00:16 <@deen> Looking for actual bugs now: Clang static analyzer found no bugs in ddnet source
00:20 <@deen> oh, or bam just fucks with it
00:20 <@deen> because bam doens't use CC and CXX variables
00:24 < eeeee> yeah bam is weird
00:28 <@deen> well, now I find out that pnglite is full of bugs, cool
00:32 < ddnet-commits> [ddnet] def- pushed 1 new commit to DDRace64: http://git.io/vvN4q
00:32 < ddnet-commits> ddnet/DDRace64 6d54de8 def: Get rid of clang static analyzer warnings
00:48 < eeeee> i also found a tw bug while porting websockets to vanilla
00:48 < eeeee> a string is copied into itself, crashes on mac
00:50 <@deen> i remember seeing that somewhere
00:51 < eeeee> https://github.com/def-/ddnet/commit/db4d4724432eca09d16ec1ad5c8f82fb92b668ee
00:51 < eeeee> ^^
00:53 <@deen> right
00:55 < eeeee> someone should pull that to vanilla... heinrich5991, fstd: what's the C+ industry standard way of dealing with this kinda thing?
00:57 < eeeee> does everyone just check if it's the same reference or what
00:59 < fstd> you'd normally not end up in a situation like that in the first place, i think :)
00:59 < fstd> i've seen that back in a while too, and it made me wonder..
01:09 < eeeee> well anyway, with that patch it seems to work
01:09 < eeeee> valgrind clean
01:09 < eeeee> deen: wanna try running vanilla servers with websockets?
01:12 <@deen> all the ddnet ones or what?
01:12 <@deen> eeeee: ^
01:15 < eeeee> yeah
01:15 <@deen> have a repo somewhere?
01:15 < eeeee> https://github.com/eeeee/ddnet/tree/vanilla_ws
01:25 <@deen> eeeee: done
01:32 < eeeee> awsum thx, seems to work
01:33 <@deen> Now the vanilla people will also be happy with the js client =)
01:34 <@deen> maybe finally someone will play on our vanilla servers
01:34 < eeeee> i wonder what should be the default settings for server filter in js client
01:35 <@deen> show all servers, sort by ping?
01:35 < eeeee> if vanilla servers had more people i'd just default to vanilla-only
01:35 < eeeee> because there's no way someone new to teeworlds can figure out how to play ddrace by just joining a random server
01:35 <@deen> yes, that's pretty unfortunate
01:36 < eeeee> in vanilla they have a chance, people naturally just start with shooting everyone with a pistol
01:36 <@deen> haha
01:37 < fstd> "wie macht man diese kette?"
01:38 <@deen> i played TW without hook for the first weeks or so
01:38 <@deen> couldn't figure that out at all
01:38 <@deen> hard enough to jump around and aim already
01:39 <@deen> but the typical noob question on ddrace is "how do i join?"
01:39 <@deen> followed by "how do i restart?"
01:40 < eeeee> yeah and it's twice as hard in js client because you have to press ESC twice to get a menu
01:43 < fstd> can't you insert some sort of clippy for that?
01:43 < fstd> or for ddrace in general, like a tutorial map/server
01:43 <@deen> sure, please do that and make a PR
01:43 < eeeee> that'd be nice, but it's a lot of work
01:43 <@deen> I'm not really good at making noob-friendly stuff
01:44 <@deen> python hates me btw. it uses twice the amount of memory it should
01:45 <@deen> somehow the GC doesn't want to clean out the old data when i load new one
01:45 < fstd> that's probably because it's a crappy language
01:45 <@deen> I'm too lazy to rewrite my script
01:45 <@deen> s
02:19 <@deen> ha, fixed the memory issue with a few manual dels, 150 MB more RAM to play around with
04:26 < eeeee> got a new domain for js client: teewebs.net
04:26 < eeeee> can you test plx
04:35 < fstd> eight A and four NS records; very srs, much elastic
04:37 < eeeee> i spent like half a year making that shit so the actual delivery better fucken work perfectly
04:51 < fstd> two nameservers is sound.  three nameservers is paranoid.  four nameservers is a public statement that the stuff is basically guaranteed to fail, and with one fail, a cascade of failure isn't all too far :\)
04:51 < fstd> so better take one nameserver out to increase reliability
05:22 < eeeee> u mad
09:29 < Ryozuki> hi
09:31 <@EastByte> hello
09:32 <@EastByte> eeeee: lol dynamic screen resize nice!
