2011-09-18T00:00:02 *** dvladim has joined #aichallenge 2011-09-18T00:28:17 McLeopold: yt? 2011-09-18T00:49:02 *** eashoka has joined #aichallenge 2011-09-18T00:49:03 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has quit IRC (Ping timeout: 260 seconds) 2011-09-18T00:50:29 Is there anyone here who can help me figure out why my submission won't compile. It works for me locally, but not on the beta server and the error message the beta server reports isn't very helpful. 2011-09-18T00:53:19 ChrishH : check for any class or some other dependency which is in ur platform but not in the bundle u submitted ... 2011-09-18T00:53:51 I already did. 2011-09-18T00:55:22 eashoka : r using a tool like Ant to package or doing it manually ? 2011-09-18T00:56:14 I have submitted a very minimal bot just to verify the server will compile it, and it claimed to once but then failed in one game. 2011-09-18T00:56:26 A couple days ago I was on here and some of the organizers were discussing some discrepencies between the main server and the workers causing my bot to fail. 2011-09-18T00:57:04 eashoka: no, but I am using Go, which apparently is not well tested on the beta server. 2011-09-18T00:57:06 eashoka : ah ok. what's the language u sued ? 2011-09-18T00:57:16 ah ok. 2011-09-18T00:57:36 So I'm here to help work it out, but I must say I'm getting frustrated. 2011-09-18T00:58:07 I posted on the forum two days ago, but no response yet. 2011-09-18T00:58:44 IRC seems to be better when I can find someone, but even then ... not sure what more I can do. 2011-09-18T00:59:11 I have been told that the servers are running Go version r.60. 2011-09-18T00:59:17 And that's what I am using as well. 2011-09-18T00:59:37 My code runs properly when compiled on my machine and tested with the latest tools from Github. 2011-09-18T00:59:54 e.g. play_one_game.sh, etc. 2011-09-18T01:00:18 Any suggestions how I can move forward to help resolve whatever the problem is? 2011-09-18T01:00:23 eashoka : there aren't many Go users I think. did u have a look at source code of Go framework classes ? 2011-09-18T01:01:17 True, gobang2 is the only other one listed on the beta server. 2011-09-18T01:01:40 I did have a look at the Go sample bot. 2011-09-18T01:01:52 I tried uploading it unchanged and it also failed to compile. 2011-09-18T01:03:12 It would really help if I could get the compile error messages, but all I see is: "Output file _go_.6 was not created." which doesn't really tell me much. 2011-09-18T01:04:16 what is _go_.6 ? 2011-09-18T01:06:36 The file extension of .6 means it is a compiled version of a Go module targeted for an 64 bit x86 architecture. 2011-09-18T01:06:47 It's like a .o file for C or C++ 2011-09-18T01:10:39 just wondering. r u submitting a similar file with ur source code as well or u r submitting only the source code ? 2011-09-18T01:11:57 Only the source. 2011-09-18T01:12:34 I'm looking through the worker/compiler.py code now to see if I can figure out how the server runs the Go compiler. 2011-09-18T01:13:58 Chrish : g8, I think that's a promising way forward. If that fails too (hope not :-) ), u can always try with another language to be familiar with the competition until this issue get eventually resolved. 2011-09-18T01:14:59 Well, I can always work on my bot locally as well, but I don't want to sink too much time into doing it in Go if the language won't really be supported. 2011-09-18T01:21:04 eashoka: I'm simulating the same compiler commands that the server uses and starting to get an idea of what's going on. 2011-09-18T01:23:39 ChrisH : glad to hear! 2011-09-18T01:23:55 *** Chris_0076 has quit IRC (Remote host closed the connection) 2011-09-18T01:24:10 Ok, I see the problem. 2011-09-18T01:25:52 I have my code split into multiple files (lexer.go, parser.go, map.go, MyBot.go, etc.) and the code in worker/compile.py doesn't handle that properly for Go. 2011-09-18T01:26:09 *** Chris_0076 has joined #aichallenge 2011-09-18T01:26:49 So I have two choices, put all my code in one big source file or get compile.py updated to allow multiple go sources for one bot. 2011-09-18T01:27:12 Any idea if I can get compile.py patched? 2011-09-18T01:30:10 *** eashoka has quit IRC (Ping timeout: 252 seconds) 2011-09-18T01:30:30 ChrisH: talk to McLeopold I rekon 2011-09-18T01:30:44 @later tell Chris H use this 2011-09-18T01:30:44 antimatroid: OK 2011-09-18T01:30:56 oh, he's even in the channel 2011-09-18T01:31:01 antimatroid: ok 2011-09-18T01:32:32 McLeopold: I have taken a look at the latest worker/compiler.py code and I know why my Go code won't compile on the server and I also have a suggestion for an improvement to compile.py that would fix the problem. Are you interested? 2011-09-18T01:33:34 ChrisH: he might not actually be here, is there much of a change to the file? you could submit a pull request on github if you know how (i'm not very knowledgeable with github) 2011-09-18T01:33:48 then someone should be able to just take a look at the changes and let it come through 2011-09-18T01:33:59 I already did a pull before browsing the code. 2011-09-18T01:34:34 yeah, i think a pull request let's you upload changes to go through once someone with privileges allows it 2011-09-18T01:34:43 (I am not good with github :)) 2011-09-18T01:35:11 hmmm, I understand pull to retrieve the latest changes from the server (I'm not great at git either). 2011-09-18T01:36:13 I can do some reading though, I'm sure I can figure it out given a little time. 2011-09-18T01:54:46 *** amstan has quit IRC (Read error: Operation timed out) 2011-09-18T01:56:23 *** eashoka has joined #aichallenge 2011-09-18T01:58:36 Well, now I'm not so sure that I have this figured out. 2011-09-18T02:00:35 I tried merging all my code into a single MyBot.go source file and submit it, but the server still fails to compile it. In addition I just noticed a part of worker/compiler.py that seems to handle the muliple source file situation for Go properly. So I'm back to being stumped and wishing I could see more information about the errors the compiler is producing. 