Time Nick Message 00:34 proller celeron55, 1cc1b93e65d6d88e20786383cb09caf22a8c6494 - wat? wider_unknown_count always 0 00:35 proller -> active_object_count_wider always = 0 00:37 proller ups, active_object_count_wider not 0 03:30 jojoa1997 VaneßaE here? 03:31 VanessaE yes but last I knew my name was spelled with a double-s, not a sharp-sz. :) 08:48 sapier https://github.com/minetest/minetest/pull/890 minetest sometimes crashes completely if a lua exception occurs 10:55 sapier https://github.com/minetest/minetest/pull/891 add backtrace to error function as that stack is gone after lua_pcall finished 10:56 kahrl wow, you're just pumping out the pull requests :P 10:57 sapier I push what I fix ;-) so it's up to you to decide and merge 10:57 sapier "you" meaning "core developers" 10:58 BlockMen btw merging, how about #849? 10:58 sapier I'm currently preparing mobf for a new release so mt is testet as sideeffect ;-) 11:01 sapier is hotbar.png automaticaly replaced by e.g. texture pack ones? 11:01 BlockMen AFAIK yes 11:05 sapier good :-) 11:11 proller https://github.com/proller/minetest/tree/next - lot of unmerged to master pulls 11:51 kahrl sapier: about #891, why only in builtin.lua, not in mainmenu.lua? 11:52 sapier hmm I guess I missed that one 11:53 sapier did you know we're drawing a full selection box even if only top is visible? ... if you mak it anaglyh you're gonna see the submerged box :-) 11:53 kahrl also I suspect it causes the backtrace to be printed twice if error() is called during the initial load of the lua file 11:53 sapier maybe but twice doesn't harm none does 11:53 kahrl because of loadScript_ErrorHandler 11:54 kahrl well we could maybe get rid of loadScript_ErrorHandler 11:54 sapier if it crashes without calling error() you'll loose the backtrace completely 11:55 sapier as far as I understood error() is a special case because it removes stack 11:55 kahrl I guess the cleanest solution would be to pass an error handler to every lua_pcall 11:55 kahrl but I don't know about the overhead 11:55 sapier I already tried that 11:56 sapier the error handler seems to be required to be on stack ... but pushing something on stack prior doing a pcall changes stack offsets for called functions ;-) 11:56 kahrl true, you have to be careful 11:57 sapier I although tried to pass the int directly while error handler beeing somewhere else on stack but that did fail too 11:58 sapier doing the backtrace on error seemed to be the best option without rewriting full api ... but if you find a better solution feel free to do so ;-) 12:01 sapier major reason for doing this was serialize crashing and I didn't have any idea where it was called from 12:02 kahrl imho if you put this into mainmenu.lua as well it's good enough to be merged 12:03 sapier done 12:05 sapier lol ... I wonder why selection box anaglyph colors are inversed to everything else 12:05 kahrl hmm, very similar commits, very similar commit hashes 12:05 * kahrl suspects the hash algorithm might not be very good :P 12:06 sapier maybe some padding done due to small commit size :-) 12:07 sapier thats crap why is my selection box colored inverse to nodes ?!?! 12:07 kahrl screenshot? 12:10 sapier http://picpaste.de/pics/anaglyph_selectionbox-Qjt7gP1X.1377432610.png 12:10 sapier as you can see the line around the red disortion is cyan 12:11 sapier ok wait it's a black border so this might be even correct 12:15 kahrl why can't I wget that picture? 12:15 sapier I don't know 12:16 sapier https://github.com/minetest/minetest/pull/893 fixes anaglyph selection box 12:16 kahrl If I use FFFFUUUUU as the user agent I can 12:17 sapier :-) 12:18 sapier 893 fixes a major cause for headaches ;-) 12:19 sapier ok I'm back to 2 digit pull requests now I'm gonna have a break ;-) 12:32 proller making dev branch: https://github.com/proller/minetest/blob/next_tools/util/next/next.pl#L34 12:46 Calinou /home/calinou/_VCS/minetest/po/uk/minetest.po:55: Section « msgstr » absente 12:46 Calinou when compiling 12:49 thexyz oh 12:49 thexyz great 12:51 thexyz this is precisely why you should use merge 13:01 thexyz Calinou: fixed 13:02 Calinou works 13:02 Calinou thanks 13:07 thexyz oh great, weblate is borked now 13:09 thexyz it'll be down for some time 13:16 sapier is there any reason why entites are drawn in larger distance than nodes? :-) 13:36 ironzorg https://boards.4chan.org/g/res/36270769 13:37 Calinou how is that relevant to Minetest development 13:40 ironzorg k 14:10 thexyz so, guys, why is https://github.com/minetest/minetest/issues/763 not fixed? 