmmmm 3 more hours until my time zone's new year 02:09 hmmmm the absolute latest I'd say is in 6 hours (PST) 02:09 VanessaE well 3 hours is enough time if people really buckle down :) 02:10 VanessaE but is it ready is the question? 02:10 hmmmm it really is, but nobody we need is on 02:10 hmmmm and ShadowNinja never gave the final say on if the serverlist fix works or not 02:10 VanessaE I don't see anything particularly wrong with it other than shadow's issue of... yeah, that 02:11 * VanessaE shrugs 02:11 VanessaE I'm on sapier's build right now for his udp/tcp testing stuff, but it's rebased to current HEAD and the server list just works for me. 02:11 hmmmm so I wonder what the biggest problems were with the UDP that he fixed 02:12 VanessaE hmmmm: the CPU usage on the server when a non-cURL client connects and downloads the initial media load 02:12 VanessaE with current vanilla, it tends to hang the server at 100% and lags everyone out 02:12 hmmmm hmm 02:12 VanessaE with his fork, this happsn far less often 02:13 hmmmm oh also 02:13 hmmmm after removing the FPS counter from window title thing 02:13 VanessaE and if you build a client with his branch, the client is far faster over curl or not curl, also 02:13 VanessaE happens* 02:13 VanessaE yes?> 02:13 hmmmm minetest idles at 0% or 1% cpu usage on all cores 02:13 hmmmm jeez 02:13 VanessaE yep, nutz isn't it? 02:13 hmmmm it's actually getting good 02:13 ShadowNinja The serverlist didn't work last I checked (after the commit that suposedly fixed it) and there haven't been any fixer commits since then. Vanessa doesn't have the issue though.. 02:13 VanessaE window managers do stupid shit when you update the titlebar 02:14 hmmmm well 02:14 hmmmm you can't blame it 02:14 hmmmm I gave the reason why it did that earlier 02:14 hmmmm multithread access to UI elements 02:14 hmmmm on Xfce there's no doubt it keeps acquiring that lock 02:15 VanessaE well you kinda can, I mean suppose the window manager updates *everything* about the decorations? I mean redraws the whole thing: side/lower handles, the drop shadows, widgets, etc, instead of just writing some new text. 02:15 VanessaE I could see that killing CPU easily 02:15 hmmmm this problem could be easily non-existent with something like Wayland 02:15 VanessaE perhaps so 02:15 hmmmm where it's only drawn once per frame no matter how often it's updated 02:16 VanessaE but I guess that ain't happening yet 02:16 hmmmm Windows solves this problem by just pushing all of the UI updates to a queue and leaving a UI thread take care of them at its own leisure 02:16 VanessaE last I saw via slashdot (insert grain of salt here), latest X still lacks support for the transition to wayland 02:16 VanessaE -( ) 02:46 VanessaE well here's another request on the forum for a way to get the client to pull a stream from e.g. youtube.. 02:48 VanessaE it would be easy for the server to tell the client where to go (...) but telling the client to do it would be.... not trivial at all. 02:48 ShadowNinja What? In-game YouTube? 02:48 VanessaE yeah 02:48 VanessaE this guy suggests youbube videos being tiled onto a node surface, similar to what we discussed a while back. 02:49 VanessaE er 02:49 ShadowNinja Oh, animated textures. 02:49 VanessaE youtube :P 02:49 VanessaE yeah, except I think he means pulling the stream live from the web. 02:49 ShadowNinja I think adding live streams is going too far. 02:50 VanessaE probably. 02:50 VanessaE but if more requests like this keep coming in, it might be worth considering in the future if it can be done gracefully, somehow. 02:51 VanessaE (I don't expect that this is possible, though. playing any kind of video is CPU-intensive, let alone youtube) 02:53 VanessaE this stream I'm watching right now on livestream, for example, is taking nearly two full cores to decode. so, no. :) 02:54 * ShadowNinja imagines someone adding 50 stream nodes and playing them all. 02:54 VanessaE I'd have to say if the client were to ever do that, only play the stream for the one side of the one node that is the closest to the player. 