2011-09-18T02:13:19 McLeopold: I did some further testing. I ran compiler.py on my machine with both a single source file and with multiple source files and both worked correctly, resulting in a fully functional bot that works with play_one_game.sh, so I can only assume that the problem is a difference in the Go environment on my machine and on the servers. I will need help from the organizers to figure that out. 2011-09-18T02:24:56 *** ChrisH has quit IRC (Quit: Page closed) 2011-09-18T02:26:49 *** Palmik has joined #aichallenge 2011-09-18T02:28:53 *** dvladim has quit IRC (Ping timeout: 276 seconds) 2011-09-18T03:48:52 *** nux67 has joined #aichallenge 2011-09-18T04:37:17 McLeopold: Do you use Eclipse + PyDev ? 2011-09-18T04:45:18 *** FireFly has joined #aichallenge 2011-09-18T04:55:18 @beta 2011-09-18T04:55:19 antimatroid: beta could be http://aichallengebeta.hypertriangle.com/. 2011-09-18T05:01:09 *** mceier has joined #aichallenge 2011-09-18T05:16:35 *** berak has joined #aichallenge 2011-09-18T05:21:39 *** cody__ has joined #aichallenge 2011-09-18T05:22:50 *** gobang has joined #aichallenge 2011-09-18T05:45:55 *** onensora has quit IRC (Ping timeout: 252 seconds) 2011-09-18T05:50:36 *** onensora has joined #aichallenge 2011-09-18T05:53:32 antimatroid: still there? 2011-09-18T05:54:00 *** berak has quit IRC (Read error: Connection reset by peer) 2011-09-18T05:54:00 *** gobang has quit IRC (Read error: Connection reset by peer) 2011-09-18T05:56:04 *** gobang has joined #aichallenge 2011-09-18T06:05:44 mleise: yeah 2011-09-18T06:06:18 *** dvladim has joined #aichallenge 2011-09-18T06:24:33 *** arlequ1 has joined #aichallenge 2011-09-18T06:36:35 antimatroid: I added an issue up for discussion about adding some Eclipse projects to the source repo. I know it is unusual, but may be handy to get a quick start on the project. 2011-09-18T06:36:58 *** dvladim has quit IRC (Ping timeout: 252 seconds) 2011-09-18T06:38:32 *** cody__ has quit IRC (Ping timeout: 252 seconds) 2011-09-18T06:53:15 *** onensora has quit IRC (Ping timeout: 260 seconds) 2011-09-18T06:54:40 *** onensora has joined #aichallenge 2011-09-18T07:03:00 *** onensora has quit IRC (Ping timeout: 240 seconds) 2011-09-18T07:04:38 *** sir_macelon has joined #aichallenge 2011-09-18T07:04:46 hi 2011-09-18T07:05:14 sir_macelon : hi 2011-09-18T07:05:49 eashoka: hi, what's up? 2011-09-18T07:06:14 see, you're leading now, congrats :) 2011-09-18T07:06:25 *** onensora has joined #aichallenge 2011-09-18T07:09:02 sir_macelon : Tx. but it's a temporary effect of the 255 cut off limit I guess :-) 2011-09-18T07:09:53 eashoka: I guess so, but anyway :) 2011-09-18T07:10:09 hopefully the real thing would start from next week :-) 2011-09-18T07:11:54 hopefully, cause during last month I did almost nothing with my bot, been on holidays etc. 2011-09-18T07:13:06 *** onensora has quit IRC () 2011-09-18T07:13:37 sit_macelon : ah g8! where to ? 2011-09-18T07:24:10 eashoka: locally in poland, by the see, but also had to attend two weddings, so travelled a lot 2011-09-18T07:24:36 *** delt0r_ has quit IRC (Ping timeout: 240 seconds) 2011-09-18T07:24:42 hopefully next year I'll be able to plan it better 2011-09-18T07:25:00 sir_macelon : BTW, SL is the number 1 ranked tourist destination these days ;-) 2011-09-18T07:25:58 eashoka: haven't heard that, but to far for me at the moment, maybe the other day :) 2011-09-18T07:26:11 *too far 2011-09-18T07:26:45 sir_macelon : I finally got my timeout issues solved. it is still not upto ur strength in battles I think. very impressed with the strategy in mathis's bot. 2011-09-18T07:27:08 sir_macelon : SL = Sri Lanka 2011-09-18T07:27:57 I've been speeding up my calculations recently also 2011-09-18T07:28:55 yep, he's very aggresive, which seems to pay off in a low-food-spawn-rate games 2011-09-18T07:30:58 I'm yet to develop a strategy for that, it is my next to-do 2011-09-18T07:32:06 eashoka : me too having a long list of improvements and defects to fix, time is the problem :-( 2011-09-18T07:34:41 *** olexs has joined #aichallenge 2011-09-18T07:37:18 *** delt0r_ has joined #aichallenge 2011-09-18T07:53:33 *** mcstar has joined #aichallenge 2011-09-18T08:08:09 *** xathis has joined #aichallenge 2011-09-18T09:06:34 When we are out of beta, you two will already have bots noone will be able to beat in 2 months work 2011-09-18T09:17:24 *** olexs has left #aichallenge 2011-09-18T09:17:26 *** olexs has joined #aichallenge 2011-09-18T09:19:20 mleise: did you write a bot? i think they are all quite simple really 2011-09-18T09:20:52 *** rabidus has joined #aichallenge 2011-09-18T09:25:00 mleise: I do, but I've been sticking to vim only for awhile 2011-09-18T09:25:57 @later tell ChrisH you bot is still breaking the server, which is ok, we are in beta. I'm trying to get it to report to you correctly. 2011-09-18T09:25:57 McLeopold: I think that worked... 2011-09-18T09:28:15 *** dvladim has joined #aichallenge 2011-09-18T09:39:13 mleise : I started coding two months ago and guess I was in first 10 in abt 1 month. So I think it's possible for a smart hardworking person to still join and storm to the top beating all of us, but yea in most cases beta testers will have an edge. may be the contest duration be increased to 3 months as antimatroid once suggested. 2011-09-18T09:40:57 mleise :BTW, what's your bot ? 2011-09-18T09:42:31 eashoka: My bot is still the starter bot. I didn't find the motivation to work on my bot. But working on the contest is nice, too. 2011-09-18T09:43:28 *** ChrisH has joined #aichallenge 2011-09-18T09:44:43 McLeopold: Ok, thanks. Let me know if there is anything I can do to help. 2011-09-18T09:45:01 McLeopold: ah ok, because I was raising the issue if it would make sense to share our Eclipse projects on github. I figure if someone would want to work on the visualizer then it would be nice to have the IDE ready to start the debugger on it. 2011-09-18T09:46:19 + the plugins detect a lot of minor bugs or stuff like shadowed identifier names, invalid characters in html etc. 2011-09-18T09:50:19 I don't think it would hurt 2011-09-18T09:50:27 soon there will be an omnipotent IDE, that will enslave humans, will use them to push the buttons on those many keyboards 2011-09-18T09:51:14 mcstar: reading too much smbc 2011-09-18T09:51:44 http://www.smbc-comics.com/? 2011-09-18T09:51:44 ChrisH: what's your patch for compiler? 2011-09-18T09:51:49 :) yes 2011-09-18T09:51:49 mcstar: Actually I can understand if people don't like that Eclipse centrism, but look how we all use Google's search engine now. It is just a convinient starting point. 