Minetest logo

IRC log for #minetest-dev, 2013-12-25

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

All times shown according to UTC.

Time Nick Message
00:09 Exio4 what do you think about https://github.com/EXio4/minetest/commit/29f21b2?
00:10 PilzAdam Exio4, builtin functions shouldnt change in creative mode
00:10 PilzAdam its up to the games to do that
00:12 Exio4 i saw https://github.com/minetest/minetes​t/blob/master/builtin/item.lua#L411 and thought this would fit too :P
00:12 PilzAdam that should be removed
00:13 sapier I'm with pa don't disable half of api by creative mode ... yes I know "half of api" is a little bit more then actually done ;-)
00:31 iqualfragile joined #minetest-dev
00:33 kaeza joined #minetest-dev
01:13 kaeza joined #minetest-dev
01:42 hmmmm hmm sapier, do the reliable packets have a checksum?
01:43 hmmmm oh god, I noticed you made a "PeerHelper" smart pointer
01:43 hmmmm didn't we agree to not use smart pointers?
01:46 sapier yes but there wasn't a other sane way to do this
01:46 khonkhortisan joined #minetest-dev
01:47 sapier I thought a lot about other options none was even close to usability of that damn smart pointer
01:48 sapier but if you have a good idea how to this without a smart pointer you're welcome
01:48 sapier no I can't add checksums as it's supposed to be a 100% compatible fix
01:49 hmmmm no no i was just wondering if they already did or not
01:49 sapier doesn't udp itself have a checksum?
01:52 sapier but keeping compatibility doesn't mean a new server can fix client side bugs ... yet it seems to reduce their occurance a little bit
01:52 hmmmm yeah udp does have a checksum
01:53 hmmmm it's funny, because I know of several games that do the whole "reliable UDP" thing and they have checksums too
01:54 sapier yes I tend to interpret this as not knowing what they're using ... or not beeing interested in
01:54 hmmmm commercial games
01:55 sapier so later one ... someone specified to have a checksum in protocol so programmer implemented it
01:57 sapier sometimes it's annoying if programmers do what they're told to do without spending a single thought on it :-)
01:58 domtron joined #minetest-dev
02:03 kahrl isn't the udp checksum kind of optional?
02:04 kahrl by that I mean the sender can set it to zero, and then the receiver won't check it
02:04 sapier optional as you can set it to 0 if you don't want it
02:04 sapier yes
02:04 domtron joined #minetest-dev
02:04 kahrl how do the different OSes handle that?
02:05 sapier no idea
02:06 sapier I wonder how to reduce oversending of udp packets without adding too much delay
02:22 kaeza joined #minetest-dev
02:25 domtron joined #minetest-dev
02:32 PilzAdam moving the FPS out of the window title gives a speedud of 50 to 100% on some window managers
02:32 PilzAdam *speedup
02:32 VanessaE didn't I say this like three months ago?
02:32 sapier I thought that was clear for months now?
02:33 PilzAdam why isnt it done yet?
02:33 VanessaE PilzAdam: because random?
02:33 VanessaE NIH?
02:33 VanessaE no one cares?
02:33 VanessaE not "our" fault?
02:33 VanessaE etc
02:33 sapier guess classical small fix everyone assuming someone else will merge it
02:33 VanessaE I said move it to F5.  everyone else said fuck off.
02:34 VanessaE (it's in F6 anyway so wtf?)
02:34 PilzAdam VanessaE, IIRC you said something about a couple of FPS, but not that much of a speedup
02:34 sapier no VanessaE always made clear it's a huge speedup (at least that's what I remember)
02:34 VanessaE PilzAdam: yeah, for me the speedup is very small.  for the person who found it, doing that change doubled his fps.
02:35 VanessaE I forget who it was now who discovered it
02:35 PilzAdam and you use xfce too?
02:35 VanessaE yes
02:35 sapier ok maybe my memory isn't as well for this point
02:35 NakedFury joined #minetest-dev
02:35 VanessaE the other person uses...eh, I forget what
02:36 domtron joined #minetest-dev
02:36 VanessaE might have been KDE
02:36 VanessaE something relatively heavy I guess
02:36 sapier hmm not much out there windowmaker/gnustep/afterstep/kde​/cinnamon/gnome2/gnome3/unity...
02:36 PilzAdam no, it doesnt have any effect on KDE
02:37 VanessaE damn it, who was it and what did he use? *grumble*
02:37 VanessaE lemme check my logs
02:37 PilzAdam the one that reported it just to me uses xfce
02:37 sapier maybe it's related to a specific xfce version too, yet I don't understand why we need to show fps in titlebar ;-)
02:39 sapier PA if you have the fix ready I agree to push it ;-)
02:39 VanessaE sapier: neither do I
02:39 PilzAdam sapier, I dont have it
02:39 VanessaE the name of the program is "minetest", not "minetest 27 fps"
02:40 sapier if we don't do it now we'll forget about it again does anyone have that fix?
02:40 VanessaE FreeNode-#minetest.log:Nov 06 2012 13:58:27 <ruskie>suggestion for the linux build... give it a "minetest" window class so that it can actually be matched by WMs because the title can't be used due to the FPS counter in there
02:40 VanessaE that doesn't seem right. seems too recent.