09:33 < eeeee> yup
09:34 <@EastByte> hmm any chance to lock the mouse?
09:34 < eeeee> it should lock
09:34 < eeeee> doesn't it ask you to lock or sth?
09:35 <@EastByte> okay I missed the dialog
09:35 < eeeee> -_-
09:35 < eeeee> those browser guise should really make it more visible, deen missed it too 
09:36 < heinrich5991> can I have a link, again?
09:36 < eeeee> teewebs.net
10:09 < eeeee> lol i keep testing my stuff on US servers and i don't think i ever seen people actually race there
10:10 < eeeee> they mostly keep pouring shit on each other in chat
10:10 < eeeee> guess should focus on testing the chat function
10:15 < heinrich5991> deen: you modified the secret password on the race_sanktoras map, huh? :)
10:30 <@deen> heinrich5991: i didn't
10:34 <@deen> eeeee: can't connect to the GER2 servers I think
10:34 < eeeee> um yeah, because tcp is filtered
10:34 < eeeee> i'll unfilter it i guess
10:34 <@deen> eeeee: and want any other mods with websockets?
10:35 <@deen> zcatch is pretty popular currently
10:35 < eeeee> idk
10:35 <@deen> anyway, can be added later
10:36 < eeeee> yeah let's see if anyone asks for it
10:36 <@deen> works really great
10:36 <@deen> haven't found a bug yet
10:37 < eeeee> i know a couple :P
10:37 <@deen> does it save settings somewhere?
10:37 < eeeee> not yet
10:37 <@deen> ok
10:37 < eeeee> i plan to add some local storage later
10:37 <@deen> ah, when you click on the editor, you can't get out again
10:38 < eeeee> uh i better just disable the editor
10:38 < eeeee> i didn't test that at all
10:38 <@deen> right
10:40 <@deen> dyncam seems a bit weird
10:40 <@deen> doesn't really look up and down for me, only left and right
10:41 < eeeee> hmm didn't notice. will check.
10:50 <@deen> eeeee: works pretty great, good job
10:51 < eeeee> thx
10:56 <@deen> even seems to work on phone browser
10:56 <@deen> missing a keyboard of course
11:00 <@EastByte> well, even starts on my 40eur tablet
11:00 <@EastByte> but then it runs out of ram
11:00 <@EastByte> one can't expect too much
11:04 <@deen> and i can't connect to our fullest blocker server because it hasn't been empty for a month, so no chance to update the server version
11:04 <@deen> eeeee: when do you want to go public with teewebs.net?
11:05 < eeeee> tomorrow i guess
11:05 <@deen> ok, then i should restart it
11:05 <@deen> i hope the 60 guys won't be mad
11:05 < eeeee> lol no
11:05 <@deen> no?
11:05 < eeeee> i'm not sure it can handle servers that big anyway
11:05 <@deen> is it bad to restart a block server?
11:06 <@deen> but many will sort by player number and try to go on the fullest server immediately
11:06 <@deen> bad if that just hangs
11:06 < eeeee> then they'll try something else? idk we can just filter out block servers (with br_exclude thing)
11:06 <@deen> i don't think we should
11:07 < eeeee> okay up to you then
11:07 < eeeee> but restarting 60 people is a bit....
11:08 < eeeee> maybe could at least slowly drain the server
11:08 < eeeee> by passwording it or something
11:08 <@deen> meh, that's bad
11:08 < eeeee> is it worse than restarting?
11:08 <@deen> yes
11:08 <@deen> their friends can't join to them and they don't even notice
11:08 < eeeee> even when you have another block server with same map on same host?