2011-09-18T09:52:04 apparently its down, i cant see whats that 2011-09-18T09:52:13 it's just a .project file, people can ignore it if they want 2011-09-18T09:52:19 sure 2011-09-18T09:52:26 i was recently confronted with eclipse 2011-09-18T09:52:40 i had to use an older version, and use its package manager 2011-09-18T09:52:44 or whatnot 2011-09-18T09:53:02 and I would put all projects in /eclipse_projects, so they don't show up inside the source directories at all 2011-09-18T09:53:24 mcstar: an older version? why is that? discontinued plugins? 2011-09-18T09:53:25 the only thing is, I havent found a syntax highlighter/object browser that does javascript well 2011-09-18T09:54:05 the software i was trying to use(an agent model) was written as an extension to eclipse, and used java 2011-09-18T09:54:24 and it was only available to an earlier version 2011-09-18T09:54:30 helios 2011-09-18T09:55:19 McLeopold: I switched to NetBeans for that, but the latest update to Eclipse made me come back 2011-09-18T09:55:37 McLeopold: I followed the instructions on the Go website to get the latest stable release using Mercurial and my compiler reports version "release.r60 9481" 2011-09-18T09:55:40 mcstar: i see 2011-09-18T09:56:13 * mcstar just doesnt understand why doesnt emacs(maybe vim) rule the developer world 2011-09-18T09:56:37 ChrisH: compiler.py can put all go files in one command 2011-09-18T09:56:50 McLeopold: Although not everyone may be happy about the current state of the type inference. It sometimes bites you, especially if you repurpose variables for different data types 2011-09-18T09:57:12 Mcleopold: Yes, I figured that out eventually, and I tested it on my machine and it worked. 2011-09-18T09:59:08 McLeopold: So that left me with the assumption that there was some type of compile error, and thus no _go_.6 file generated (which is all the server tells me), and if I saw the compile errors I might have a better idea what the real problem was. 2011-09-18T09:59:16 whats the difference between suns jre and openjdk? are the hotspots different? or just the libraries? whats the diff. in startup speed? 2011-09-18T09:59:28 ChrisH: It's probably the SplitN issue 2011-09-18T09:59:44 *** jako has joined #aichallenge 2011-09-18T09:59:45 I'm checking why you are not getting a full error message now. 2011-09-18T09:59:50 McLeopold: What's that? I don't use SplitN in my code, BTW. 2011-09-18T10:00:13 McLeopold: Doesn't mean it isn't used by something I do call though. 2011-09-18T10:00:16 or the __MACOSX directory 2011-09-18T10:00:29 mcstar: I think openjdk is a complete open-source implementation of java 2011-09-18T10:00:44 I gotta go for a bit, be back later. 2011-09-18T10:00:49 *** ChrisH has quit IRC (Quit: Page closed) 2011-09-18T10:00:53 that is - for many - reason enough to use it :) 2011-09-18T10:00:55 i was under the impression that only one JVM exists 2011-09-18T10:01:11 hmm, then i must be wrong 2011-09-18T10:01:22 im not saying that 2011-09-18T10:01:35 but you need a better arguemtn to convince me 2011-09-18T10:01:37 :) 2011-09-18T10:01:48 like, one jvm at all, or just for linux? 2011-09-18T10:02:03 from oracle's/sun's side 2011-09-18T10:02:08 there are others, i know 2011-09-18T10:02:08 it is the default in many linux distributions 2011-09-18T10:02:14 JRockIt 2011-09-18T10:02:20 from Oracle 2011-09-18T10:02:22 so i think it must be open-source 2011-09-18T10:02:53 McLeopold: so my question is: are the jvm's for sun's proprietary java and openjdk's java are the same? 2011-09-18T10:03:59 jvm startup: ~10sec, sbcl ~2 sec, python ~0 sec 2011-09-18T10:04:11 python 0 ? Oo 2011-09-18T10:04:14 im staying with python for my scripts 2011-09-18T10:04:19 essentially 0 2011-09-18T10:04:41 yeah, don't load up a 'platform' just for a script ;) 2011-09-18T10:04:44 i dont like it(the stupid identation) but python is very good for certain tasks 2011-09-18T10:06:17 i came across a hint that jvm can be started in "server" mode? to reduce initial startup time, and after that all processes spawn from that one 2011-09-18T10:06:30 do you know anything about this? i couldnt find details 2011-09-18T10:06:55 indentation is clean, I like it, so does CoffeeScript 2011-09-18T10:06:59 sry, to amortize initial startup time* 2011-09-18T10:07:12 mcstar: no idea, but I know that applets are loaded into one VM 2011-09-18T10:07:27 McLeopold: its very hard to follow levels of indentation through pages 2011-09-18T10:07:27 so a web site with many applets has the startup time of only one VM 2011-09-18T10:08:15 mcstar: yes it is. I've heard the argument that if it is too hard, you probably need to refactor 2011-09-18T10:08:42 yeah, i heard that too :) 2011-09-18T10:08:57 the problem, is that indentation changes semantics 2011-09-18T10:09:10 well, that's the benifit if you like it :) 2011-09-18T10:09:23 it happended to me that i un-indented a code-segment, which belonged to a conditional 2011-09-18T10:10:02 basically, i think what im saying is that an end-delimiter would come handy 2011-09-18T10:10:20 it would help the editor to understand the code too 2011-09-18T10:10:37 no semicolon, but {} ? 2011-09-18T10:10:59 that last argument doesn't make sense, the editor should be able to read indentation just like the interpreter 2011-09-18T10:11:38 unless by editor, you mean you? 2011-09-18T10:11:40 McLeopold: it would understand if something were to belong to a something: 2011-09-18T10:11:45 hehe 2011-09-18T10:11:46 nooo 2011-09-18T10:11:55 something: 2011-09-18T10:12:00 statement1 2011-09-18T10:12:03 statement2 2011-09-18T10:12:14 the editor doesnt know where sattement2 belongs 2011-09-18T10:12:25 forget thet 1 space 2011-09-18T10:12:30 of course it does, it uses the indentation levels 2011-09-18T10:12:34 nono 2011-09-18T10:12:39 the editor should indent for me 2011-09-18T10:12:51 you see what i mean? 2011-09-18T10:12:57 i dont want to keep track of indentation 2011-09-18T10:13:10 oh, then no, you have to type tab or backspace at some point in your life 2011-09-18T10:13:18 hehe 2011-09-18T10:13:22 maybe i f up, idk, but i put in an end clause, than C-M-q and voila, its done 2011-09-18T10:14:24 another problem, is that python should reject beginning of line tabs 2011-09-18T10:14:39 it should be standard to accept only sapces 2011-09-18T10:15:00 i think even hard-ass pythoners agree with this 2011-09-18T10:15:02 dont you? 2011-09-18T10:15:07 no 2011-09-18T10:15:14 :D 2011-09-18T10:15:32 some people want tabs, even if it is against pep8 2011-09-18T10:15:52 against what? 2011-09-18T10:16:10 http://www.python.org/dev/peps/pep-0008/ 2011-09-18T10:16:57 forgive me if i dont read that page 2011-09-18T10:17:12 i guess it argues for my point right? 