14:10 thexyz I mean, it's been in official minetest build for 3 months and there seems to be no bugreports related to it 14:10 thexyz moreover, without the patch applied the game won't even start 14:11 PilzAdam thexyz, I already told you a couple of times to push it to master already 14:12 BlockMen thexyz, what do you say to #849 ? 14:14 thexyz really? I must've been sleeping, I do that sometimes 14:14 PilzAdam IIRC you replied that it might cause bugs 14:15 thexyz oh, right 14:15 PilzAdam and then I told _you_ that its in the official builds for some time and there werent any bug reports ;-) 14:19 thexyz https://gist.github.com/xyzz/f2f1b7121212eb469ce6 14:20 thexyz PilzAdam: ^ 14:21 PilzAdam is a setting really needed here? 14:23 thexyz Specifies if the device should use high precision FPU setting. 14:23 thexyz This is only relevant for DirectX Devices, which switch to low FPU precision by default for performance reasons. However, this may lead to problems with the other computations of the application. In this case setting this flag to true should help 14:23 thexyz on the expense of performance loss, though. Default value: false 14:24 thexyz what if they want performance 14:24 thexyz I dunno 14:24 PilzAdam add a setting for it if you want, I dont care 14:27 thexyz okay 14:31 kahrl should we add a label "halfapproved" to pull requests, to mark ones that one dev approved but need another to be merged? 14:32 kahrl or "semiapproved" or whatever 14:32 sapier halfapproved sounds odd ... what about "final_approvment_pending" 14:32 sapier ok too long 14:33 BlockMen "dev_checked" 14:37 kahrl then when a dev is bored, he could announce "going to merge these dev_checked pull requests: #ABC, #DEF, #GHI", wait a few minutes and then do it 14:38 kahrl of course only if those were checked by a different dev 14:38 sapier :-) just wanted to mention that ... "ready_to_merge" could be another option ... I know it shouldn't happen devs approve but don't merge ... but it happens 14:43 PilzAdam thexyz, can I run and push updatepo.sh now? 14:52 thexyz yeah 14:57 PilzAdam done 14:59 BlockMen devs, what about #849? 14:59 BlockMen can it be merged? 15:00 PilzAdam yes 15:00 PilzAdam if someone else agrees 15:09 thexyz updated weblate as well 15:16 hmmmm #849 looks good, but it could be even better if you were to cache the result of whether or not the hotbar image is present. i already do this for the crosshair image. 15:16 hmmmm i hope everybody realizes that TextureSource::isKnownSourceImage() is not free 15:17 hmmmm especially when it's being called hundreds upon hundreds of times per second when it doesn't need to 15:21 BlockMen hmmmm, you mean like here? -> https://github.com/BlockMen/minetest/blob/master/src/hud.cpp#L66 15:21 hmmmm yes 15:23 BlockMen ok, gonna change that 15:23 hmmmm thanks 15:25 hmmmm oh while you're at it, can you add a space before "of hud element ID " on line 225? 15:33 BlockMen sure 15:41 kahrl should I push the shapeless crafting fix from yesterday? 15:42 kahrl https://gist.github.com/kahrl/6330631 15:48 BlockMen done and rebased #849 15:50 hmmmm err, i meant to say to add a space inside of the quotes 15:51 hmmmm so it doesn't do "ignoring drawform listof hud element id" 15:52 hmmmm also i think you forgot to add hud.h 15:58 BlockMen oh, damn 16:13 PilzAdam kahrl, of course! 16:14 kahrl done 16:43 BlockMen hmmmm, fixed 16:44 hmmmm yup, looks good 16:46 kahrl BlockMen: looks good to me as well, I'll merge it 16:50 BlockMen ok, thx 17:56 * Taoki[laptop] wonders if there's any news on the colored fog & parallax mapping shader. Mostly if theye're in and finished now 18:09 BlockMen gettext ignores statements with #,fuzzy. does someone know a solution for that? 18:10 PilzAdam translate the string properly and remove the fuzz flag? 18:11 PilzAdam +y 18:15 BlockMen for every language? 18:16 PilzAdam of course 18:17 BlockMen ok, just give me a moment to learn all those :P 18:32 thexyz BlockMen: there's `--use-fuzzy` flag for `msgfmt` 18:32 thexyz which does what you want 18:36 BlockMen thexyz, thx 18:36 BlockMen where do i add? 18:36 BlockMen *the flag 18:39 thexyz BlockMen: src/CMakeLists.txt, line 582 before "-o" 18:46 BlockMen thexyz, thx. works 18:47 thexyz great 20:23 kahrl thexyz: I can do the sha1 filenames, but is there any compatibility support needed? 