Time Nick Message 09:37 Megaf [4597527.257472] Out of memory in UB 102: OOM killed process 21319 (minetestserver) score 0 vm:1209008kB, rss:507976kB, swap:519396kB 09:37 Megaf Killed 09:37 Megaf Can you please do something about it? 10:40 Megaf I found yet another bug, something that used to work and no longer works 10:40 Megaf retporting 10:48 Megaf https://github.com/minetest/minetest/issues/1878 10:50 kahrl Zeno`: you bwoke it! :P 10:55 Zeno` yes :) 10:55 Zeno` Megaf, are you sure the world exists? 10:55 Zeno` hmm 10:55 Zeno` it shouldn't exist 10:55 sol_invictus Zeno`: it happens when connecting to a remote servere 10:55 * Zeno` looks 10:55 Megaf Zeno`: It shouldnt ask for a world in first place... 10:58 Zeno` this is kahrl's fault 11:00 Zeno` I like your password 11:00 Megaf :) 11:00 Megaf safest password ever 11:00 Zeno` yep 11:02 kahrl Zeno`: could be, I don't remember... apparently github deleted some of my commit comments 11:02 Zeno` I remember fixing --go ages ago 11:03 Zeno` obviously it has crept back in somehow 11:03 Zeno` I wasn't really blaming you btw :P 11:05 kahrl I guess this is somehow related to current_address not being set properly 11:06 kahrl really, let's get rid of those variables as they are confusing as heck 11:08 Zeno` yeah 11:08 Zeno` which is one reason I tried to get rid of them in the first place 11:08 Zeno` if they were not there this may not have happened but I get confused every time I look at them 11:09 Zeno` they're pointless (apart from causing confusion) 11:09 kahrl well... the reason they were added was singleplayer mode 11:10 Zeno` Probably made more sense when it was a 1000 line function 11:10 kahrl which sets the player name to "singleplayer" but you don't want this to replace the name field when you go back to the menu 11:10 kahrl and similar for the other variables 11:12 kahrl I think putting them into GameParams makes sense 11:14 kahrl simple_singleplayer_mode would fit there better as well 11:18 Zeno` yes ok 11:19 Zeno` kahrl, are you working on this? If you are I won't duplicate the effort (I'm too tired anyway, so I hope you are :)) 11:20 kahrl not atm 11:20 Zeno` hmm ok 11:20 kahrl might do so later 11:21 * Zeno` will get some caffeine just in case 11:31 Zeno` MegaF, ok fixed. Will do some more tests and make a PR 11:33 Megaf Cool 11:33 Megaf One more critical bug for ya 11:33 Megaf https://github.com/minetest/minetest/issues/1880 11:33 Megaf #1880 11:33 ShadowBot https://github.com/minetest/minetest/issues/1880 -- minetestserver will keep consumig RAM till it uses all available RAM and SWAP. 11:37 Megaf can we close #1650 giving a reason to not merge? 11:37 ShadowBot https://github.com/minetest/minetest/issues/1650 -- Smoother player movement (Faster jumping, slight height boost) by TriBlade9 11:41 kahrl why? 11:41 Megaf because it is not going to be merged 11:41 kahrl why do you think so? 11:53 Zeno` kahrl, https://github.com/Zeno-/minetest/commit/093974633cac4d83459fd2507a93d1df575b5fa8 11:56 kahrl Zeno`: wouldn't that lead to problems when the game is launched with both an address and a world_path? 11:56 kahrl I guess I don't completely follow the logic there 11:56 Megaf you need and if address is submited then ignore world_path 11:56 Megaf s/and/an 11:57 Zeno` an address AND a world path? That can happen? 11:57 Megaf Zeno`: never trust the user 11:57 Zeno` Yes, but does it make sense? 11:57 Megaf Zeno`: that patch works 11:58 Zeno` I mean, is address + world_path a valid condition? 