02:45 VanessaE well I can't find the original reference but plenty of chatter here and there.  so, let's just take it the fuck out of the title bar a;ready?
02:45 VanessaE already*
02:46 VanessaE I have a simple patch that takes it out but it's not "right".
02:46 sapier can you post it to gist?
02:46 VanessaE http://pastebin.ubuntu.com/6632217/
02:47 VanessaE I don't recall whose code this is though.
02:47 VanessaE so don't attribute it to me :)
02:47 sapier what's wrong with it?
02:47 VanessaE it leaves a couple of defined-but-not-used-variable compiler warnings, but other than that, it works fine.
02:48 sapier I guess I can handle this
02:49 rambomedic joined #minetest-dev
02:49 VanessaE I'd suggest moving the fps that's in the F6 display over to F5 while you're at it
02:49 VanessaE or maybe not
02:49 VanessaE idk
02:49 rambomedic while you're at what?
02:49 VanessaE seems like it would be less easy to get lost in the f5 display :)
02:49 VanessaE rambomedic: he's ripping minetest all to hell and back ;)
02:49 Exio4 i would add it next to the coords
02:49 rambomedic who? what? I just joined VanessaE
02:50 VanessaE haha
02:50 VanessaE rambomedic: nevermind, sapier's just adding a small patch to the engine
02:50 sapier sad to tell vanessae but that code isn't even close to the location where titlebar fps is removed ;-)
02:50 VanessaE wanna bet?
02:50 Exio4 DOGEs to the rescue
02:50 rambomedic what does the patch do?
02:51 sapier I don't bet for good reasons ;-)
02:51 VanessaE sapier: don't ask me how it does it, but somehow it still works :)
02:51 PilzAdam sapier, Im doing it currently...
02:51 sapier ok pa then I wont start to do same thing
02:52 rambomedic ugh joining IRC mid-conversation and without context is so frustrating
02:52 sapier rambomedic: there's a log ;-P
02:52 kaeza that's why there are logs on the topic :)
02:52 VanessaE rambomedic: http://irc.minetest.ru/minet​est-dev/2013-12-25#i_3514905
02:52 * rambomedic curls up in chiar
02:52 sapier I guess he got it ;-)
02:52 * rambomedic doesn't know how to spell chiar
02:52 VanessaE lol
02:53 rambomedic damnit I messed it up twice
02:53 sapier try chair
02:53 rambomedic (I know how to spell it, I just messed up)
02:53 PilzAdam https://github.com/PilzAdam/minetest/commi​t/4b31a38ae9b2d2019ddf75d09c1fdfd67f512067
02:53 sapier sorry ;-)
02:54 PilzAdam rambomedic, shit-chat goes to #minetest
02:54 VanessaE PilzAdam: ah, it was that extra static at 3385 I forgot to crop out wasn't it.
02:55 sapier https://github.com/PilzAdam/minetest/commi​t/4b31a38ae9b2d2019ddf75d09c1fdfd67f512067 agreed
02:55 VanessaE this seems good to me.
02:56 iqualfragile gdamn, why does github not support ipv6…
02:57 PilzAdam "Double performance" >> changelog.txt
02:57 iqualfragile :D
02:57 PilzAdam + (by PilzAdam)
02:58 VanessaE haha
03:02 domtron joined #minetest-dev
03:04 Miner_48er joined #minetest-dev
03:05 rambomedic hmmmm: can we merge https://github.com/minetest/minetest_game/pull/220
03:06 hmmmm who's 'we'
03:07 rambomedic america
03:07 rambomedic *you
03:07 VanessaE the current batch of textures should probably be given another go with optipng and/or advpng rather than directly accepting that pull, though
03:08 * kaeza also pokes #minetest-dev with https://github.com/minetest/minetest/pull/1060 ...again
03:08 VanessaE I'm pretty sure there have been one or two changes since that pull + a few new ones, no?
03:08 hmmmm huh??!
03:08 rambomedic let me check
03:09 hmmmm I don't get it, how is that enum element inconsistent
03:09 VanessaE what, huh?
03:09 VanessaE oh
03:09 hmmmm and isn't that a bit too trivial for a pull request
03:09 kaeza hmmmm, all elements follow the same name as the HUD element def
03:10 kaeza plus lua_api.txt documents "position" (not "pos") for hud_change
03:10 hmmmm ah
03:12 kahrl I guess we can close 1069 now?
03:13 kahrl #1069
03:13 ShadowBot https://github.com/minetest/minetest/issues/1069
03:13 rambomedic VanessaE: just checked and confirmed, no textures have been added/changed since my pull requ
03:13 VanessaE ok then
03:13 VanessaE I'm okay with it, provided that it doesn't break the PNG loader.
03:13 VanessaE there's a potential glitch with Windows 8.1 and optipng'd images
03:14 rambomedic link?
03:14 VanessaE email.
03:14 VanessaE someone reported it to me after an HDX update.