20:23 thexyz don't think it's crucial 20:24 kahrl allowing multiple servers will be tougher, because 1) how to set them in the config; 2) I'm not sure if the current httpfetch code does keepalive well when there are different servers 20:25 kahrl (others: we are talking about remote_media) 20:25 VanessaE comma-separated list, or just put multiple remote_media entries maybe. 20:26 kahrl the former would work because you can %-escape commas 20:26 kahrl the latter probably wouldn't because of how the settings parser works 20:26 thexyz or maybe we need a way to specify remote media server on client? 20:27 thexyz most of data will be hosted at minetest-cdn.net and missing files can be served from other remote_media server specified in server's config 20:27 thexyz that way it'll be faster even for servers without remote media setting 20:28 thexyz so 1) client tries to fetch data from global remote media server (specified in its config) 2) if not found then server's remote_media 3) if not found then the usual slow way 20:29 kahrl that might add a lot of useless requests if some server uses a game that isn't represented on minetest-cdn.net 20:30 thexyz true… 20:31 VanessaE I don't like the idea of defaulting to minetest-cdn.net 20:31 thexyz why? 20:31 VanessaE no offense, but I prefer to maintain control of what files my server sends to its clients 20:31 thexyz um 20:31 kahrl VanessaE: I don't either, I thought the idea was to allow the client to specify it 20:31 VanessaE (I'm kinda anal like that) 20:31 thexyz I assume client verifies sha1 20:31 VanessaE well.. 20:31 VanessaE mm 20:32 VanessaE I suppose.. 20:32 thexyz yes? 20:32 VanessaE *shakes head* 20:32 VanessaE I guess if it's a client-side setting, that's fine 20:32 thexyz well there's a certain problem with client-side setting as kahrl pointed 20:33 VanessaE but to have it enabled and defaulted to that server (or any other) feels wrong 20:33 thexyz okay, why? 20:33 VanessaE thexyz: call it paranoia. 20:33 thexyz no, no, explain 20:33 thexyz client gets exactly the same files 20:33 VanessaE well for starters, there's a lag problem - what if my server's 10x faster than cdn? 20:34 VanessaE but that's not the major issue 20:34 thexyz yeah, that's possible 20:34 thexyz but what's the major one? 20:34 kahrl also, there's a privacy problem 20:34 VanessaE cdn is guaranteed never to be always up-to-date 20:34 VanessaE because it would be pretty much impossible 20:34 kahrl why should minetest-cdn.net get to know whenever someone is playing on a non-public server without a remote_media server? 20:34 VanessaE which means no matter what, clients are gonna hit whatever the remote_media URL is that the server they're connecting to has specified 20:35 thexyz okay, let's say we have server-side remote_media setting 20:35 VanessaE that much is true too, though I think that's less of a problem (i.e. we can probably trust the admin thereof) 20:36 thexyz client-side one was stupid, I agree 20:36 VanessaE client hits cdn first, sees "ok, half of these files are out of date". Downloads what it can, then hits the configured URL for the rest, then the server itself 20:36 thexyz especially for LAN servers and whatnot 20:36 VanessaE three connections instead of two is guaranteed to slow stuff down some 20:36 kahrl given the current code it would even try to contact minetest-cdn.net when starting a singleplayer world 20:36 VanessaE (assuming they're in series) 20:36 thexyz VanessaE: not "out of date" 20:36 thexyz they either exist or not 20:36 VanessaE well if they don't exist, they're out of date :) 20:37 thexyz no, if they don't exist then they don't 20:37 VanessaE you get my point 20:37 thexyz pretty much 20:37 VanessaE "out of date" relative to the mod that supplied the media originally 20:44 thexyz kahrl: regarding this privacy issue, isn't all media cached on client? 20:46 kahrl true, although not the first time 20:47 kahrl anyway, what would this setting solve that multiple remote_media servers wouldn't? 20:52 thexyz kahrl: it will work for any server, even for those without remote_media set; anyway, I too think there're too many flaws in it 21:25 kahrl thexyz: does this look sensible? https://gist.github.com/kahrl/6336428 21:35 thexyz kahrl: will this work correctly? I'm not sure how featchrequest works, does it only fetch if media file is not fetched yet? 21:36 thexyz and why is timeout = 0? 