11:58 Megaf Zeno`: will try no giving the world path too 11:58 Zeno` If not I'll add an extra check; if so I will amend 11:58 kahrl Zeno`: what if the address is set in minetest.conf and a world path is passed on the command line? 11:58 Zeno` karhl, it will probably break 11:59 Megaf kahrl: if world is provided it will be ignored and server will still be connected to 11:59 Megaf just tested 11:59 * Zeno` checks source code; I'm actually pretty sure that's tested for somewhere else 11:59 Megaf Zeno`: no need, it works 11:59 kahrl ok, so this if clause is (and was) just defensive programming 12:00 kahrl Megaf: ... there's a reason professional QA teams do both black box and white box testing 12:01 Zeno` It looks like it. There are a lot of things I would like to change in main.cpp but the initial goal was to refactor and get it in a state where you would not lose your sanity doing so :) 12:01 Zeno` e.g. there are duplicate checks in more than one place, but I left them there trying to follow the original code as closely as possible 12:02 kahrl yeah, that makes sense 12:02 Zeno` I'd rather the conditions were closer together 12:02 Zeno` And will work on that maybe next week 12:02 kahrl just remove this condition in particular if it serves no need 12:05 Zeno` triple checking 12:06 Zeno` I wonder why it was there in the first place 12:07 kahrl brb 12:18 Megaf "Massif runs programs about 20x slower than normal." 12:19 Megaf That will be fun 12:19 Megaf we have to make minetest 20x faster so we can run massif and still have fun :P 12:20 Zeno` kahrl, I'm leaving it there because although it's terrible it is needed 12:21 Zeno` next iteration of the refactor this can be fixed; I'll add a FIXME comment 12:25 Megaf Zeno`: can you creat a PR please? 12:25 Zeno` getting to it :) 12:25 Megaf cool 12:26 Megaf Zeno`: kahrl: I'm making a snapshot of my server now, when it's complete I will run valgrind on it to try to find what is using all that ram 12:26 Megaf by the way, I think #1881 should not be too difficult to fix. 12:26 ShadowBot https://github.com/minetest/minetest/issues/1881 -- on shutdown minetestserver will throw ERROR[CurlFetchThread]: servers.minetest.net/announce not found (HTTP response code said error) (response code 500) 12:30 Zeno` #1882 12:30 ShadowBot https://github.com/minetest/minetest/issues/1882 -- Fix regression with command line --go option by Zeno- 12:32 Zeno` kahrl, it's not ideal but can I merge #1882 for now (to make it not a blocker any more) 12:32 ShadowBot https://github.com/minetest/minetest/issues/1882 -- Fix regression with command line --go option by Zeno- 12:33 kahrl Zeno`: yeah, seems good 12:33 Zeno` ok cool 12:33 Zeno` I'll look more closely at the structure of main.cpp next week (it has more than one confusing bit as you know) 12:37 Zeno` I think Travis is having a snooze 12:38 Zeno` what is massif? 12:39 Zeno` oh nice... a new tool! 12:39 Zeno` (new for me) 12:41 Megaf Zeno`: is for me too :) I'm recompiling my server in debug mode now to run massif 12:59 Zeno` any comments on #1876 and #1879? 12:59 ShadowBot https://github.com/minetest/minetest/issues/1876 -- Optimise VoxelManipulator::copyFrom by Zeno- 12:59 ShadowBot https://github.com/minetest/minetest/issues/1879 -- Improve performance by Zeno- 12:59 Zeno` Also, does anyone speak/read Japanese? (#1787) 12:59 Megaf Zeno`: on the client side I haven't notice any regression with them 13:00 Megaf s/notice/noticed 13:01 sfan5 Zeno`: i do not speak japanese but 1787 looks ok 13:02 Zeno` I'll google translate it :) 13:03 sfan5 that won't get you anywhere 13:03 Zeno` it will in a way 13:05 Megaf um, valgrind makes thing very slow indeed 13:05 Megaf I have dug some cobbles and placed soem torches and minetestserver still haven't noticed that. 13:05 celeron55 yes it will; you won't be able to play but if it consumes RAM anyway, just leave it until it's risen plenty 13:05 Zeno` ok, the updated .po looks good. I'll merge it 13:06 celeron55 leave it for like 10 times the time it would take for RAM consumption to increase enough 13:06 celeron55 also note that it will consume multiple times the memory as massif is keeping track of it 13:06 Zeno` or 20 times 13:06 celeron55 so you need a