03:14 rambomedic !g glitch optipng windows 8.1
03:14 ShadowBot rambomedic: https://github.com/jorgebastida/g​lue/blob/master/docs/optipng.rst
03:14 VanessaE I optipng'd all of that format in all HDX sizes.
03:14 rambomedic do clarifyt
03:15 VanessaE the person reported that win 8.1 + minetest 0.4.8 (presumed stable) aborts with a bad alloc error (but he gave no log output)
03:15 VanessaE says that some images are not recognized as images at all (let alone PNGs) but they all work perfectly for me.
03:15 rambomedic and before compressing the textures it doesn't?
03:15 VanessaE (but then, I use Linux)
03:15 VanessaE supposedly so, yeah
03:16 VanessaE so there may be something about the png format that optipng uses that win 8.1 can't handle, so be wary of it?
03:16 VanessaE I can't test this, so I can only report what's reported to me.
03:17 rambomedic then I'm assuming it may be a bug with windows' png library? ask him for a log
03:17 VanessaE that's what I told him.
03:17 VanessaE can anyone else confirm that?
03:31 Miner_48er joined #minetest-dev
03:36 domtron joined #minetest-dev
03:45 us_0gb joined #minetest-dev
04:03 domtron joined #minetest-dev
04:37 domtron joined #minetest-dev
04:44 hmmmm oh snap
04:44 hmmmm we need to get ready to release
04:44 hmmmm thexyz around?
04:44 hmmmm anybody who can do some windows builds?  debian repo updates?
04:45 sapier I can but I'm not sure if we didn't add a bug recently
04:46 hmmmm all of the commits to master have been bugfixes for more than a week
04:46 sapier I can join to my fixed server branch with fixed version (of course) as well as 0.4.8 ... but master seems to be stuck at loading item definitions
04:46 hmmmm hmm
04:46 sapier I'd like to have a closer look to this
04:46 sapier at least try to reproduce it ;-)
04:47 hmmmm if we can fix it within the next, say, 12 hours, I'd say we can do a christmas release
04:47 hmmmm vanessae tried to bisect it but was getting weird results
04:47 sapier it's not even sure to be a core bug so maybe it's not relevant for release
04:47 sapier I know whats cause for beeing stuck at downloading textures
04:47 hmmmm so it's not readily reproducible?
04:48 sapier I need vanessae's server to reproduce on localnet timing doesn't seem to be suitable to cause it
04:48 hmmmm is it readily reproducible for vanessae's server?
04:49 hmmmm I thought vanessae was having problems connecting to other servers though, not to her onw
04:49 hmmmm own*
04:49 VanessaE sapier: I've pulled and am putting your currend HEAD in now
04:49 sapier cheapie seems to be stuck everytime and I had it at least once
04:49 VanessaE hmmmm: only redcrab's 0.4.4 server was a problem.
04:49 VanessaE sapier: my servers are ready to restart with yours now.
04:50 VanessaE hmmmm: and I bisected it to kahrl's httpfetch patch.
04:50 sapier ok you should have way less cpu load now
04:50 sapier kahrl already proposed/pushed some fixes for it
04:51 VanessaE oh ok, didn't try it recently.
04:51 hmmmm what do you guys think about this https://github.com/kwolekr/minetest/commit​/208247f65351dab93c89ede0b4b07f4e915268bf
04:51 hmmmm it's to fix proller's thing
04:51 VanessaE lemme check with current HEAD.
04:51 VanessaE ah yes, it's fine now.
04:51 VanessaE ok, so that's solved.  guess I didn't get the memo :)
04:52 VanessaE sapier: my servers are ready for your tests.
04:52 sapier seems to be fine hmmmm
04:52 hmmmm I think it's just one of those transient errors
04:52 sapier what about cpu load vanessae?
04:52 hmmmm sapier, I mean is there a better way to fix it, not if it's fine or not
04:52 VanessaE oh that's MUCH better
04:52 hmmmm i.e. changing the order in which things are deinitialized
04:52 sapier a better way ... throw an exception fix all destructors on it's way to the top ;-)
04:52 VanessaE CPU loads are all < 3% except for survival, and that's only because I'm actually *on* it.
04:53 VanessaE this is about like I would expect it to be.
04:53 sapier ok lets try master version again
04:53 VanessaE in fact, Nostalgia world is actually the one using the most (about 9%)
04:53 VanessaE ok
04:53 VanessaE Survival shot up to 90% for a moment
04:53 VanessaE it's hanging at about...
04:53 VanessaE well was hanging at 80%-odd.
04:54 sapier and stuck at item definitions :-(
04:54 VanessaE now it's at nearly zippo.
04:54 sapier high cpu load again?
04:54 VanessaE no
04:54 VanessaE "zippo" is a synonym for "zero"
04:54 sapier :-)
04:54 VanessaE it's nearly zero
04:54 VanessaE varying of course
04:55 VanessaE ~0...5...10...5..20....5...  you get the idea
04:55 sapier yes if there's work to do my branch wont hesitate to make your cores to do it ;-) and as I split sending from receiving maxing out 2 cores is possible now
04:56 NakedFury isnt there a way so all the devs can see real time the cpu load of a tested server? like a live video feed just showing all the statistic they would need.