03:41 hmmmm ShadowNinja, so no word on the serverlist fix? 03:49 ShadowNinja hmmmm: Nothing further ATM. Maybe I can get a second server running. 03:49 hmmmm well, it's not like it matters unless we have the rest of the people needed to do a release 03:51 VanessaE well how many people does it take? 03:53 hmmmm someone to do debian repo updating, someone to do windows compiling, someone to do website updating, and then me :/ 03:54 VanessaE eek, right 03:54 VanessaE didn't think of it from that perspective. 03:54 hmmmm minetest_game tag as well 03:54 hmmmm I don't do marking/tagging or really anything with minetest_game 03:55 ShadowNinja Here's the error: http://pastebin.ubuntu.com/6671550/ (And whoever decided to use GET for JSON data needs to be konked on the head) 03:57 hmmmm what the .... 03:58 hmmmm well first of all it's not properly escaped 03:58 hmmmm or nevermind, it is, I didn't see the periods 03:59 hmmmm that is ridiculous 03:59 hmmmm the JSON should not be in the query parameters, but in the POST body 03:59 hmmmm I am pretty sure this is proller's fault actually, since he's the one who did this 03:59 hmmmm the one who made this* 04:02 VanessaE I wonder if that's why it works for me? 04:02 VanessaE I don't use anything json related 04:02 VanessaE at least not deliberately. 04:03 ShadowNinja VanessaE: Um, the serverlist is JSON-based and you use the serverlist. 04:03 VanessaE oh. heh 04:03 VanessaE figured this was happening minetest-server-side e.g. something in the config 04:03 hmmmm my guess is that one of the parameters are invalid and it's not producing a response because the server threw an exception of some sort 04:03 VanessaE or something that's compiled-in 04:04 hmmmm you can't really figure out what the error is without looking at servers.minetest.net 04:04 ShadowNinja It may be possible that the URL is too long and the request was dropped. 04:04 VanessaE http://pastebin.ubuntu.com/6671577/ 04:04 VanessaE there's what I use, if it's of any help at all. 04:05 ShadowNinja Well, VE probably has a much longer one just from mods. 04:05 VanessaE indeed so 04:06 VanessaE 160-some mods I think 04:23 VanessaE well, there's plenty of Jan 1. remaining to make said release and still be able to call it a "new years" release. 04:24 VanessaE *looks at clock* 4:24am UTC 04:24 VanessaE lots of time yet. 04:24 VanessaE get it out by 11:59pm UTC and you can still legitimately call it that. 04:25 VanessaE depending on how drunk everyone is ;) 04:26 ShadowNinja Not without a working serverlist. But if that's fixed by then I'm O.K. with a NY release. 04:27 VanessaE ShadowNinja: I know this sounds like a stupid idea but could you try setting up a temporary server with my game, 04:27 VanessaE http://minetest.digitalaudioconcepts.com/VanessaE_Game.tar.bz2 04:27 VanessaE and a substantial copy of the config I posted? 04:27 ShadowNinja I notice there are an awfull lot of %20s in that URL(space). 04:27 VanessaE see what it does then? 04:27 VanessaE (I mention my game because of the sheer number of mods) 04:28 VanessaE with a clean copy of current git HEAD of course. 04:28 VanessaE just for shits and giggleds 04:28 VanessaE -d 04:28 ShadowNinja Um, I tested this with my server's workd and minimal minetest_game world. The game can't touch the serverlist other than by modifying settings. 04:28 VanessaE yes I know, but the game's selection of mods DOES affect what info is sent to the master server list. 04:29 VanessaE it could just as easily be a dummy game with 150 fake mods 04:29 ShadowNinja This may be usefull too: http://pastebin.ubuntu.com/6671633/ 04:29 VanessaE as long as it populates the mod list somehow 04:29 ShadowNinja VanessaE: It is populated, and the shutting down message has the same issue. 04:29 VanessaE hm 04:31 * ShadowNinja restores his working directory. 