better machine than the one that oomkilled minetest at half a gigabyte 13:06 Megaf I will keep an eye on it 13:06 sfan5 massif has a massif memory consumption ;) 13:07 Megaf -/+ buffers/cache: 280 743 13:07 Megaf Swap: 1024 0 1024 13:07 Megaf so far so good 13:07 Megaf total | used | free 13:07 Zeno` lol, you need to give it long than a few minutes :P 13:07 Megaf oki 13:08 Megaf you know what, I will run it on my desktop where I have 8 GB of ram and four 3,2 GHz cores 13:08 celeron55 there you might even be able to play it 13:09 Megaf I keep a fresh snapshot of my server on my desktop as backup 13:09 celeron55 make sure to use the same world, could be due to it too 13:09 Megaf How do I stop valgrind? 13:09 celeron55 ^C 13:10 Megaf ok, shutting down... 13:11 celeron55 if you run it long enough, you should end up in a situation where clearly something is consuming the vast majority of memory 13:11 celeron55 if you now look at the output (rtfm to do that), it should look like nothing in particular is doing that 13:11 celeron55 as it ran so short time 13:13 celeron55 basically just browse through ms_print's output to where the most memory is being continuously used and see what's taking most of it 13:13 celeron55 (also, i am assuming it doesn't leak particularly much memory, which you could check with memcheck first) 13:14 celeron55 (that won't matter much in the end though, it's just that massif won't tell you whether the memory was leaked or not) 13:18 Zeno` memcheck shows nothing 13:24 Fritigern I was wondering if anyone knows if anybody has ever gotten MT to work on a Raspberry Pi, which has an ARM CPU 13:24 Zeno` Fritigern, Megaf has 13:24 Zeno` (I think) 13:25 sfan5 Fritigern: minetest works on android, android runs a raspberry pi => yes 13:25 sfan5 the rpi CPU sucks tho' 13:25 Fritigern Coolness. Do you happen to remember if its performance was anywhere near acceptable? 13:25 sfan5 no idea 13:25 sfan5 i didn't ever have mt running 13:26 Fritigern Actually, the Raspberry Pi FAQ states that although there is an android version for RaspPi, it is highly unstable 13:27 Megaf sfan5: android doesnt works on Pi 13:27 Megaf not properly 13:27 sfan5 that is probably because android 4.x only (really) works on armv7 13:27 Fritigern http://www.raspberrypi.org/help/faqs/#softwareAndroid 13:27 Megaf Fritigern: minetestserver runs fine on raspberries 13:27 Megaf I used to have a server 13:28 Fritigern What about the client? 13:28 Megaf Fritigern: I managed to compile it but not to run it 13:28 Megaf Fritigern: https://github.com/Megaf/MinetestPi-Raspbian 13:29 Fritigern Thing is that i have zero experience with RaspPi, but i am interested. I think it could be a lot of fun playing with it and eventually have a few laying around the house for various purposes 13:30 Megaf [09:57] a good example of a well done game is Quake III 13:30 Megaf [09:58] it can run with filterings, settings on max and in full HD, 1920x1080 with 256 MB of ram on a 600 MHz ARM CPU, Raspberry Pi 13:30 Megaf ^ Fritigern 13:30 Megaf sfan5: Dont say that it CPUs sucks 13:31 sfan5 it does 13:31 Fritigern I could imagine using one with OpenElec for movies and music, antoher one in the kitchen to keep a shopping list and a DB of recipes. Yet another one with a camera in the nursery to keep an eye on the baby, etc.... 13:31 Megaf minetest is just not optimized of it 13:31 Megaf Fritigern: OpenELEC runs very well 13:31 Megaf it's perfect actually, but not the topic for here 13:31 sfan5 it can run with filterings, settings on max and in full HD, 1920x1080 with 256 MB of ram on a 600 MHz ARM CPU, Raspberry Pi 13:31 sfan5 wat 13:31 sfan5 the most is done by the gfx card 13:31 sfan5 not the cpu 13:32 Megaf maybe that's the biggest problem in minetest xP 13:32 Fritigern And who knows, there will eventually prolly be a RaspPi C or D which is more powerful, which i might use as a gaming machine for a 5-6 year old 13:32 Zeno` I'm working on it! 13:32 Megaf https://www.youtube.com/results?search_query=quake+iii+on+raspberry+pi 13:33 Fritigern Megaf: Wouldn;t it be fun if you could compile Singu for RaspPi? :-) 13:33 sfan5 Megaf: the cpu itself sucks 13:33 Zeno` But... I'm giving her all she's got, Captain! 