2011-09-18T10:17:47 McLeopold: did you use emacs with python? 2011-09-18T10:18:21 no, I haven't gotten into emacs yet 2011-09-18T10:18:30 i still find python editing unconfortable as long as i cant dynamically execute my code, and re-evaluate revious whole expressions 2011-09-18T10:18:41 IPython 2011-09-18T10:18:47 thats not good enough 2011-09-18T10:18:52 at least not the terminal version 2011-09-18T10:18:59 maybe the gui a bit better 2011-09-18T10:19:03 http://ipython.org/ 2011-09-18T10:19:09 what IS better, is pySclices 2011-09-18T10:19:20 "For new projects, spaces-only are strongly recommended over tabs." 2011-09-18T10:19:21 Slices 2011-09-18T10:19:33 and the other is sage's we interface 2011-09-18T10:19:37 I like my tabs 2011-09-18T10:19:42 McLeopold: have you heard of sagemath? 2011-09-18T10:20:25 its a glue-every-opensource-math-project-with-python-together CAS 2011-09-18T10:20:45 it has a very good notebook-like interface, that you access from the browser 2011-09-18T10:21:05 you have cells, you type in it, it can send that input to different interpreters 2011-09-18T10:21:16 python/maxima/sage/bash/mathematica 2011-09-18T10:21:38 and displays the results, inline graphics, inline 3d java visualization 2011-09-18T10:22:00 and you can save the notebooks, and manage them, and easily share them on the network if you want to 2011-09-18T10:22:13 it was quite good for python editing 2011-09-18T10:22:26 but ultimately i want to use emacs 2011-09-18T10:23:06 but i havent succeeded with all the dependencies yet, the py-shell throuws errors 2011-09-18T10:24:44 back to the indentation/syntax: one of the main reasons i really lile lisp, is that indentation doesnt change the semantics, its indentation is very canonical, and it doesnt have precedence rules! 2011-09-18T10:34:08 sorry, did i short-circuit something? 2011-09-18T10:35:47 *** GeorgeSebastian has joined #aichallenge 2011-09-18T10:35:47 *** GeorgeSebastian has joined #aichallenge 2011-09-18T10:35:51 ah, i forgot to add latex to the list of interpreters, sage notebook displays inline latex too 2011-09-18T10:38:19 *** ChrisH has joined #aichallenge 2011-09-18T10:56:25 @later tell janzert the sandbox compilation is what causes golang to fail, the temp _go_.6 file would be deleted between commands 2011-09-18T10:56:25 McLeopold: I think that worked... 2011-09-18T10:57:53 how does the c compilation work btw? if the compiler script looks for MyBot.c? 2011-09-18T11:01:36 I need to review that... 2011-09-18T11:03:32 it uses the TargetCompiler 2011-09-18T11:09:11 *** McLeopold has left #aichallenge 2011-09-18T11:20:12 *** Palmik has quit IRC (Remote host closed the connection) 2011-09-18T11:21:01 *** Palmik has joined #aichallenge 2011-09-18T11:33:06 *** sigh has quit IRC (Remote host closed the connection) 2011-09-18T11:41:35 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has joined #aichallenge 2011-09-18T11:56:23 *** dvladim has quit IRC (Ping timeout: 260 seconds) 2011-09-18T11:58:10 mleise :any idea abt the planned date to go official beat stage ? 2011-09-18T11:58:36 fix your keyboard! what did you ask? :p 2011-09-18T11:58:45 out of beta stage? 2011-09-18T11:58:53 hahaha 2011-09-18T11:58:59 start? 2011-09-18T11:59:05 (succint enough?) 2011-09-18T11:59:08 or: 2011-09-18T11:59:11 start when? 2011-09-18T11:59:22 next week as far as I know eashoka 2011-09-18T11:59:37 oh sorry. I meant when are we planning to reach official beta stage ? :-) 2011-09-18T11:59:53 but we are in the beta stage 2011-09-18T12:00:11 *** Accoun has quit IRC () 2011-09-18T12:00:15 in other words we reached it months ago 2011-09-18T12:01:03 oh yea. I red someone coining the 'official beta" phrase to mean everything running on final env but still at testing stage 2011-09-18T12:01:23 ok, then you know more than I do :) 2011-09-18T12:01:31 i know what you mean, lets just call this stage beta, the other no 2011-09-18T12:01:42 t* 2011-09-18T12:01:56 live-test ? 2011-09-18T12:02:02 armed/official/is on 2011-09-18T12:02:27 mleise: i think mcleo meant the short period after launch, when he fixes some thigns 2011-09-18T12:03:40 how does he know it will be short, har har. ok i understand. then this must be the alpha phase. ahh screw that wikipedia must help 2011-09-18T12:04:06 ah it will be a release candidate 2011-09-18T12:04:28 So next week would be the grand official launch ? (being programmers, everybody knows there could be fixes even after the launch) 2011-09-18T12:05:43 yea release candidate sounds much better :-) 2011-09-18T12:05:48 eashoka: This is how I understood it. 2011-09-18T12:06:08 eashoka: the launch means one thing: the contest will be hosted on another servers 2011-09-18T12:06:20 you get resources, registrations will go live too 2011-09-18T12:06:50 mcstar: what 'resources' and what 'registrations' that aren't there already? 2011-09-18T12:06:54 and surely, if some bugs arise, they will get fixed, before the end, when the final rankings will be made 2011-09-18T12:07:20 *** Accoun has joined #aichallenge 2011-09-18T12:07:40 mleise: there will be more workers available, abnd people will need to register again 2011-09-18T12:07:54 dont they? 2011-09-18T12:07:59 all accounts will be deleted 2011-09-18T12:08:02 yes, you are correct 2011-09-18T12:08:19 some people have multiple accounts that wouldn't be allowed anyway 2011-09-18T12:08:40 yeah, and the many starter bots.. 2011-09-18T12:09:14 but pls, do tell me, if there will be more computational resources availabl? 2011-09-18T12:10:38 mcstar: yes 2011-09-18T12:10:59 I'll name my bot ᓚᐂᓗ 2011-09-18T12:11:20 is that a mexican hat ? 2011-09-18T12:11:27 <_flag> Why is the turn limit on the server 255? 2011-09-18T12:11:51 _flag: i think they are testing trueskill, and they want fast matches 2011-09-18T12:11:59 (not sure) 2011-09-18T12:12:02 <_flag> Ah, okay 2011-09-18T12:12:06 eashoka: hmm... maybe then I'll use ᖧᐂᖨ 2011-09-18T12:12:36 <_flag> Because this way the matches between top bots rarely actually finish 2011-09-18T12:13:30 i hope i can name my bot A.Smith 2011-09-18T12:14:50 Is there a Neo to counter you? 2011-09-18T12:15:12 why mr anderson? why? why do you persist??? 