04:33 VanessaE http://pastebin.ubuntu.com/6671641/ 04:33 VanessaE ^^^ just unquoted it with a little python script I found on the web a while back 04:33 VanessaE not that it's useful at all 04:34 VanessaE wait a minute 04:34 VanessaE um 04:34 VanessaE that can't possibly be a legit request 04:37 VanessaE and all I get when I hit that complex URL with firefox, decoded or not, is "{}" 04:49 ShadowNinja We probably shouldn't be using Json::StyledWriter here, it should be Json::FastWriter. 05:07 hmmmm happy new year 05:11 VanessaE happy new year :) 15:40 pitriss Hi I want to ask.. Is there any possible way to do backup of MT world without stopping mtserver? 15:41 VanessaE nope. 15:41 VanessaE if the server is running, it might write to the world file while the file is being copied to the backup 15:41 VanessaE stop the server, copy the file, start the server 15:42 VanessaE then do whatever you need to do with the file, e.g. archive/compress it, send it to off-site backup server, etc. 15:43 pitriss ok thank I thought so.. 15:43 pitriss *thanks 16:20 specing VanessaE: how about doing a FS snapshot? 16:20 specing such as what btrfs/zfs provide? 16:25 VanessaE possible 16:25 VanessaE if such a snapshot can lock out changes completely while it's being taken 16:25 Naked happy new year 16:25 VanessaE but I've never used either of those filesystems, couldn't say for sure. 16:30 sapier VanessaE any new issues with fixed client? 16:30 sapier ähh server 16:35 VanessaE not so far, no 16:35 VanessaE I still use an old client 16:35 VanessaE people are saying it's lighting fast now 16:35 VanessaE lightning* 16:36 sapier ok because I just completed next possible improvement ... a full blown mixed mode TCP/UDP client/server 16:37 VanessaE cooooooool 16:38 sapier https://github.com/sapier/minetest/tree/tcp_implementation_2 old clients can still connect, but can't use tcp of course ;-) 16:38 VanessaE I may be about at the limit of what my VPS can do though (being as slow as it is, I mean) 16:39 VanessaE so I'm not sure how much further of an improvement I'll be able to see :) 16:39 sapier ok transmission of data trough tcp might reach the performance borders :-) 16:40 VanessaE remind me a little later and we'll give that one a try 16:41 sapier ok next step is adding improved unit tests to judge different solutions, after that one I'm gonna give enet a try ... after refactoring connection mess it's quite straight forward adding/removing new protocols ... and interface gets even more clean with each protocol 16:45 nore sapier, I have a question: is the minetest protocol different for each of tcp, udp, etc, or is it the same each time and there is only a backend to know which network protocol to use? 16:45 sapier I assume when you say "protocol" you actually mean "packet format" no packet format is same for all 16:46 sapier put of course for network transmission there is added an additional envelope around that format 16:48 nore ok... and yes, when I said minetest protocol, I meant packet format 16:50 sapier I guess replacing packet format will never be a quick task no matter how often you refactor 16:51 nore I wonder whether we could save some bandwidth by changing the packet format and dropping compatibility... 16:51 sapier no we couldn't 16:52 sapier as of bandwith usage our current format is quite close to optimum 16:52 nore good then 16:54 sapier we could add compression of course ... but media cant be compressed and other datare are typicaly < 100 bytes so compression overhead might even add additional data 16:54 sapier assuming media is compressed png in most case 16:55 nore and are mapblocks sent compressed? 16:55 nore if not, we could save some space for example, if the mapblock is only stone... 16:55 sapier hmm mapblocks could be an interesting point yes haven't thought about those 16:56 hmmmm ... 16:56 hmmmm have you people never looked at mapblock serialization? 16:56 sapier but that's finetuning first we need to find a way to decide what future network protocol will be 16:57 nore hmmmm, no, I haven't 16:57 hmmmm also a bare binary protocol isn't close to optimum :)... you can do even better with data prediction 16:57 sapier no I haven't done anything about maps by now hmmmm I usually ask you for those things ;-) 16:57 nore how are they serialized? 