11:09 <@deen> anyway, i can restart at night, when only 30 players are on
11:09 < eeeee> well they would notice since you'd also have a broadcast announcing that
11:09 <@deen> they don't/can't read etc
11:09 < eeeee> :/ okay then
11:12 <@EastByte> they probably cannot read at all
11:12 <@deen> you can announce that you will restart 10 times and afterwards they still think someone just hacked the server
11:12 <@EastByte> the shutdown cmd doesn't accept a reason right :/
11:12 <@deen> i think so, never checked
11:12 <@EastByte> heh
11:13 <@EastByte> Console()->Register("shutdown", "", CFGFLAG_SERVER, ConShutdown, this, "Shut down");
11:13 <@EastByte> but you could kick all with a sh script
11:13 <@EastByte> with reason
11:23 < ddnet-commits> [ddnet] def- pushed 1 new commit to DDRace64: http://git.io/vvAAS
11:23 < ddnet-commits> ddnet/DDRace64 84cdcdb def: Remove wrong include
11:42 <@deen> laxadedi: any idea why nimda doesn't like http://ddnet.tw/releases anymore? is it because i release 5 maps at a time?
11:56 < ddnet-commits> [ddnet] def- pushed 1 new commit to DDRace64: http://git.io/vvxTM
11:56 < ddnet-commits> ddnet/DDRace64 d4d5109 def: Timestamp index for race table for improved performance of /players/
11:57 <@deen> time down from 600 to 100 ms on first request on http://ddnet.tw/players/deen/ (further request are cached anyway)
11:58 <@deen> all the time is spent to determine the 10 last finishes anyway, rest is prepared
12:10 <@deen> So, ddnet with mysql uses threads, but that's not supported by the Mysql Connector
12:11 <@deen> I'm wondering if that's the cause of our slow memory leaks
12:11 <@EastByte> gawd, I hate all the sql libs out there
12:12 <@deen> I guess we should have used a single thread with a queue instead, but i think in practice that may turn out to be less reliable
12:12 <@EastByte> I think it's the proper way
12:12 <@deen> haven't seen any problems with our threaded mysql, but that's probably because we have a low amount of threads around at any time
12:13 <@EastByte> would be easy to just add locks instead of making a queue system
12:13 <@deen> oh, it has locks!
12:13 <@deen> my bad
12:13 <@EastByte> !
12:13 <@deen> so that's probably fine
12:14 <@deen> but a queue system could be nicer and cheaper I guess
12:14 <@deen> not that i'm going to implement it, but should be simple, because all the data is already passed to the thread, could just be passed to the queue in the exact same way
12:15 <@deen> ah, and i think it causes the memleaks, because only the ddrace and zcatch servers leak memory, exactly those that have mysql on
12:17 <@EastByte> great
12:17 <@EastByte> maybe do some valgrind magic?
12:17 <@EastByte> or do you think a queue would fix the problem?
12:17 <@deen> i tried many times to find that with valgrind
12:17 <@deen> i believe a queue could
12:18 <@deen> valgrind never reported anything, but it's probably not easy to report memory problems with all the threads flying around
12:18 <@deen> and i can't simulate it well, on my local system memory never increases, only on real servers, after a few days
12:19 <@EastByte> would be too easy otherwise :P
12:30 <@EastByte> deen: I looked into c2nim recently
12:30 <@EastByte> writing a C lexer/parser definitly is a PITA
12:47 < laxa> deen: isn't the caching supposed to be on the SQL instance ?
12:47 < laxa> or it's implemented on the SQL lib ?
12:49 <@deen> laxa: what caching?
12:51 <@deen> need to think more about mysql indices. down to 0 ms now =)
12:53 < ddnet-commits> [ddnet] def- pushed 1 new commit to DDRace64: http://git.io/vvxle
12:53 < ddnet-commits> ddnet/DDRace64 230ad1a def: Revert "Timestamp index for race table for improved performance of /players/"...