2011-09-18T12:15:25 because i chose to!!!!! 2011-09-18T12:15:32 whoa, thats huge 2011-09-18T12:16:04 i can have my reasons to call my bot A.Smith, its very logical 2011-09-18T12:16:13 i cant see why i would call it Neo 2011-09-18T12:16:27 not that i dont like the character, i do very much 2011-09-18T12:16:35 but Smith is the grater antagonist ever 2011-09-18T12:16:40 greates 2011-09-18T12:16:42 t 2011-09-18T12:16:54 Your username must be longer than 6 characters and composed only of the characters a-z, A-Z, 0-9, '-', '_', and '.' 2011-09-18T12:17:01 eashoka seems to interfere with our keyboards 2011-09-18T12:17:12 my other bot is lore 2011-09-18T12:17:20 .Lore. 2011-09-18T12:17:27 for that reason i added the dots 2011-09-18T12:17:47 its 6, though, but it was accepted 2011-09-18T12:19:26 mleise : so there goes your mexican hats :-( I really like those. 2011-09-18T12:20:11 --007-- would be a cool id too 2011-09-18T12:20:28 is that the police? or is it 911? 2011-09-18T12:20:45 Bond. James Bond 2011-09-18T12:24:23 blow. job blow 2011-09-18T12:25:30 i just dont see why would that be cool 2011-09-18T12:25:38 :D 2011-09-18T12:25:59 *** amstan has joined #aichallenge 2011-09-18T12:25:59 *** ChanServ sets mode: +o amstan 2011-09-18T12:27:09 eashoka: you know im being sarcastic all the time, and i theoretize that A.Smith cant be topped 2011-09-18T12:28:12 *** cody__ has joined #aichallenge 2011-09-18T12:28:24 hihi: http://aichallengebeta.hypertriangle.com/profile.php?user=1 2011-09-18T12:30:23 i would laugh at rank 361 too 2011-09-18T12:31:24 mleise : woa!! is that you! 2011-09-18T12:31:28 ? 2011-09-18T12:31:45 eashoka: Just a quick db hack, yes it is me 2011-09-18T12:32:48 that's easily the coolest I have seen yet! 2011-09-18T12:33:11 uh.. ok thx 2011-09-18T12:39:08 mleise: i suggest you disable unicode 2011-09-18T12:39:25 do you really want to refer people by kanji charaters? 2011-09-18T12:39:33 *** Kingpin13 has joined #aichallenge 2011-09-18T12:40:43 no, I hate that with online games ^^ 2011-09-18T12:41:17 um.. you ... with that cryptic name, could you stop shooting your teammates? 2011-09-18T12:42:12 :) 2011-09-18T12:42:22 i just refer them as fucks 2011-09-18T12:42:33 he you f, stop teamkilling me you a.whole 2011-09-18T12:44:18 for some reason, workers are running blazing fast right now. a whopping 25 players per minute! 2011-09-18T12:44:57 i got to seet hat 2011-09-18T12:46:18 eashoka: keep in mind that games are cut at 255 turns 2011-09-18T12:48:30 *** eashoka has quit IRC (Ping timeout: 252 seconds) 2011-09-18T12:49:22 *** eashoka has joined #aichallenge 2011-09-18T12:49:46 lol the timeout almost matches the turn limit 2011-09-18T12:50:08 :) 2011-09-18T12:50:41 i dont really understand why is the timeout soo long? 2011-09-18T12:50:45 3 minutes? 2011-09-18T12:51:17 it happened many times, that i typed up quite some in 3 minutes, just for my friend to timout than of course when he reconnects al is gone 2011-09-18T12:51:18 That strikes me sometimes as well. The internet is so fast nowadays that you don't need that. 2011-09-18T12:51:28 cause my private channel is not logged :) 2011-09-18T12:53:16 mleise: do you communicate with friends through irc? or you use some other IM? 2011-09-18T12:53:35 others: ICQ and MSN networks 2011-09-18T12:53:53 logged of course ;) 2011-09-18T12:54:07 i tried pidgin with google-talk 2011-09-18T12:54:16 the channel was always noise 2011-09-18T12:54:23 so we only use irc atm 2011-09-18T12:54:34 noisy* 2011-09-18T12:55:13 (i dont mean the "channel" was noisy, of course our equipment was a cheapshit noisy thing) 2011-09-18T13:02:53 *** xathis has quit IRC () 2011-09-18T13:12:37 *** eashoka has quit IRC (Quit: Page closed) 2011-09-18T13:19:06 *** sigh has joined #aichallenge 2011-09-18T13:22:58 *** cody__ has quit IRC (Ping timeout: 252 seconds) 2011-09-18T13:29:54 *** Hammerok has joined #aichallenge 2011-09-18T13:31:42 *** dvladim has joined #aichallenge 2011-09-18T13:32:41 what's a fun random name for a mathematical object? 2011-09-18T13:32:48 i need to stick a name on something and it seems there isn't one 2011-09-18T13:33:02 antimatroid: lemur? 2011-09-18T13:33:08 in the case that i have the Cartesian product of two sets my object is a bijection 2011-09-18T13:33:56 i think THAT is an antimatroid 2011-09-18T13:34:17 so, lets just call it nick, for short 2011-09-18T13:34:46 say what? :P 2011-09-18T13:36:27 oh, you might be right aha 2011-09-18T13:36:36 :D 2011-09-18T13:37:02 it's a lot like it otherwise 2011-09-18T13:37:33 L is just one set right? I sort of have to take the alphabet as different for each place in the word 2011-09-18T13:37:45 and the words are fixed length 2011-09-18T13:39:13 matroid? 2011-09-18T13:39:24 there are too many unknows so that i really cant be more contructive 2011-09-18T13:39:37 antimatroid: problem with naming stuff after yourself is.. you might need to name something else after yourself later 2011-09-18T13:39:47 it gets weird when a lot of things are named after you 2011-09-18T13:39:53 amstan: i would never name something after myself 2011-09-18T13:39:59 so you'll have to make a wikipedia disambiguation page 2011-09-18T13:40:14 \o/ 2011-09-18T13:40:28 matroids and antimatroids are mathematical objects 2011-09-18T13:40:34 *** sir_macelon has quit IRC (Ping timeout: 252 seconds) 2011-09-18T13:40:42 i named myself after something rather than vice versa :P 2011-09-18T13:46:50 haha 2011-09-18T13:46:53 *** GeorgeSebastian has quit IRC (Read error: Connection reset by peer) 2011-09-18T13:46:56 just what i was talking about 2011-09-18T13:47:08 i wrote 5 lines, and he timed out 2011-09-18T13:47:57 its kind of funny that my slow adsl is more stable here on the countryside, than my friend's cable-net in the capital city 2011-09-18T13:51:13 *** arlequ1 has quit IRC (Ping timeout: 252 seconds) 2011-09-18T13:53:07 should a fair game be considered symmetric? 2011-09-18T13:53:29 or should a game be symmetric if each player is playing every other player symmetrically? 2011-09-18T13:54:24 deja vu 2011-09-18T13:54:35 i know 2011-09-18T13:54:43 i still don't have a clear answer :P 2011-09-18T13:54:48 :) 2011-09-18T13:54:53 http://pastebin.com/4tLWiM78 2011-09-18T13:55:02 did you talk to that physics professor you told me you will? 2011-09-18T13:55:04 that's an example of a fair game that i'm not sure of whether to call symmetric 2011-09-18T13:55:17 not sure what about? 