16:57 sapier data prediction is something completely different ;-) 16:57 hmmmm have you ever seen Efficient XML? 16:58 sapier xml isn't a binary format ;-) 16:58 hmmmm well 16:58 hmmmm the structure of XML doesn't need to be 16:58 hmmmm binary XML isn't anything new 16:58 hmmmm see infoset 16:58 sapier for what I see we don't have a lot of redundant information in our packets 16:59 hmmmm well Xpath infoset to be precise 16:59 iqualfragile https://github.com/minetest/minetest/pull/1076 merge plz, tested with server using those patches, its still possible to connect using a 0.4.6 client (and one using that code obv.) 16:59 sapier if the information is relevant for the sending entity is out of scope of packet format ;-) 16:59 iqualfragile 0.4.6 is the version of minetest included in jessie 17:00 sapier 1076 is not meant for direct merge 17:01 iqualfragile why not? 17:02 sapier lots of cleanup missing 17:05 sapier most of that cleanup is in my tcp branch right now but I guess I'm gonna extract it soon 17:05 sapier ok maybe not soon ... first I'm gonna write the protocol unit tests 17:07 hmmmm wouldn't it be a better idea to use clock_gettime() instead of gettimeofday() and just include a wrapper for windows around gettimeofday? 17:07 hmmmm iirc gettimeofday() is poor on performance because on some platforms it does a system call 17:08 sapier give me a hint where in code you are right now 17:08 hmmmm I'm just speaking in general 17:08 hmmmm but the new JSemaphore wait 17:09 hmmmm apache's use of gettimeofday() used to be a really big performance bottleneck on freebsd 17:09 sapier you could be right yes but we need to be carefull about right clock usage (not sure if current way is correct in any case) 17:09 hmmmm don't know if it's still the same way anymore 17:10 sapier especially we need to be sure to use the same clock as semaphore will use for timout 17:12 hmmmm CLOCK_REALTIME should be essentially the same as gettimeofday() but with more resolution 17:13 hmmmm I really don't like sem_timedwait's use of an absolute time, because it can be affected by changing the clock 17:13 sapier I know 17:13 hmmmm gotta wonder what they were thinking when they came up with it 17:13 sapier that's ugly if you have a better idea how to do it you're welcome 17:14 hmmmm the obvious thing is to use a relative time 17:14 hmmmm but you can't go back in time and change the spec 17:14 hmmmm well SysV 17:14 sapier relative time would've been more work to implement ;-) 17:17 sapier hmm semaphore timeouts may use a different timesource as CLOCK_REALTIME too ... :-( another ugly thing 17:18 hmmmm the timesource should be the same, the UTC on-the-wall clock 17:18 sapier "If the Timers option is not supported, the timeout shall be based on the system clock as returned by the time() function." I just don't know what is meant wit timer option 17:20 sapier "Timers" is an optional semaphore feature as of IEEE Std 1003.1 ... guess best thing would be use clock_gettime and fallback to gettimeofday ... or just live with errors in case no realtime clock is available ;-) 18:35 celeron55 i managed to make a ridiculously hacky qmake build of irrlicht in the sailfish sdk 8) i had to write my own wayland backend for irrlicht, to which i hand the window and display pointers from a qt program 18:36 celeron55 this mer/sailfish/whatever sdk is the most obscure thing i have ever seen 18:39 celeron55 it's theoretically possible to use this toolchain with cmake, but the information for actually figuring out how to do that simply doesn't exist, so i guess next i'll try to just cram minetest into this same qmake project 18:47 celeron55 t3 18:47 celeron55 -t3 19:33 VargaD celeron55: do you have adevice to run mer/salfish? 