13:33 sfan5 not minetests fault 13:33 Megaf I have no idea what Singu is, and again, this is not the channel to talk about that 13:33 Zeno` All she's got isn't good enough! What else ya got? 13:34 sfan5 Fritigern: or you just buy a banana pi.. 13:34 Zeno` Um... Okay, if we eject the core and detonate, the blast could be enough to push us away! I cannae promise anything, though! 13:35 Zeno` I actually can get a huge performance increase by not rendering leaves on trees 13:35 Zeno` but it looks a bit stupid 13:37 Fritigern Could that rendering somehow be off-loaded to the GPU? 13:37 sfan5 umm 13:37 sfan5 the GPU already does the rendering 13:37 sfan5 not using the GPU would be kinda slow 13:40 celeron55 quake 3 is a highly static game 13:40 celeron55 it doesn't anything for the CPU to do except some simple physics 13:41 Megaf ok, but since ABMs run on server side, the client should work nice on a raspberry pi 13:41 celeron55 lol 13:41 Megaf mapgen and abms and heavy stuff are server side 13:41 celeron55 the CPU is probably like 50 times slower than your desktop CPU 13:41 Fritigern I don;t know how to explain what i mean really. My reasoning was that for instance most Nvidia cards can use CUDA to help with number crunching, AMD cards have something that is supposed to work similar (i forgot what it's called), so perhaps those technologies (when present) could be used to improve performance? 13:42 exio4 opencl 13:42 Megaf celeron55: the whole problem I saw in minetest running on a pi is that it is not using OpenGL ES 13:42 Megaf its doing all work via software 13:42 celeron55 Fritigern: maybe, but that's a lot of development work; none of that is automatic and none of that works on less gaming hardware on which minetest has to run too 13:43 celeron55 Megaf: sounds like a driver problem 13:44 Fritigern celeron55: Of course, if those technologies are absent, the sofware should default to the traditional methods. Anyway. it was just an idea. I have no idea how much work it would be 13:44 Megaf celeron55: it is, but instead of devs trying to fix that they just keep saying that the (insert hardware here) is too slow 13:45 Megaf the problem I'm reporting is that there is no, or little support to OpenGL ES 13:45 Megaf the repply I get is, (Your CPU sucks) 13:45 Megaf ^ sfan5 13:45 kilbith it could work very well - look at Minecraft-Pi.. 13:46 Megaf In this channel alone there are at least three people very interested in running Minetest on the Pi, kilbith Fritigern and me 13:46 Megaf imagine elsewhere 13:46 kilbith some people are interested on the official raspi forum too 13:46 celeron55 what i'm saying is that dynamism (is that a word?) always costs performance, no matter what; not sure that the theoretical limit is here in various aspects though 13:47 celeron55 or practical limit 13:47 exio4 my phone has about the same CPU as the rpi, and it lags badly for anything, I would like to know how it would be with the rpi, maybe its gpu is better 13:48 Megaf and besides all that, there is already a working port of MT for android 13:48 Megaf and it runs on some low end phones 13:49 Megaf not too hard to just include the Pi GPU on the list.. 13:49 Megaf since MT already supports in, on Android... 13:49 Fritigern The GPU is a VideoCore 4, should anyone wish to know 13:51 Megaf its the same GPU on Galaxy S3 I believe 13:51 Megaf and on some iPhones 13:51 Megaf pretty decent GPU 13:51 exio4 the cpu isn't that decent anyway 13:51 Megaf exio4: still, cpu is not the problem 13:51 exio4 even in their site they say it is more like a pentium 3 with a shiny gpu :P 13:51 Megaf MT lack of OpenGL ES support is 13:52 exio4 I would have bet it had support of opengl es 13:53 Zeno` sfan5 patched irrlicht to support opengl es (right?) 