2011-09-18T13:55:36 1 is the row player, 2 is the col player and 3 is the matrix player 2011-09-18T13:55:36 these things, mainly symmetry 2011-09-18T13:55:42 as far aas i can remember 2011-09-18T13:55:48 though, it was couple months ago 2011-09-18T13:55:55 but you never told me what happened 2011-09-18T13:56:12 you can find a symmetry of the game with (123) as the permutation of the players, but not for a transposition (12),(13),(23) of the players 2011-09-18T13:56:38 oh, i've solved most of my maths problems, i'm just caught on what to call thigns etc. now 2011-09-18T13:58:20 ill take a look at that, but i cant give advices and think about game theory at once 2011-09-18T13:58:36 "/onhold #aichannel 2011-09-18T14:03:01 i don't need it verified, i've checked :P 2011-09-18T14:03:47 wikipedia says " If the symmetry group of x is the trivial group then x is said to be asymmetric, otherwise symmetric.", but that would have a game be symmetric even if you could only swap 2 players and not all of them 2011-09-18T14:04:17 antimatroid: i didnt mean that, i would like to think thatover, but i cant atm 2011-09-18T14:04:52 :) fair enough 2011-09-18T14:07:44 *** GeorgeSebastian has joined #aichallenge 2011-09-18T14:10:38 *** jako has quit IRC (Ping timeout: 252 seconds) 2011-09-18T14:19:36 *** GeorgeSebastian has quit IRC (Ping timeout: 258 seconds) 2011-09-18T14:32:33 *** GeorgeSebastian has joined #aichallenge 2011-09-18T14:34:52 mcstar: openjdk and sun JDK are not the same 2011-09-18T14:35:08 delt0r_: the jvm? 2011-09-18T14:35:24 also called hotspot if im correct 2011-09-18T14:35:26 openjdk is a little behind and there are still some things that can't be OS 2011-09-18T14:35:43 yea JVM+standard libs 2011-09-18T14:36:08 for the most part openJDK works well and is only a little slower 2011-09-18T14:36:26 im just wondering what would be better for clojure, since it mostly uses the jvm, and some other code 2011-09-18T14:36:26 but there are some areas where it dies a horrible death 2011-09-18T14:36:44 I would always recommend the sun JVM 2011-09-18T14:37:23 you know recently happened that distros can no longer distribute it 2011-09-18T14:38:03 oh god not more RS bs? 2011-09-18T14:38:21 you know outside the US it is perfectly legal 2011-09-18T14:38:23 bs i know, whats rs? 2011-09-18T14:38:32 Richard Stallman 2011-09-18T14:38:36 no 2011-09-18T14:38:44 java changed the licence 2011-09-18T14:38:59 openjdk 2011-09-18T14:39:00 or sun 2011-09-18T14:39:01 sun jre jdk disappeared from arch's repositories 2011-09-18T14:39:16 sry, i mean sun of course 2011-09-18T14:39:22 not java 2011-09-18T14:39:23 since the sun jre has never been distributed in any of the linuxs i am exposed 2011-09-18T14:39:24 to 2011-09-18T14:39:25 :) 2011-09-18T14:39:33 *** sigh has quit IRC (Remote host closed the connection) 2011-09-18T14:40:08 ok... non story... most already didn't --apple are going to stop too 2011-09-18T14:40:36 but then its like they have a lisp VM or anything either... installing one is not so hard. 2011-09-18T14:40:46 not like 2011-09-18T14:41:16 i know its not hard, thats not the point 2011-09-18T14:41:30 the point is that you have to update it manually 2011-09-18T14:41:46 at anyrate I think you could easily use openjdk for ants. 2011-09-18T14:41:47 fortunately, its available from AUR, a user maintaned repo 2011-09-18T14:42:11 i wasnt talking about ants, in particular 2011-09-18T14:42:20 but i think they already use the openjdk thing 2011-09-18T14:42:21 but it would be more painless to use the sun one 2011-09-18T14:43:45 also i install a jre with my programs to solve the "what jvm are you running" 2011-09-18T14:44:05 so my install does not look like a java app 2011-09-18T14:47:44 mcstar: also java is stupid easy to install --user space or global install..even without sacrificing to the package management gods 2011-09-18T14:48:08 (IMO package management is a joke, on all OSes ) 2011-09-18T14:48:51 not really 2011-09-18T14:49:03 note the IMO part ;) 2011-09-18T14:49:07 at least my fixpoint in package-manager space was Arch's pacman 2011-09-18T14:49:11 you won't convince me otherwise :D 2011-09-18T14:49:12 yeah 2011-09-18T14:49:15 noted 2011-09-18T14:49:19 prefix mine with imho too 2011-09-18T14:49:35 I don't bother with humble 2011-09-18T14:49:47 oh, just a typo 2011-09-18T14:49:51 lol 2011-09-18T14:50:16 so whats the word... will ants go live this week? next week? 2011-09-18T14:50:20 what is outstanding 2011-09-18T14:50:38 of course tha packaga manage ris not everything, you need people who can sort the packages and dependencies sensibly 2011-09-18T14:51:08 it is outstanding how many people are still asking this question 2011-09-18T14:51:24 i guess next week it gonna be ... 2011-09-18T14:51:27 wait for it... 2011-09-18T14:51:37 what was a saying? 2011-09-18T14:51:41 i* 2011-09-18T14:52:19 soon this channel will leak from the pressure of my humor 2011-09-18T14:57:02 contestbot: later tell McLeopold where or how are the temporary files being deleted? 2011-09-18T14:57:02 janzert: Yes master! 2011-09-18T14:58:58 contestbot: later tell McLeopold or are these files not created within the bots directory but live in /tmp or something? I can understand those not be preserved 2011-09-18T14:58:58 janzert: I come to serve. 2011-09-18T15:03:39 *** boegel has joined #aichallenge 2011-09-18T15:12:54 *** xathis has joined #aichallenge 2011-09-18T15:13:20 *** Hammerok has quit IRC (Ping timeout: 252 seconds) 2011-09-18T15:19:52 *** xerus has joined #aichallenge 2011-09-18T15:21:19 *** trololo has joined #aichallenge 2011-09-18T15:22:48 *** xerus has quit IRC (Client Quit) 2011-09-18T15:22:51 *** boegel has quit IRC (Quit: Leaving) 2011-09-18T15:24:58 *** delt0r_ has quit IRC (Ping timeout: 252 seconds) 2011-09-18T15:30:59 *** McLeopold has joined #aichallenge 2011-09-18T15:31:57 *** TTE has joined #aichallenge 2011-09-18T15:33:11 janzert: it may be that they are created in the wrong dir, or my check for the files is done in the wrong dir 2011-09-18T15:33:42 *** TTE has quit IRC (Client Quit) 2011-09-18T15:35:12 aichallenge: Scott Hamilton epsilon * r2a423a4 / (website/api_compile_result.php website/sql.php): remove worker id requirement from compile status posting - http://git.io/sskMqg 2011-09-18T15:38:38 *** delt0r_ has joined #aichallenge 2011-09-18T15:43:01 *** Accoun has quit IRC () 2011-09-18T15:43:11 *** nux67 has quit IRC (Quit: Page closed) 2011-09-18T15:48:12 *** Palmik has quit IRC (Read error: Connection reset by peer) 2011-09-18T15:51:26 McLeopold: the distribution of games played per map seems quite odd. For example since the turn limit has been set to 255 map 42 (symmetric_36) has had 2 games played on it, map 21 (symmetric 10) has had 215 games and map 53 (symmetric_26) has had 225 games 2011-09-18T15:51:38 all three maps are at priority 5 2011-09-18T15:52:53 oops, map 21 is symmetric 14 2011-09-18T15:55:46 *** Accoun has joined #aichallenge 2011-09-18T15:59:40 *** gobang has quit IRC (Read error: Connection reset by peer) 2011-09-18T16:00:52 *** testing has joined #aichallenge 2011-09-18T16:01:03 hey 2011-09-18T16:01:46 any idea on when the actual event might begin ?? 