04:57 VanessaE brb
04:57 sapier ok now I can connect with master
04:57 NakedFury merry holiday
04:57 sapier ok not exactly
04:58 VanessaE looks like it crashes.
04:58 VanessaE crashed*
04:59 sapier crashed?
04:59 sapier can you provide a log?
04:59 VanessaE checkin
04:59 VanessaE http://pastebin.ubuntu.com/6632585/
05:00 sapier ok that's not as quick to fix as before
05:00 VanessaE recommendation?
05:01 sapier as master connect doesn't work for some reason ... and I need to get some sleep ... back to master and another try once you've got time again
05:01 VanessaE ok
05:01 VanessaE no prob
05:02 rambomedic serious bug with gnome 3 #1071
05:02 sapier I wonder if this really is a regression in master
05:02 ShadowBot https://github.com/minetest/minetest/issues/1071
05:02 hmmmm you know
05:02 hmmmm I think I should m_con.Stop() in Server::~Server()
05:03 sapier and wait for it
05:04 hmmmm oh
05:04 hmmmm Stop doesn't wait?
05:04 sapier no
05:04 sapier just requests thread to stop
05:04 hmmmm does Kill wait?
05:04 sapier no kill does kill
05:04 hmmmm nonsense, I see a pthread_join in there
05:04 sapier which usually isn't what you want to do
05:04 sapier kill is bad
05:05 sapier hmm
05:05 sapier wait I'll have a look
05:05 hmmmm oh nevermind
05:05 hmmmm pthread_cancel first
05:06 sapier ok kill doesn't wait
05:06 hmmmm it only does that first join if it's not running but started
05:06 sapier yes
05:06 hmmmm right
05:06 sapier but if you kill a running thread it just kills
05:07 hmmmm so at the moment, Server::m_con is simply being killed after ~Server finishes
05:07 hmmmm yes?
05:08 hmmmm so instead of ignoring EBADF what I should do is indeed m_con.Stop(); m_con.Wait();
05:08 hmmmm before deleting the peers
05:08 sapier seems to be yes
05:08 hmmmm that won't work though
05:08 hmmmm I need to cancel out of the select()
05:08 sapier isn't there a timeout?
05:09 hmmmm is there?
05:09 sapier I think so
05:09 sapier but I may mix it up with my fixes there it is
05:09 hmmmm I guess it's settable via the config?
05:11 sapier I can't tell for sure I guess I need to get some sleep prior beeing helpfull again .. sorry
05:11 hmmmm not it's not
05:11 hmmmm it's called with a timeout of 0 * 1000
05:11 hmmmm which if I recall is INFINITE
05:11 hmmmm so like..... what.... signal(SIGUSR1) or something (which won't work on windows anyway, so that adds even more complication)
05:12 sapier is this UDPSocket wait?
05:12 hmmmm yes
05:12 hmmmm Connection::receive -> UDPSocket::WaitData()
05:12 sapier if you call it with 0 there's no wait ... just ask vanessae I maxed out her cpu usage by doing it this way ;-)
05:13 VanessaE lol
05:13 hmmmm erm
05:13 hmmmm the wording says "to effect a poll, the timeout argument should point to a zero-valued timeval structure"
05:13 hmmmm whatever that means
05:14 sapier maybe there's something else modifying behaviour there are some layers pf code till actual select call
05:14 hmmmm I'm confused on the behavior of a select() with a 0-valued timeval structure
05:15 hmmmm because it's clearly what's being passed as the argument to that select call
05:15 rambomedic joined #minetest-dev
05:15 sapier blocking/non-blocking socket?
05:16 sapier if it's non blocking it's always ready to read
05:16 hmmmm it's blocking, is it not?
05:16 sapier I don't know
05:16 sapier I'm just throwing random related thoughts in right now ;-)
05:16 hmmmm how do you not know, you rewrote the entire connection code
05:17 sapier but not the udpsocket ;-)
05:17 hmmmm mm
05:17 sapier that part did surprisingly work well
05:17 * hmmmm searches
05:18 sapier I didn't rewrite connection code I just put it into mixer, puzzled and glued resulting parts together afterwards ;-)
05:19 sapier good night :-)
05:19 hmmmm >ctrl+f O_NONBLOCK
05:19 hmmmm >0 results found.