13:53 Megaf if anyone is interesed, I can try to build stock MT on my Pi today, once again 13:54 Fritigern Megaf: The results of that would be interesting. 13:54 kilbith Zeno`: Irrlicht team has patched; there's an official branch on Git 13:55 Zeno` kilbith, yeah I think sfan5 (or someone) added some additional patches. Not sure now 13:55 kilbith for MT 13:55 Zeno` yes 13:55 Zeno` I think we should just port to Ogre3d :D 13:56 Megaf +1 13:56 Megaf I totally support droping irrlicht 13:56 kilbith Megaf: i'm interested 13:58 Megaf exio4: Pentium II 13:58 Megaf "Overall real world performance is something like a 300MHz Pentium 2, only with much, much swankier graphics. " 13:58 PenguinDad Zeno`: I hope you're joking 13:58 Zeno` pity it involved so much work 13:58 Zeno` PenguinDad, not 100% joking 13:59 exio4 Megaf: even worse 13:59 Zeno` there is no reason irrlicht and Ogre3d cannot both be supported. Apart from a lot of work needed 13:59 Megaf there is no reason Lua and Python cannot both be supported. Apart from a lot of work needed 13:59 Zeno` (and I really mean a lot) 14:00 Megaf :P 14:01 Fritigern Ooh, here's a video of someone running 0.4.3 on a RaspPi http://www.youtube.com/watch?v=nCbnF9OAne4 14:02 Megaf celeron55: when I try to run the server on valgrind on my desktop it crashes 14:02 Megaf In thread 4062540: 14:02 Megaf /home/megaf/RAM/Server/src/script/cpp_api/s_base.cpp:75: ScriptApiBase::ScriptApiBase(): Assertion 'm_luastack' failed. 14:03 Zeno` Megaf, you have to build with LuaJIT disabled 14:04 Megaf why? 14:04 celeron55 oh yes, valgrind doesn't like just-in-time compiling 14:04 Megaf and it works with luajit on a Xeon 14:04 Megaf how, how do I disable LuaJIT? 14:04 celeron55 i have never seen luajit and valgrind working simultaneously 14:05 celeron55 -DDISABLE_LUAJIT=1 should work 14:05 celeron55 yes, not documented 14:06 Zeno` I think it's documented now 14:07 Zeno` err... well, not the issue with LuaJIT 14:08 Zeno` You *can* make a special build of LuaJIT that works with valgrind but it's a bit of a hassle 14:09 Zeno` Megaf, http://repo.or.cz/w/luajit-2.0.git/blob/HEAD:/src/Makefile#l123 14:10 Zeno` It's too much hassle for me though 14:11 Megaf since I will just keep the server on doing nothing I wont use that 14:16 Megaf is it possible to run a MT client without video? 14:16 Megaf just to test a server from a headless box 14:32 sfan5 Megaf: opengl es support is entirely on Irrlichts side, telling us won't help 14:32 Megaf Zeno`: Let's make an Ogre port of MT :P 14:33 sfan5 Zeno`: not my patches, official irrlicht stuff 14:36 Megaf LuaJIT disabled and valgring running = max_lag 0.86 14:36 Megaf the power of AMD Phenom II x4 14:58 Megaf 5.8% of 8 GB is? 15:00 Megaf about 500 MB? 15:01 PenguinDad Megaf: ~475.136 MB 15:01 Megaf ok 15:01 Megaf celeron55: I think I can stop vagrind now, can't I? 15:02 Megaf ServerThread is already using 5.8% of 8 GB 15:06 Megaf server stopped 15:08 Megaf Zeno`: now, what do I do with the output of massif? 15:11 Megaf sfan5: any idea on what I do now? 15:12 sfan5 dunno 15:12 sfan5 maybe find what ate all the RAM 15:12 Megaf yep, I have the output, but I dont know how to read it 15:12 Megaf lol 15:13 sfan5 <celeron55> if you now look at the output (rtfm to do that), [...] 