19:42 celeron55 yes 19:42 celeron55 Using renderer: OpenGL ES 2.0 V@6.0 AU@ (CL@2961769) 19:42 celeron55 Qualcomm 19:42 celeron55 8) 19:44 VargaD nice :) 19:49 celeron55 ../../test3/minetest/mtsrc/src/game.cpp:3477:1: internal compiler error: in extract_insn, at recog.c:2113 19:49 celeron55 ohsit 19:50 celeron55 the sailfish sdk uses gcc 4.6; apparently it's a good idea \o/ 20:44 VargaD Does it compilacted to add support for more "worlds" in a server? (and also possibility to move to other world in the game) 21:07 VanessaE um, wut> 21:07 VanessaE terminate called after throwing an instance of 'con::PeerNotFoundException' 21:07 VanessaE what(): GetPeer: Peer not found (possible timeout) 21:07 VanessaE /home/minetest/Scripts/minetestserver-survival.sh: line 22: 25197 Aborted 21:12 sfan5 VargaD: shouldn't be that hard, you would just need multiple server environments 21:36 celeron55 http://c55.me/random/2014-01/tscrot-2014-01-01_23-34-11.png 21:36 celeron55 lol epic 21:36 VanessaE well at least it RUNS :D 21:37 VanessaE (insert Xzibit/yo dawg here) 21:37 VanessaE all you need now is one of those mods in the game that runs a virtual computer e.g. forth or whatever 21:37 celeron55 i have sacrificed my brain to the god of awful crap 21:37 celeron55 you won't believe what this thing already contains 21:38 hmmmm guys 21:38 hmmmm come on i'd like to make a release this century so we can get new stuff 21:38 VanessaE celeron55: at this point, I think I'd believe almost anything :P 21:38 celeron55 also, this doesn't for example actually work 21:39 celeron55 there's no input system at all in place (no mouse, no keyboard, no tocuh), and sailfish thinks this isn't responding (while it is) and kills it after a few seconds 21:39 celeron55 hmmmm: how about hilighting the people who you actually need for that 21:39 hmmmm I don't know who does it normally 21:40 celeron55 see wiki 21:40 hmmmm I need someone who updates the debian repo, updates the website, and builds the windows versions 21:41 hmmmm sapier and sfan might build windows versions so that's why I'm not singling any one person out for that 21:41 celeron55 i can try the launchpad stuff (but it's kind of a post-release thing) 21:41 hmmmm alright 21:41 hmmmm after release then 21:41 sfan5 hmmmm: s/sapier/xyz/ that is how it was last time 21:41 celeron55 and after that there's plenty of people who can update the website 21:41 celeron55 so you need the windows build and nothing else really 21:41 hmmmm and I also need word from shadowninja to see if the serverlist works or not 21:41 celeron55 and it needs to be tested before tagging 21:41 xyz oh god what's the point of porting it to an os no one uses 21:42 hmmmm xyz, woudl you like to build a windows version for minetest? 21:42 xyz right now? sure 21:42 PilzAdam hmmmm, when do you plan to release? 21:42 hmmmm RIGHT NOW 21:42 hmmmm alright 21:42 hmmmm we have the gang together 21:42 PilzAdam without a feature freeze? 21:42 hmmmm about that 21:42 hmmmm what we need is for the master branch to be in what you could consider a stable state 21:43 hmmmm a feature freeze is just a formal way of telling people to stop adding new shit 21:43 celeron55 xyz: well, why not, if it's doable; sailfish is going to get a lot more users after some time 21:43 hmmmm there has been nothing but bugfixes and very minor enhancements for the past 2, no 3 weeks 21:43 hmmmm s/has/have/ 21:43 PilzAdam https://github.com/minetest/minetest/issues?labels=&milestone=2&page=1&state=open 21:43 sfan5 can we add the tab playerlist (w/o lua) before 0.4.9? 21:43 hmmmm set_sky() is controversial, dunno about that yet 21:43 hmmmm tab playerlist, NO. 21:43 hmmmm it doesn't work 21:43 sfan5 >(w/o lua) 21:44 hmmmm it doesn't matter 21:44 xyz celeron55: dunno; I heard it runs android apps so I see very little point in porting to it; it'd be much better for someone to write a generic touch screen ui instead 21:44 sfan5 mhm, ok 21:44 hmmmm it'll be in the next release 21:44 PilzAdam formspec table needs to get in to make the menu more useable 21:44 hmmmm it's fine if this happens once every month and a half 21:44 celeron55 xyz: well yeah, it runs freeminer just fine 21:44 hmmmm