05:20 sapier so it may be interesting to find out why that socket returns immediatly in my new code
05:21 sapier ... i didn't check that one for multithreading issues yet ... guess I'm gonna do that tomorrow
05:21 sapier left #minetest-dev
05:24 rambomedic serious gnome 3 bug #1071
05:24 ShadowBot https://github.com/minetest/minetest/issues/1071
05:27 Miner_48er joined #minetest-dev
05:29 hmmmm well, to sapier:
05:29 hmmmm the thing is, the way Connection uses UDPSocket is that it always has a zero wait
05:30 hmmmm what that documentation means, "effect a poll" just says that it immediately returns *as if* the socket was nonblocking
05:30 hmmmm so the wait happens elsewhere within Connection, not at all in the UDPSocket
05:30 hmmmm that's why your stuff is driving up the cpu, because you are responsible for waiting
05:31 hmmmm and as for my answer, the timeout is at the mercy of however the hell long the Connection thread decides to wait
05:32 hmmmm it might be a good idea to just simply check EBADF after all... there's no harm in select() failing on that if we're shutting down
05:36 domtron joined #minetest-dev
05:39 Miner_48er joined #minetest-dev
06:15 rambomedic joined #minetest-dev
06:23 OldCoder joined #minetest-dev
07:10 VanessaE joined #minetest-dev
07:27 rambomedic joined #minetest-dev
08:00 darkrose joined #minetest-dev
08:00 darkrose joined #minetest-dev
09:07 john_minetest joined #minetest-dev
09:45 ImQ009 joined #minetest-dev
09:56 VanessaE joined #minetest-dev
10:13 Jordach joined #minetest-dev
10:20 Jordach joined #minetest-dev
10:27 john_cephalopoda joined #minetest-dev
10:35 nore joined #minetest-dev
10:36 proller joined #minetest-dev
10:36 troller joined #minetest-dev
10:40 domtron joined #minetest-dev
10:42 Gethiox joined #minetest-dev
10:43 john_minetest joined #minetest-dev
10:58 sapier joined #minetest-dev
10:59 VanessaE joined #minetest-dev
11:05 nore joined #minetest-dev
11:08 domtron joined #minetest-dev
11:40 domtron joined #minetest-dev
11:56 VanessaE joined #minetest-dev
11:59 VanessaE joined #minetest-dev
12:01 domtron joined #minetest-dev
12:13 NakedFury joined #minetest-dev
12:18 iqualfragile joined #minetest-dev
12:21 VanessaE joined #minetest-dev
12:28 VanessaE joined #minetest-dev
12:40 VanessaE joined #minetest-dev
13:08 john_minetest joined #minetest-dev
13:43 john_minetest Wow... It would be epic to have books you can write to and bookshelves are just like chests for books only. So you can put books in the book shelve and store REAL knowledge in there instead of just decoration.
13:44 john_minetest Is it possible to allow special things only in a chest?
13:56 sapier you can already do this in lua john
13:59 domtron joined #minetest-dev
14:03 jin_xi joined #minetest-dev
14:06 john_minetest Gooood.
14:06 john_minetest I will write a mod.
14:27 domtron joined #minetest-dev
14:36 zat joined #minetest-dev
14:41 john_minetest Lol, I took a lot of code from the books mod.
14:41 kaeza joined #minetest-dev
14:42 ImQ009 joined #minetest-dev
14:53 proller joined #minetest-dev
15:13 sapier john_minetest:  whats "lol" about it? I haven't had a look at it by now
15:14 EvergreenTree joined #minetest-dev
15:17 Zeitgeist_ joined #minetest-dev
15:29 hmmmm joined #minetest-dev
15:30 domtron joined #minetest-dev
15:31 hmmmm release!
15:31 hmmmm also merry christmas
15:36 iqualfragile_ joined #minetest-dev
15:39 kaeza joined #minetest-dev
15:49 domtron joined #minetest-dev
15:55 Gethiox joined #minetest-dev
16:07 john_minetest release? Nothing on the forums, nothing on github.
16:07 proller фтв ыешдд ыргевщцт скфыр умукнцруку
16:08 proller and shutdown crash everywhere
16:14 domtron joined #minetest-dev
16:16 Megaf joined #minetest-dev
16:33 john_minetest oh, yeah. That, too. And startup crash sometimes, too.
16:33 john_minetest left #minetest-dev
16:34 proller 101% for me 8(
16:38 ShadowNinja And the serverlist is broken!
16:39 ShadowNinja (Unless the config changed or something like that)
16:43 domtron joined #minetest-dev
16:56 Gethiox joined #minetest-dev
17:00 domtron joined #minetest-dev
17:12 Gethiox joined #minetest-dev
17:14 us_0gb joined #minetest-dev
17:34 sapier joined #minetest-dev
17:35 sapier ShadowNinja: I can see serverlist, can you be more precise what is broken?
17:39 sapier thexyz did you stumble uppon excessive memory usage due to simultaneous joing of clients on servers with lots of textures yet?
17:47 thexyz sapier: with enet or what?
17:48 sapier it's not a issue with networking protocol but the way minetest transfers texutres
17:49 sapier once a client connects mintest prepares all chunks to be sent queuing them for that particular client
17:49 sapier obviously if that data isn't out by the time next client connects it's going to requeue same data
17:50 sapier if your server has about 170 mb of media data it's gonna blow
17:50 thexyz I didn't test it
17:50 sapier I'd expect any protocol will be affected by this behaviour so we gonna need to fix it anyway
17:50 thexyz also if your server has about 170 mb of data, you plan to host a large number of clients and don't have remote media/httpfetch then it's your fault
17:51 sapier it's not about planing
17:51 thexyz btw, I didn't manage to benchmark your server improvements because every client was eating about 100% of cpu
17:51 sapier it's a dos attack ... a quite simple one
17:51 ShadowNinja sapier: Serverlist announcing I meant.  It works for Vanessa after the fixer commit, but not for me.