12:58 < laxa> deen: sql caching
12:59 <@deen> the sql server caches, yes. so?
12:59 < laxa> well, my question is
12:59 < laxa> caching should be on sql server side right ?
12:59 <@deen> right
12:59 < laxa> I am just curious :)
13:11 <@deen> EastByte: on the other hand, the ddnet test servers without sql also use much memory =/
13:24 < ddnet-commits> [ddnet] timgame opened pull request #194: resend player/dummy info that filtered the server (DDRace64...DDRace64) http://git.io/vvxuw
13:28 <@EastByte> meh
13:29 <@EastByte> deen: we could try my own memory leck detection method
13:29 <@deen> ?
13:29 <@EastByte> counting the number of allocs and frees an substract :)
13:29 <@EastByte> has to be zero 
13:30 <@EastByte> I guess valgrind is more advanced in this case :D
13:30 <@deen> some memory is never meant to be freed and should exist until program ends
13:30 <@EastByte> hm such memory has to be excludet
13:31 <@EastByte> I always used this method to verify my memorymanagement
13:34 < ddnet-commits> [ddnet] def- pushed 2 new commits to DDRace64: http://git.io/vvx2N
13:34 < ddnet-commits> ddnet/DDRace64 c4f011d timgame: resend player/dummy info that filtered the server
13:34 < ddnet-commits> ddnet/DDRace64 b0085ae Dennis Felsing: Merge pull request #194 from timgame/DDRace64...
13:36 <@EastByte> deen: do you know why connecting dummy takes so much time?
13:36 <@deen> EastByte: yes, that's because of the new security connection token stuff
13:36 <@EastByte> ah okay
13:36 <@deen> but please find a way to make it shorter
13:37 <@EastByte> I have 5 final exams the next 6 weeks :P
13:37 <@deen> coding is the best diversion?
13:38 <@deen> distraction*
13:38 <@EastByte> why would I want to be distracted? xd
13:38 <@EastByte> I use coding to learn math though
13:38 <@deen> because studying all day is boring
13:38 <@EastByte> ah so I stop studying...
13:39 <@EastByte> well one thing I would add first
13:39 <@EastByte> and indicator for dummy connecting process (button not clickable etc.)
13:39 <@EastByte> an*
13:39 <@deen> yes, I'm working on that right now
13:39 <@EastByte> ah great, good luck
13:47 < Teetime> deen: regarding the memory-leaks, do you know if exceptions are thrown? because then the memory allocated for the results/etc won't be deleted since the delete-parts will be skipped
13:49 <@deen> Teetime: i checked for that. The exception handling looks fine and should delete objects fine. Also didn't find a single thrown exception in the logs
13:54 < Teetime> hm.. guess it'll be difficult to find
13:54 <@deen> it's probably not even a leak
13:54 <@deen> i think some object store just grows slowly until it reaches a limit
13:54 <@deen> the limit is just a bit too high for my taste
13:56 < Teetime> but what object store could it be?
13:57 <@deen> bullets for example
13:57 <@deen> new CProjectiles get created a lot, but where are they removed?
13:57 <@EastByte> aren't they added to the entities list
13:57 <@EastByte> ?
13:57 <@deen> yeah, something like that
13:58 <@EastByte> hmm, someone needs to look into that
14:06 < ddnet-commits> [ddnet] def- pushed 1 new commit to DDRace64: http://git.io/vvxXW
14:06 < ddnet-commits> ddnet/DDRace64 c1aeebb def: Clean up a bit and hopefully fix player move on dummy connect
14:07 < ddnet-commits> [ddnet] def- force-pushed DDRace64 from c1aeebb to 9c1e65a: http://git.io/-BwJ2w
14:07 < ddnet-commits> ddnet/DDRace64 9c1e65a def: Clean up a bit and hopefully fix player move on dummy connect
14:10 < ddnet-commits> [ddnet] def- pushed 1 new commit to DDRace64: http://git.io/vvx1B
14:10 < ddnet-commits> ddnet/DDRace64 d690877 def: At least don't access invalid memory on server request. Should fix #173
14:16 < ddnet-commits> [ddnet] def- pushed 1 new commit to DDRace64: http://git.io/vvxDU
14:16 < ddnet-commits> ddnet/DDRace64 2e8ee2b def: Disallow spectators to participate in kick votes
14:16 < ddnet-commits> [ddnet] def- force-pushed DDRace64 from 2e8ee2b to 14b5766: http://git.io/-BwJ2w
14:16 < ddnet-commits> ddnet/DDRace64 14b5766 def: Disallow spectators to participate in kick votes...