2011-09-18T16:03:21 *** onensora has joined #aichallenge 2011-09-18T16:03:56 testing: sometime 2011-09-18T16:04:21 now >50% likely within the next month 2011-09-18T16:04:29 I think ;) 2011-09-18T16:04:52 nice to hear that .... looking forward to it 2011-09-18T16:05:27 maan, i posted on the interplanetary bulletin board, the IBB, that it starts tomorrow 2011-09-18T16:05:46 mcstar: why would you do that? 2011-09-18T16:06:07 to notify mars, and other planets 2011-09-18T16:06:08 mcstar: like.. we had false alarms like that a few times during the summer 2011-09-18T16:06:18 they need time for the signal to reach them 2011-09-18T16:06:20 man.. the martians are going to be so pissed 2011-09-18T16:06:40 :) 2011-09-18T16:06:46 obviously i didnt post anything 2011-09-18T16:11:48 *** dvladim has quit IRC (Ping timeout: 260 seconds) 2011-09-18T16:12:34 amstan: when did you have false alarms? 2011-09-18T16:12:39 i think i missed that 2011-09-18T16:12:52 did somewhere appear that the contest is srtating? 2011-09-18T16:13:13 starting* 2011-09-18T16:20:08 mcstar: yeah, well, we were having pre launch meetings in july 2011-09-18T16:21:05 janzert: I'm still working on a fix for the matchmaking to even out the player games, I can look at the map selection after that 2011-09-18T16:22:41 janzert: also, I believe I'm ignoring priority now 2011-09-18T16:23:52 just having priority as a binary setting (>0 are used) should be fine 2011-09-18T16:24:55 McLeopold: also could you add a field to the map table to record the last time the turn limit was adjusted for the map (something like max_turns_adjusted maybe) 2011-09-18T16:25:05 yes, I'm checking for >= 0 2011-09-18T16:25:24 janzert: you can get that info from the game table 2011-09-18T16:25:36 kinda 2011-09-18T16:25:37 I don't think so 2011-09-18T16:25:55 I'm limiting it to games with the same turn limit right now 2011-09-18T16:26:10 but that doesn't cover if the turn limit is set back to a previous value 2011-09-18T16:26:17 which I kind of want to do 2011-09-18T16:26:33 nope, that wouldn't work then 2011-09-18T16:28:50 aichallenge: McLeopold epsilon * r2a3cf95 / sql/0_schema.sql : add timestamp field to map table - http://git.io/ICxG0Q 2011-09-18T16:29:00 *** nplus has quit IRC (Read error: Connection reset by peer) 2011-09-18T16:29:56 *** nplus has joined #aichallenge 2011-09-18T16:30:00 *** nplus has joined #aichallenge 2011-09-18T16:30:48 janzert: done 2011-09-18T16:32:02 Thanks. I knew the changes needed to the schema but I'm not sure how to apply that to the live database. 2011-09-18T16:32:32 I just issue alter table commands 2011-09-18T16:32:57 so far I haven't blown up anything 2011-09-18T16:33:01 too much 2011-09-18T16:33:05 ok, so there isn't an automatic process 2011-09-18T16:33:10 hehe 2011-09-18T16:33:15 no 2011-09-18T16:33:27 I could do it the right way, but that is so much more work :) 2011-09-18T16:33:33 *** Naktibalda has joined #aichallenge 2011-09-18T16:33:37 :) 2011-09-18T16:33:55 Once we launch, I'll check in proper alter table statements and version the database. 2011-09-18T16:35:25 janzert: was your map stats for all time? 2011-09-18T16:35:38 *** Naktibalda has quit IRC (Client Quit) 2011-09-18T16:35:41 limited to games with max_turns of 255 2011-09-18T16:35:47 okay 2011-09-18T16:35:59 that is bad :( 2011-09-18T16:36:14 oh, my logic doesn't though 2011-09-18T16:37:31 yeah, there are half a dozen maps with single digit games and 3 with over 200 :( 2011-09-18T16:41:50 Add some fun maps while you are at it. Mazes, pictograms, etc... where is the map editor *sigh* 2011-09-18T16:42:58 They aren't good gameplay wise, but a themed map now and then would be a real highlight. 2011-09-18T16:48:52 janzert: it looks like for all time the matchmaker is doing a good job and keeping them even, given the fact that for 2 months it was out of wack 2011-09-18T16:49:28 ok 2011-09-18T16:49:44 we just need a way to reset the game count to a point in time 2011-09-18T16:49:47 but then the question is why are they so out of whack for the last week(?) 2011-09-18T16:50:10 ahh, the matchmaker is still looking at all time numbers? 2011-09-18T16:50:13 because the algo is trying to keep stuff even, and a bunch a maps where played too much before 2011-09-18T16:50:16 yes 2011-09-18T16:50:21 ok 2011-09-18T16:50:41 how did you do finals last year? 2011-09-18T16:50:48 did you wipe the database? 2011-09-18T16:50:48 yeah, we need a way to limit it to look at things only after a certain point 2011-09-18T16:51:14 I moved all the game information from before the finals into archive tables 2011-09-18T16:51:27 so basically from the point of view of the matchmaking system, yes 2011-09-18T16:51:27 maybe a stored procedure to do that would be nice... 2011-09-18T16:52:14 adding a timelimit may be less disruptive? 2011-09-18T16:52:22 whichever is easier I suppose 2011-09-18T16:56:14 well, I would want something that works for the finals 2011-09-18T16:56:45 archiving makes queries faster, so I'll go with that 2011-09-18T16:56:59 but it disrupts looking and old games... 2011-09-18T16:57:03 yes, I was imagining the timelimit would be set to the start of the finals 2011-09-18T16:57:36 right, any pages that refer to games need to be modified to query the archive table if they don't find the game in the primary table 2011-09-18T16:57:55 which is why I thought the timelimit might be less disruptive 2011-09-18T16:58:12 so whichever way you want to do it :) 2011-09-18T16:58:38 this would only affect matchmaking? 