17:51 thexyz if you want to fix it then go on, but I don't think this is important at all
17:52 sapier cpu usage is already fixed was missreading of 0 timeout not meaning wait for ever but return immediately
17:52 thexyz a dos attack? since when do you need to dos minetestserver?
17:52 thexyz I mean, just play on it
17:52 thexyz anyway, it's your time so feel free to spend it on fixing this
17:53 thexyz sapier: okay, I'll check it later today once again then
17:53 sapier if you don't have any interest in it I won't fix it because without cooperation it needs to be done twice once for mine and again for future enet versions
17:53 sapier and it's not even worth doing this as of educational purpose
17:54 thexyz future enet? I thought you're not interested in it
17:54 thexyz anyway, you got it correctly -- I'm not interested in fixing this
17:54 sapier as far as I know decision for enet has been made unless there ain't a unfixable flaw in ther
17:54 sapier e
17:55 sapier it doesn't matter if I still think it's wrong or not if the decision is made this way
17:56 thexyz as far as I know the decision was to go with UDP + TCP
17:56 sapier is anyone working at this?
17:57 sapier it's sad we don't even know for sure what has been decided
17:58 sapier I guess that's the real flaw in our development process ;-)
17:58 thexyz I don't know
17:58 thexyz I checked the logs that day
17:59 thexyz http://irc.minetest.ru/minet​est-dev/2013-12-20#i_3505769
17:59 thexyz everybody agrees that we need TCP+UDP but nobody is working on it
18:00 sapier that's really ugly
18:00 sapier problem with tcp+udp is it's even more work then enet
18:00 VanessaE joined #minetest-dev
18:01 Yepoleb joined #minetest-dev
18:01 sapier I guess I could replace the "reliability" code from my aproach quite easy but hard to test that hybrid aproach
18:05 iqualfragile joined #minetest-dev
18:05 thexyz well? I know
18:05 thexyz tell that to them
18:05 rambomedic joined #minetest-dev
18:27 Gethiox joined #minetest-dev
19:42 VanessaE joined #minetest-dev
20:27 john_minetest joined #minetest-dev
20:47 proller joined #minetest-dev
20:58 sapier joined #minetest-dev
21:05 Exio4 new year release?
21:05 john_minetest more probable than christmas release
21:06 john_minetest Although I am not sure what is missing.
21:09 proller coredumps everywhere
21:13 sapier proller could you please stop beeing destructive only and start beeing helpfull?
21:13 proller i can make bt 5 times
21:14 sapier a backtrace isn't enough, you need to describe exactly what you're doing and what version you're using
21:15 hmmmm alright, new year release
21:15 sapier did you try already check out the fix kahrl suggested tonight?
21:15 hmmmm what fix
21:15 hmmmm if it's about the unhandled socket exception on shutdown I've got that covered, stop freaking worrying about it
21:15 sapier or was it hmmmm ... let me look at logs, guess I'll never sort the fill word guy and kahrl :-)
21:16 sapier no offence hmmmm ;-) just joking
21:16 hmmmm sapier, did you see what I said yesterday after you left
21:16 hmmmm I mentioned the reason why your code takes up 100% cpu
21:16 hmmmm you are responsible for all sleeping, UDPSocket::receive() returns instantly
21:17 sapier yes I already added a fix so if there's a explanation for it it's fine
21:18 sapier as I split send from receive I can do a blocking wait using a timeout there
21:19 sapier but back to prollers problem is this your shutdown/logout/leave problem only or is there a reproducable case in game?
21:22 NakedFury joined #minetest-dev
21:23 john_minetest How far has colored chat come in minetest?
21:23 john_minetest Pulled yet or still somewhere in the "we-won't-pull-but-freeminer-has-it"-zone?
21:26 Sokomine colored chat? that sounds fine! i'd love to see that
21:26 VargaD Hi, does freeminer changes merged back to minetest?
21:27 VargaD can I help in merging some of the interesting features?
21:28 iqualfragile VargaD: thexyz stated that he would not put any effort into merging features back into minetest, proller is unable to create clean pull requests so i guess they will not get merged  back
21:29 iqualfragile VargaD: you can therefore help, try creating clean pull requests, but it might be smart to ask if people want that feature in here first.
21:30 VargaD I'm willing to provide clean pull requests one-by-one, but then I ask the feature first, thanks iqualfragile
21:39 djdduty joined #minetest-dev
21:39 john_minetest grblargh
21:39 john_minetest Is colored chat upstream or just freeminer now?
21:40 john_minetest Those are the features that are of interest.
21:40 sapier john_minetest: we did only bugfixing the last week if it wasn't there it isn't there yet ;-)
21:41 Miner_48er joined #minetest-dev
21:41 john_minetest Okay. Just because this isn't really visible. Bugfixes aren't so visible.
21:41 john_minetest Although they are often more important than features.
21:42 sapier it's quite normal the large amount of work is the one not visible at all ;-)
21:42 john_minetest It is somehow not visible.
21:42 VargaD the changes highlight performance improvements but it seems the mutexed queue still used there unchanged
21:43 john_minetest That isn't optimal. It leaves the impression that the devs aren't doing anything.