15:14 Megaf I will try to make a graph 15:17 celeron55 just pastebin the ms_print output 15:17 celeron55 other than that, google the massif manual 15:17 celeron55 it's not meant to be used blindfolded 15:21 Megaf right, the raw output is here, https://gist.github.com/Megaf/79fdd09dec33e94f00e5 15:21 Megaf I will use ms_print now 15:32 Megaf celeron55: Zeno`: kahrl: ms_print output of valgrind. https://gist.github.com/Megaf/50aac763e9f0bf85adb9 15:33 Megaf RAW -> https://gist.githubusercontent.com/Megaf/50aac763e9f0bf85adb9/raw/f91788eee3a586d292f54302ce8c3a6d74da0f48/massif.out.1536.ms_print 15:53 Zeno` hmmmm, I am going to sleep now, but can you check #1876 and #1879 please 15:53 ShadowBot https://github.com/minetest/minetest/issues/1876 -- Optimise VoxelManipulator::copyFrom by Zeno- 15:53 ShadowBot https://github.com/minetest/minetest/issues/1879 -- Improve performance by Zeno- 15:53 Zeno` night all 15:58 Megaf Zeno`: I can, not today 15:58 Megaf good night Zeno` 17:25 hmmmm Zeno`, the fast boolean patch is not thread-safe 18:27 Krock h#1883 18:30 kahrl I wonder why init_gettext doesn't simply use GetCommandLine() instead of that argc,argv hack 18:31 kahrl sapier: ? 18:32 kahrl but yeah Krock's PR looks fine 18:39 Krock strange mesage whenever I join a singleplayer world: 19:43:50: ERROR[main]: WARNING: Overriding gamespec from "base" to "minetest" 18:45 kahrl oh gawd... the w32api is so dreadful http://stackoverflow.com/questions/3292686/process-name-missing-from-getcommandline 18:48 sapier https://github.com/minetest/minetest/pull/1561 not a single comment to the updated version???? I'm gonna merge this one at the end of this weekend so don't complain afterwards. It's NOT a "NOTHING CHANGES" fix but really fixes the scaling issues. Some visible things have to change to get this. 18:49 kaeza kahrl, still makes sense 18:50 kahrl kaeza: you mean the GetCommandLine behaviour? 18:50 kaeza his second example is akin to calling execv("foo", "bar", "baz") (i.e. you run `foo`, but `argv[0]` is `bar`, not `foo`) 18:50 Krock sapier, looks good 18:51 kahrl kaeza: I would agree if there was any way whatsoever to tell from the second process which of the two cases is happening 18:52 kahrl but really, since CreateProcess is supposed to be low level, it shouldn't be trying to parse the command line to find the executable anyway; that's the job of an API like ShellExecute 18:52 kahrl it's too late to change that though 18:54 sapier we do parse the executable name from command line args? 18:54 kaeza for the record, I'm not saying winapi is rainbows and unicorns :P 18:55 * Megaf looks at #1561 18:56 ShadowBot https://github.com/minetest/minetest/issues/1561 -- Implement proper font handling by sapier 18:56 kahrl sapier: I think the MSVC startup code does, by calling CommandLineToArgvW 18:56 kahrl which I guess does some magic (comparing with GetModuleFileName(NULL) probably) to decide if the command line contains the executable name or not 18:56 Megaf sapier: does it depends on freetype? 18:57 sapier megaf you mean 1561? no it doesn't depend on freetype 18:57 Megaf ok 18:57 sapier kahrl ugly... but I guess windows is always ugly 18:58 Megaf big patch sapier 18:58 Megaf still downloading her 18:58 Megaf here* 18:58 Megaf you worked a lot on it, didnt you? 18:59 sapier partial, fontengine is from me, the formspec adaptions aren't 18:59 Megaf $ git apply 1561.patch 18:59 Megaf 1561.patch:78702: space before tab in indent. 18:59 Megaf settings->setDefault("freetype", "false"); 18:59 Megaf warning: 1 line adds whitespace errors. 18:59 sapier I did formspec fixes myself but those would've required all formspecs to be fixed. This version doesn't. 18:59 Megaf sometimes I get these 1 line adds whitespace errors... 19:00 sapier Thanks, I'll check this 19:07 sapier megaf fixed. How do you check things like that megaf? 19:08 Megaf sapier: wget https://github.com/minetest/minetest/pull/1561.patch (add .patch after pull/). then git apply 1561.patch 19:09 Megaf sapier: git will tell you that 19:09 sapier why doesn't tell git this on checkin? :-) 19:09 sapier https://github.com/minetest/minetest/pull/1879 why do ppl think hardcoding things improves performance? 