21:41 kahrl timeout = 0 because otherwise big files would end up hitting the timeout 21:41 kahrl it is setting the connect timeout to 0 as well though, that's a bug and I'll fix it 21:42 kahrl the code only fetches the files that aren't in the cache because file_requests only contains those 21:42 kahrl I haven't changed it to use the sha1 yet 21:43 harrison http://youtu.be/KEMuf86NuPU 21:44 kahrl thexyz: is the support for multiple servers mainly meant for reliability, or for load distribution? 21:45 thexyz no and no; the problem is when main server doesn't have some media 21:46 thexyz (I assume server owners will use minetest-cdn as primary media source and then set up their media server for files that are not present) 21:46 kahrl well my code doesn't work in that case 21:47 kahrl as in, it falls back to the conventional method instead of another remote_media server 21:47 kahrl falling back onto another server means a whole lot more code... 21:47 thexyz oh, right, I'm falling asleep already 21:51 thexyz how about make HTTPFetchRequest's url a vector of URLs and then make it cycle through them until success? 21:51 kahrl I don't want to complicate httpfetch at this point 21:52 kahrl it's already complicated enough :P 21:52 kahrl I think I'll make a helper class for it in client.cpp though 21:53 thexyz okay 22:03 PilzAdam the inventory location "player:" seems to only work if you are the player with 22:06 kahrl PilzAdam, in formspecs? 22:06 PilzAdam yes 22:06 kahrl yeah because its contents aren't sent 22:07 PilzAdam is there a way to fix that? 22:07 kahrl probably but I don't know how 22:08 Exio4 about the "main content server", would be having a list of "known sha1(s)" something that could work? 22:16 thexyz Exio4: elaborate 22:17 Exio4 minetest-cdn having a "knsha1.txt" that lists the sha1 of the files that are there, and remote_media having one too with it, in case some file with X sha1 needed, check where you can get it 22:18 Exio4 i don't think i explained properly 22:18 Exio4 but more like poorly :P 22:18 thexyz how huge would that list be? 22:18 kahrl Exio4, thexyz: a bloom filter could be useful 22:19 Exio4 thexyz: 256kb? 128kb? 22:20 thexyz Exio4: I dunno, I only know that sha1 hash is 20 bytes 22:20 Exio4 20+separator (\n) 22:20 thexyz Exio4: also, do you want to download this .txt every time client tries to connect to server? 22:20 PilzAdam kahrl, I think the problem is that in client.cpp:2584 m_env.getPlayer() is called 22:20 thexyz Exio4: no need for separator 22:20 Exio4 true 22:20 PilzAdam but the client doesnt have a list of players, only the player himself is in the list 22:21 kahrl PilzAdam: doesn't he? how does nick completion work? 22:21 PilzAdam it gets the playernames when creating the GenericCAO and saves them in env::m_connected_players (or so) 22:23 thexyz client could as well POST this list of sha1 hashes to server, then server sends back list of hashes it has 22:25 kahrl thexyz: what about the bloom filter idea? 22:27 thexyz kahrl: I don't get how's that related 22:29 kahrl just put the sha1's into a bloom filter, to test if a given sha1 is available on the server check the bloom filter 22:29 kahrl this has a chance to give false positives but the space required for it is much smaller than a full list of the sha1's 22:30 kahrl so instead of requesting full_list_of_sha1s.txt you request sha1_bloom_filter.txt 22:30 kahrl probably not .txt though 22:32 thexyz that's good but I think we don't really need it 22:33 thexyz (for now) 22:33 thexyz do you think something is wrong with one additional POST request? 22:34 kahrl I just think we shouldn't waste bandwidth when we don't have to 22:34 kahrl not everyone has unlimited bandwidth on their servers 22:34 thexyz well let's calculate it then 22:35 thexyz 10000 sha1 hashes is 0.19 megabytes 22:36 thexyz in the worst case client sends 0.19 megabytes, receives 0.19 megabytes and then downloads 10000 files 22:37 thexyz and the server has already sent list of those hashes anyway 22:37 kahrl well okay 22:47 VanessaE I'm noticing an odd glitch - my pointer keep re-appearing after a few seconds while playing the game 22:47 VanessaE and jitters around when I move/turn 22:47 VanessaE keeps* 23:56 VanessaE seeing a variable performance improvement in Linux systems by removing the FPS display from the window title bar. Can we please move that to the F5 display (put it next to the 'drawtime' figure) ? 23:57 VanessaE disabling the code that puts it in the title bar gives this guy on #minetest at least a 2x improvement. I see a 10 to 20 percent improvement myself.