21:43 Exio4 so what? is your suggestion stop fixing shit and just add features?
21:43 sapier writing a feature usually is only about 1/3 of way to completion
21:44 sapier mutex queue is considered to be a risky fix, not something to do days prior release
21:44 iqualfragile VargaD: i would recommend not to care about those performance improvements
21:44 sapier it's a very special case only varga
21:45 VargaD sapier: can I help you in any way to merge the mutexed queue changed back?
21:45 sapier mutexed queue is ready to be merged, it's just not done yet
21:45 VargaD OK, thanks :)
21:46 sapier but you've been looking for a change to extract from freeminer, I don't know about this anything but maybe you could create a pull request for this?
21:46 sapier this beeing colored chat ;-)
21:47 hmmmm I don't want just colored text
21:47 hmmmm I want it to be a well designed, preferably standard format for text formatting
21:47 hmmmm if freeminer is okay with quick solutions that turn hacky later on, great
21:47 VargaD we do we need text formatting and color chat?
21:48 VargaD :)
21:48 hmmmm like celeron said, bbcode is okay
21:48 sapier that's why I said extracting it and create a pull request to develop until completion ... don't underestimate this task vargad
21:52 VargaD writing bbcode isn't that simple and fast, are you sure that it is useful in a chat of a realtime game?
21:55 kaeza how about some sort of (literal) escape sequence? that should be really simple
21:55 VargaD chat clients use simple shortcuts to bold/underlined text and also for colors
21:55 kaeza like \c1OMG RED!!!!\c0
21:55 sapier what about some sort of per player coloring only?
21:55 Exio4 \vXXXXXX
21:56 VargaD specing: that should e a different feature
21:56 Exio4 like the actual implementation kaeza?
21:56 VargaD and only color the name
21:56 kaeza Exio4, why do you need full 24 bit color for chat? :I
21:56 VargaD sorry I wanted to write sapier
21:57 Exio4 32*
21:57 Exio4 \vRGBA
21:57 Exio4 the alpha is optional
21:57 kaeza that's plain overkill
21:57 VargaD I don't think that creating a new solution to this very old problem is the best way to solve this task
21:58 sapier is it really usefull to write "\vFF0000FFHello"?
21:59 VargaD how about that widely used irc way: http://wiki.swiftirc.net/index.php?title=Colours
21:59 Exio4 mirc colors? no please
22:00 Exio4 i would prefer writing my messages in json instead of mIRC colors
22:00 Exio4 (something like kaeza's solution)
22:00 VargaD that actual message format can be anything
22:00 VargaD just it is easier to use simple shortcuts
22:00 sapier lol
22:01 kaeza it's easier for me to type \c0 than C-b :)
22:01 kaeza err... whatever the shortcut is
22:01 Exio4 something like '1
22:02 Exio4 something like 1 = red, but as shorcut is good
22:02 Exio4 as long as some "full color" solution is available for mods
22:04 VargaD Exio4: you are right mods need 24bit color with alpha, but I think that simple coloring should be easy to use
22:05 thexyz for now it uses \vRRGGBB format
22:05 thexyz \v being vertical tab
22:06 thexyz dunno about the ability for client to send colorized chat
22:08 thexyz if you're feeling like implementing bb code for such task then do it, but I think that's just wrong
22:08 Exio4 bbcode for realtime chat looks like overkill
22:08 Exio4 wasn't it a joke at first?
22:09 thexyz sapier: you won't write this, colors are purely for mods
22:09 thexyz like in shell
22:09 thexyz or in syntax highlighting
22:09 sapier oh ok good to know
22:10 thexyz well, who knows, it's possible to make a mod which does weird things
22:10 thexyz like parsing the message and adding colors to it
22:10 kaeza ^ useful for pings
22:11 thexyz yes, I've already done some modifications to the chatplus mod
22:12 hmmmm yes bbcode is a half-joke
22:12 thexyz they'll be here when finished https://github.com/xyzz/chatplusplus
22:12 hmmmm the point is to use something that exists and is something of a standard
22:12 thexyz but
22:12 thexyz what exists?
22:12 hmmmm we want more than color, we want to encode if the text is bold, italic, strikethrough, underline, the size, font,
22:12 thexyz you do?