those two bugs, I have no idea about 21:44 PilzAdam it should have been done last release alraedy 21:44 hmmmm well that's alright 21:45 celeron55 xyz: i mean... buildcraft 21:45 hmmmm it needed to get in a more finished state then 21:45 celeron55 xyz: but i want to learn the toolchain of this thing so i might just as well do it with minetest 21:45 hmmmm and then it was enhanced and now it's in a semi-broken state 21:45 xyz I see 21:45 hmmmm I need to consult with sfan more about that 21:45 hmmmm so PilzAdam, bump the version of minetest_game and tag it, would you? 21:46 PilzAdam lets wait for the win builds first 21:46 hmmmm sure 21:46 hmmmm man, I wanted to do this last night 21:46 * sfan5 makes a win build 21:47 PilzAdam https://github.com/minetest/minetest/blob/master/doc/menu_lua_api.txt#L1 <- can we get this right this time? 21:47 sfan5 umm.. aren't windows builds pointless now when the version was not bumped yet? 21:47 hmmmm yeah 21:47 PilzAdam sfan5, to see if it builds at all 21:48 hmmmm I'm doing that in a second 21:48 hmmmm that also 21:48 hmmmm and PilzAdam, yes I did that 21:49 hmmmm so I'll push the version bump if windows compiles successfully 21:50 PilzAdam is the changelog up to date this time? the one for 0.4.8 is still not updated 21:51 hmmmm I guess not 21:51 hmmmm who cares about the changelog, we can do that later 21:51 hmmmm or whoever does the changelog 21:51 PilzAdam "we can do that later" <- see 0.4.8 changelog 21:51 PilzAdam nobody cares about that after a release 21:52 hmmmm just point to the URL of the commit log 21:52 VargaD nice window manager celeron55 21:52 PilzAdam we write the changelog for users who dont understand / dont have enough time to look through the commit log 21:53 sfan5 hmmmm: builds successfully 21:54 hmmmm great 21:54 celeron55 VargaD: it's the best one out there; i recently tried gnome, kde and mate when i bought a new computer and those just sucked 8) 21:55 hmmmm wtf is with the -dirty thing 21:55 hmmmm that's giving me problems 21:55 VargaD Which one you are using celeron55? 21:56 PilzAdam it appends that if you have uncommited changes in the git repo 21:57 celeron55 VargaD: this is icewm, with the glorious yellowmotif theme; almost default configuration; let's not offtopic more 21:58 hmmmm alright I can't get it 21:58 hmmmm I tried refreshing the cmake, I tried make cleaning, I tried everything 21:58 VargaD sorry, I was just curious, btw I'm running awesome 21:58 PilzAdam hmmmm, should I try it? 21:59 hmmmm hrmm 21:59 hmmmm I dunno 21:59 hmmmm it'll probably work for you 21:59 hmmmm something is busted with my own setup 21:59 celeron55 VargaD: awesome is probably equally designed for real use 8) 22:00 xyz hmmmm: just edit cmake/modules/GenerateVersion.cmake 22:00 hmmmm that's the hacky way though 22:01 xyz this shit is so annoying though 22:01 hmmmm I know 22:01 hmmmm I want to soo get rid of it 22:01 PilzAdam rm .git and build 22:01 xyz that's the easiest way 22:01 xyz if we're releasing in a rush anyway 22:02 hmmmm oh 22:02 hmmmm you need to make the tag first 22:03 hmmmm got it 22:03 PilzAdam https://github.com/PilzAdam/minetest/commit/ed640d5daaed2f9781ac9623355800476510a08b 22:04 hmmmm I already did the same exact thing 22:05 PilzAdam no, you didnt 22:05 PilzAdam there is a "=" missing in doc/menu_lua_api.txt 22:05 hmmmm oh come on 22:05 hmmmm one =? 22:06 PilzAdam .... and you already taged it without waiting for MSVC build 22:06 hmmmm you needed to tag it in order for GenerateVersion to produce the right version string 22:06 PilzAdam only if you build in a git repo 22:06 xyz firstly you need to confirm that msvc build works 22:06 hmmmm sfan5 did that 22:07 PilzAdam no, sfan5 builds with mingw 22:07 hmmmm well 22:10 xyz by the way, was the subsections thing by celeron55 accepted? 22:22 xyz I've uploaded my build to github release draft 22:31 * sfan5 is still building 22:42 sfan5 uploaded mine! 23:02 VanessaE still about 1 hour to go if you guys still wanna get that release done on Jan 1 (it is 23:03 UTC)