19:10 Megaf sapier: he actually tested the performance 19:10 sapier for a lot of usages of setFastBool to last element of his if/else cascade? 19:11 sapier reading may be faster but setFastBool is most likely not (in any case) 19:12 sapier and it's more then ugly to place a temporary variable store within a class managing permanent settings 19:12 Megaf hm 19:12 Megaf anyway, I can't give any opinion there, 19:12 Megaf since I don't know that stuff :P 19:13 sapier actually what he added is a settings cache ... a good idea indeed but implementation isn't 19:13 Megaf I know that I don't like NullDriver high on this list. http://i.imgur.com/nK41xkP.jpg 19:14 sapier where exactly should I look at? 19:14 Megaf 7 item on the left list 19:15 Megaf CNullDriver 19:15 kahrl sapier: https://github.com/mx4492/demystifying-whitespace-in-git 19:16 sapier if I don't remember wrong null driver is used for some texture operations 19:16 sapier ahh I see I have git 1.7 19:17 Megaf sapier: hm, I was thinking about removing the null driver and software and directx driver of minetest 19:17 Megaf leaving only OpenGL 19:17 sapier no 19:17 kahrl CNullDriver is just the base class for the other driver classes 19:18 sapier ppl tend to write opengl only code if it looks like minetest only supports opengl but that's wrong e.g. for android we needed ogles ... this would've been way more hard if we had a lot of opengl dependencys 19:19 kahrl also some windows video card drivers don't support OpenGL well but DirectX works 19:20 sapier and directx is more performant on windows 19:20 sapier is faster 19:20 kahrl not sure, I think it depends on the driver 19:20 sapier true I forgot the "in most cases" :-) 19:21 sapier what are thos malloc calls from? 19:21 kahrl lua? 19:22 sapier that's ugly 19:22 kahrl why 19:22 sapier malloc is a quite heavy operation 19:23 sapier and if it's really in lua we can't change anything about it 19:24 sapier same for new/delete of course 19:25 kahrl sapier: actually you can: lua_setallocf 19:26 Calinou what is the debugger Zeno- is using on the screenshot? http://i.imgur.com/nK41xkP.jpg 19:26 kahrl but I doubt it's worth it unless someone writes a really good allocator 19:26 sapier well I don't think we can implement malloc faster then os, the better option would be avoiding allocating/freeing memory 19:27 sapier we'd have to know exactly how malloc is used to write a better allocator ... and hope it's usage never changes ;-) 19:27 Megaf sapier: what should I get/have/notice with the proper font handling? 19:27 sapier formspec fonts scale 19:27 PenguinDad Calinou: callgrind afaik 19:27 sapier if a formspec is shown twice the size as before the fonts are twice a size too 19:28 sapier for non freetype fonts the "twice" is just a rough number, there are only some fonts provided and best match is used 19:30 sapier you should check things like console too megaf 19:37 Megaf sapier: it changed so many things that I didnt know where to look 19:37 Megaf overall fonts look much better now 19:37 sapier that's been the intention, if you don't find any "no go" changes it's most likely fine 19:51 Megaf https://github.com/minetest/minetest/pull/1561#issuecomment-64924467 19:56 kahrl sapier: I see that Hud doesn't update its font and text height and simply uses the ones passed to its constructor 19:57 kahrl instead of calling the font engine 19:57 kahrl is that intentional? 19:58 sapier it's intentional, there've been ppl argueing about their hud scaling to screensize so it's been disabled 19:59 sapier wait only font's don't scale? 19:59 sapier let me have a look at it, it's some time since that patch was written 19:59 Krock fonts never scaled until now except you changed it in minetest.conf 20:00 sapier well maybe the hud just was forgotten 20:00 sapier kahrl where's the cunstructor you're talking about 20:00 sapier ? 