22:13 hmmmm a general rich text format
22:13 hmmmm this isn't just for chatting
22:13 hmmmm this is for UI layout as well
22:13 thexyz ah
22:13 thexyz just throw webkit in
22:13 hmmmm and it just so happens that the same power and flexibility (or maybe a subset)
22:13 hmmmm can be present in the chat
22:13 hmmmm so when you want to make some text in the formspec bold you can use the same escape code as you would in chat
22:14 hmmmm we use the same parsing routine
22:14 thexyz yeah sure
22:14 hmmmm I am sorry, we are engineering software here, not slapping some random shit together to make a mud hut
22:14 hmmmm I'd like designs to be well thought out
22:14 thexyz http://librocket.com/
22:15 hmmmm lol
22:15 thexyz you're adding more hacks to the faulty concept formspecs are
22:15 hmmmm I was thinking about formspec
22:15 Sokomine mud huts arn't the worst in mt. cobble boxes are much more evil
22:15 hmmmm I think it's totally plausible to do an overhaul to formspec to make it into something nice
22:15 hmmmm and I'd like to work on that sometime soon
22:16 Sokomine an option to format text would be very welcomed. i just had the problem to format text and to highlight some numbers - which is beyound what can be done well with formspecs currently
22:17 pitriss hmm hmm.. Guys maybe coloring should'nt be possible in chat.. Colour benefits are in different colors for join/part messages, system messages, /me etc.. and it should color whole line
22:17 thexyz then do it
22:17 pitriss it will make chat more readable just because when you want to chat then you can ignore lines just by color
22:18 Sokomine i don't see much usage for player-controled coloring either, but it's something younger users seem to love
22:19 Sokomine as for the rest, pitriss is right: colored text helps to analyze the information therein much faster in your mind
22:19 thexyz and then I'll replace those crude ugly awful hacks I'm doing with your sane perfect extensible solution
22:26 robmyers joined #minetest-dev
22:31 thexyz so what's the correct way to disable mapgen?
22:31 thexyz if I just say `allow_generate = false` then I guess server thread will just ask for the block again and again?
22:33 Exio4 a maybe-hacky solution would be answer with a block full of content_ignore
22:33 sapier can you enter a content_ignore node?
22:34 sapier left #minetest-dev
22:34 Exio4 you can't, afaik
22:34 thexyz I'd like it to not store this at all
22:34 Exio4 i didn't say "store"
22:35 thexyz won't the server save it?
22:35 VargaD proller has 447 commits, 58 in minetest master branch
22:35 thexyz I'm not sure, probably better to go read the code
22:35 thexyz but that's so boring :(
22:36 Yepoleb joined #minetest-dev
22:36 sepi Hi, can anyone please tell me how to skin a model in blender?
22:36 Exio4 how does the server handle the 'mapgen limit'?
22:37 Exio4 i see
22:38 iqualfragile sepi: i recommend asking that question in #blender
22:39 sepi iqualfragile: Ok, I just wanted to know if I can use regular UV mapping
22:39 Exio4 or in #minetest
22:40 thexyz by the way!
22:40 thexyz here's a set of some abm-related performance improvements https://github.com/freeminer/freeminer/comm​it/b1a64f466226dd8c2086fd2b0358661543ec9080 https://github.com/freeminer/freeminer/comm​it/e226c9f6d4ae9efdd77a93858737bb3ece07f54e  https://github.com/freeminer/freeminer/comm​it/7ae8f655b2ef3e8db58aab094fd079b1b82915e5
22:43 thexyz this changes required_neighbors query complexity to O(1) (previously O(logN)) at expense of some overhead in memory usage
22:43 thexyz same for m_aabms
22:43 thexyz (my profiler told me this part of code takes the most time)
22:47 iqualfragile thexyz: thanks for the effort, these changes realy look good
22:47 iqualfragile (will obviously kill the cache, but it seems that does not matter)
22:49 thexyz this abmhandler also seems to be recreated on every step which isn't good
22:49 thexyz but meh
22:53 iqualfragile VargaD: you could make a pull request out of those changes
22:53 rambomedic joined #minetest-dev
22:53 VargaD I'm reviewing those changes at the moment
22:54 thexyz do note that colored chat isn't ready yet
22:55 PilzAdam joined #minetest-dev
22:59 VargaD does using bitset actually improves performance?
23:00 thexyz using my very own bitset sure does
23:00 thexyz since all operations are O(1) now
23:00 thexyz and it's called very often
23:02 iqualfragile when i first looked at minetest modding one of my first thoughts was: wow, that must be expensive
23:05 Megaf joined #minetest-dev
23:09 thexyz what is?
23:10 RealBadAngel joined #minetest-dev
23:10 iqualfragile thexyz: allowing to set neighbor nodes in abm
23:10 iqualfragile s
23:12 specing thexyz: std::vector<bool>?
23:12 thexyz specing: no, also never use std::vector<bool>
23:12 thexyz because it does evil things like packing your bits
23:13 specing oh, it is space efficient
23:13 thexyz yes, I dunno if it's reasonable to be space efficient
23:13 specing but we dont want space efficiency
23:13 specing we want
23:13 specing 13337 SP3333D!
23:13 thexyz it's not that hard to implement similar stuff without using vectors
23:14 thexyz but it does only eats 64 kb per abm so I honestly don't see the problem, you of course can decrease it to 8 kb/abm
23:14 thexyz just replace std::vector<char> with std::bitset and I guess that's it
23:14 specing thats a lot of bits 0.o
23:15 thexyz that's the size of content_t
23:15 thexyz (just using std::bitset won't work because its count() is O(N) apparently)
23:20 specing I vote not a bug
23:20 specing currently 2:1 in favor of nota bug
23:20 iqualfragile specing: wtf?
23:21 specing iqualfragile: sup?
23:21 thexyz that's not how you decide if something is a bug
23:21 thexyz because it for sure is
23:56 khonkhortisan joined #minetest-dev

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