14:17 <@EastByte> https://github.com/def-/ddnet/commit/d690877117ce1537200d34bb728c874637d0058b looks like a non-trivial client vuln
14:24 < ddnet-commits> [ddnet] def- pushed 1 new commit to DDRace64: http://git.io/vvx9o
14:24 < ddnet-commits> ddnet/DDRace64 16106d8 def: Fix menu bug with button pressing (more of a hack)
14:40 < ddnet-commits> [ddnet] def- pushed 1 new commit to DDRace64: http://git.io/vvxdk
14:40 < ddnet-commits> ddnet/DDRace64 8f02de3 def: Make dummy connecting a bit nicer: Show that dummy is being connected, disable unusable buttons
14:41 <@deen> EastByte: that fine?
14:48 <@EastByte> deen: well seems to work
14:49 <@EastByte> but the button still looks clickable
14:50 <@deen> yep, how to fix that?
14:50 <@deen> i didn't find a way
14:50 <@EastByte> hmm
14:51 <@deen> there is a "Checked" parameter, but that didn't change how it looks
14:52 <@EastByte> I guess there is no option for it
14:59 < heinrich5991> EastByte: there are lots of "non-trivial" client vulns
15:00 <@EastByte> seems so
15:55 <@deen> eeeee: GER2 tcp ports still look blocked
16:13 <@deen> eeeee: zcatch should now also have websockets, at least the servers that i could restart so far
16:14 <@deen> and i made http://web.ddnet.tw/ a link to http://teewebs.net/ , hope that's fine
21:46 < eeeee> deen: unblocked GER2, zcatch also seems to work now
22:21 <@EastByte> eeeee: what do you mean by "unblocked"?
22:21 <@EastByte> allowed to be requested using websockets?
22:23 < eeeee> EastByte: tcp was firewalled pretty hard on ger2, because GER2 was a ddos attack fallback for GER
22:23 <@EastByte> okay I understand
22:36 <@deen> eeeee: great, thanks
23:36 < eeeee> do you think i should make the js client auto-join on ddrace servers?
23:37 <@deen> either that, or we do auto-joins on ddnet ddrace servers in general
23:38 <@deen> anyone have an opinion on that?
23:38 <@deen> most people play
23:38 <@deen> only few spectate
23:38 <@deen> so it would make sense
23:41 <@deen> I'll just do that right now, let's see how many people complain
23:41 <@deen> if it's less than 50 it's fine i guess
23:42 < eeeee> kk
23:43 <@deen> done
23:46 <@deen> so, can I announce the teewebs link tomorrow?
23:46 <@deen> and put it in an iframe on ddnet.tw? :P
23:47 < eeeee> should prolly test the iframe ahead of time :P
23:47 < eeeee> i'll make a forum post for on teeworlds.com and ddnet forum later today
23:47 <@deen> wel, today is over in 10 minutes for me
23:48 < eeeee> tomorrow it is for you, then
23:54 <@deen> Looks like it loses focus easily: http://ddnet.tw/index2.php
23:56 <@deen> when i'm in it it's fine, but then i accidentally press TAB and lose focus. any trick to fix that?
23:57 < eeeee> tab doesnt unfocus it for me
23:58 <@deen> only after esc and changing windows
23:59 < eeeee> lol i closed the whole browser tab and now my pointer is permanently locked in firefox
23:59 <@deen> so many bugs^^
23:59 < eeeee> shit i have to restart the whole firefox now