20:01 sapier ahh there 20:23 sapier kahrl better this way? 20:24 kahrl sapier: looks good 20:24 kahrl (but I haven't tested it) 20:24 sapier I don't know if there's any visible difference too :) 20:24 sapier it may be a little bit slower but not sure 20:25 kahrl btw for consistency: rename glb_fontengine to g_fontengine 20:26 sapier :-) ok ... where's that style written again? ;-) 20:27 kahrl well basically every global is named like that 20:27 kahrl apparently it's not mentioned on the wiki 20:27 kahrl wait, it is 20:27 sapier really? for what I remember globals are usually not even marked as globals :) 20:27 kahrl http://dev.minetest.net/Code_style_guidelines second-to-last point 20:28 kahrl sapier: see e.g. main.h 20:28 kahrl if you find some that aren't marked, they should be fixed imo 20:28 sapier well no need to discuss about it wasting more time then actual work to do ;-) 20:29 kahrl true :) 20:30 sapier but I'm not gona change it a third time ;-) 20:30 sapier fixed 20:35 sapier https://github.com/minetest/minetest/pull/1768 shadowninja I don't really understand what this feature is supposed to do can you explain it to me? 20:45 Jeija What is m_nothing_to_send_pause_timer in clientiface.cpp good for? I deleted it and didn't experience any disadvantage, only a remarkable increase in mesecons performance. 20:46 Jeija That is because minetest was usually in the process of waiting until that 2 second timer elapsed and therefore didn't send mesecon-modified blocks to the clients. 20:46 Jeija Practical demonstration: https://www.youtube.com/watch?v=B0YwfGVp3o8 20:53 sapier for what I remember it's been meant to reduce server cpu usage if nothing important is to update 20:53 sapier of course that doesn't fit to something like mesecons 20:54 sapier minetest wasn't meant for the sort of "dynamic terrain" mesecons introduce 20:54 Jeija Ah ok, would a lower value also be sufficient for that? Because I guess checking for those blocks is quite CPU-intensive... 20:54 Jeija Like instead of waiting 2 seconds wait like 0.4 seconds 20:55 sapier I don't think the 2 second value had some special meaning, but it's hard to predict what will change 21:29 celeron55 what the hell 21:29 sapier can you be a little bit more precise celeron? ;-) 21:29 celeron55 i have so far gotten two emails about i think buildcraft, altough the first one didn't mention which android version of minetest they were playing 21:30 celeron55 like, users asking stuff 21:30 celeron55 how do they get my email, why is it included in those 21:30 sapier ask you about a build neither approoved nor wantes by us? :) 21:30 celeron55 well whatever, i guess i'll just laugh at these and do nothing 21:31 celeron55 or maybe i should send them advertisements 21:31 celeron55 they sure like those i bet 21:32 sapier for sure yes :-) 21:32 celeron55 yes, i hope i get as many of these as possible and then i'll just cross-promote my other projects to them 21:32 sapier or you do answer them and attach a nice bill onto that message :-) 21:32 celeron55 altough, i don't like shitty users like this... and they don't even have money, otherwise they'd be playing MC PE 21:33 sapier come on our own android port ain't that bad :-) 21:33 celeron55 well it's bad, but it doesn't have advertisements 21:34 sapier I only said it's not "that" bad ;-) 21:34 kahrl we could publish a build with advertisements and only link buildcraft users to it 21:34 sapier hopefully the scalable fonts will make android port a good deal better 21:37 celeron55 (also, i meant that i got these within one week from each other) 21:37 celeron55 (forgot that word) 22:58 Megaf sfan5: is the NDK still needed to compile Minetest for Android? 22:59 Megaf sfan5: And I believe this needs updating http://dev.minetest.net/Android 23:01 ShadowNinja Megaf: Yes, since Minetest is native (C++) code.