2011-09-18T16:59:02 I can't think of anything else offhand it should effect 2011-09-18T16:59:37 I guess leaderboard generation but that is completely incremental so shouldn't need it 2011-09-18T16:59:52 right, submissions can be reset if we want 2011-09-18T17:00:17 just game counts in the rankings table 2011-09-18T17:00:23 yep, I believe reseting mu and sigma should be all that is needed there 2011-09-18T17:00:57 hmm, yeah 2011-09-18T17:01:18 I'm storing those in the submission table right now, so that the rankings are faster 2011-09-18T17:01:25 although I think I'd like for the rate field to go away at least in the default view before we go live 2011-09-18T17:01:33 it's duplicate data, but it allows us to reset the tally 2011-09-18T17:01:40 ahh 2011-09-18T17:01:53 not at all? 2011-09-18T17:02:05 I mean, you don't want to display that info? 2011-09-18T17:02:23 the rate information would be nice to still have for admins or those interested to view, but I think it's bad for the normal view 2011-09-18T17:02:42 ok 2011-09-18T17:04:31 *** testing has quit IRC (Quit: Page closed) 2011-09-18T17:05:10 *** trololo has quit IRC (Ping timeout: 252 seconds) 2011-09-18T17:07:46 *** GeorgeSebastian has quit IRC (Quit: Leaving) 2011-09-18T17:09:34 *** xathis has quit IRC (Ping timeout: 252 seconds) 2011-09-18T17:17:12 *** nux67 has joined #aichallenge 2011-09-18T17:17:43 *** foRei has quit IRC (Read error: Connection reset by peer) 2011-09-18T17:17:50 *** foRei has joined #aichallenge 2011-09-18T17:18:33 *** mcstar has quit IRC (Quit: WeeChat 0.3.5) 2011-09-18T17:38:06 *** mceier has quit IRC (Quit: leaving) 2011-09-18T17:42:28 *** mathis has joined #aichallenge 2011-09-18T17:47:25 janzert: I have an idea for a new matchup generator 2011-09-18T17:49:26 hmm, do I want to hear it... ;) 2011-09-18T17:49:31 ok... 2011-09-18T17:49:41 select a seed, last to have played 2011-09-18T17:50:06 select 10% of the other bots with the best match quality, according to trueskill 2011-09-18T17:50:21 pick the player that has played the least with the seed 2011-09-18T17:50:58 for the 3rd player, again select 10%, but the match quality is calculated for all 3 2011-09-18T17:51:23 again, select the player that has played the least, but with both player 1 & 2 2011-09-18T17:51:31 repeat until you have enough 2011-09-18T17:51:47 I think this removes the need for my "no repeat" logic 2011-09-18T17:52:08 And this will solve the problem of match quality favoring those with low sigma/high game counts 2011-09-18T17:52:46 all based off user_id, so submitting new bots won't increase your chances of playing again 2011-09-18T17:53:08 hmm, let me think about it for a minute, initial is impression isn't bad though ;) 2011-09-18T17:53:42 if you suddenly submit a new bot that improves, putting you in a new percetile of the field, you'll start to get picked alot, which I think is good 2011-09-18T17:54:12 the older bots will want to play against the newer good bot right away 2011-09-18T17:54:36 *** eashoka has joined #aichallenge 2011-09-18T17:55:37 *** ltriant has joined #aichallenge 2011-09-18T17:55:37 am I write that increasing the selection percentage to 100% would bring cause this to be a round robin? 2011-09-18T17:55:44 (that's a good thing by the way) 2011-09-18T17:55:48 :) 2011-09-18T17:55:52 I believe so 2011-09-18T17:56:00 heh write->right 2011-09-18T17:56:02 for 2-player games, yes 2011-09-18T17:56:08 * janzert nods 2011-09-18T17:56:17 *** Kingpin13 has quit IRC (Quit: quit) 2011-09-18T17:56:21 3 player might be skewed a little, but it would remain fairly even 2011-09-18T17:57:10 tweaking the percentage will be the trick 2011-09-18T17:57:51 we could even base it off your ranking 2011-09-18T17:58:07 top bots stay at 5%, bottom bots increase to 20% or something 2011-09-18T17:58:13 it's too bad the cutoff can't be more probababilistic 2011-09-18T17:58:50 how so? 2011-09-18T17:59:25 we could limit it to match quality within 5% of best match quality or something 2011-09-18T17:59:33 i.e. the closest in skill bots would be included for consideration 100% of the time and the chances would go down in a pareto distribution as they got further away 2011-09-18T17:59:53 we can do that 2011-09-18T18:00:31 ahh, I thought that would be quite difficult to do in sql 2011-09-18T18:00:35 so 2011-09-18T18:00:37 no 2011-09-18T18:00:54 I just need a way to calc the pareto 2011-09-18T18:03:13 hmm, it was encoded in the last contest code at one point 2011-09-18T18:03:23 really? 2011-09-18T18:03:37 but I see the final version just used the one built into python :} 2011-09-18T18:03:48 pretty sure it was anyway 2011-09-18T18:07:29 percent of highest match quality would probably be good enough 2011-09-18T18:07:40 10% of total field might be good enough now 2011-09-18T18:08:01 yeah, it just misses out on the occasional outlier game 2011-09-18T18:08:10 but probably works out ok in practice 2011-09-18T18:08:19 where a really good bot plays a bad bot? 2011-09-18T18:08:31 right 2011-09-18T18:08:40 so, I'm okay with that not happening 2011-09-18T18:09:05 they should ideally happen very occasionally 2011-09-18T18:09:26 well, I could muliply match quality by a random number and sort? 2011-09-18T18:10:12 I still need to take a top % and do the least played logic to avoid the game counts not being even 2011-09-18T18:11:10 really pairings outside of the 10% range should be less than 1 in a couple hundred anyway so is probably fine not to worry about them 2011-09-18T18:11:27 I'll just to the multiply anyway 2011-09-18T18:12:19 isn't your method just making the border 'fuzzier' not actually occasionally pulling in an opponent from far beyond the nominal cutoff? 2011-09-18T18:12:39 it would be giving them some chance 2011-09-18T18:12:55 hmm, mayby not 2011-09-18T18:13:14 I guess what is the range and distribution of your random numbers? 2011-09-18T18:13:20 0-1 2011-09-18T18:13:29 uniform 2011-09-18T18:13:41 *** foRei has quit IRC (Quit: Bye) 2011-09-18T18:14:26 hmm, seems like it would be too strong and dominate the actual quality measure 2011-09-18T18:14:53 I'm thinking it wouldn't be strong enough now 2011-09-18T18:15:55 take the numbers 1-100 and multiply by a random 0-1 and see how often 1 is chosen, probably not much 2011-09-18T18:16:21 chosen being 1*rand() is greater than all the others numbers times their own rand 2011-09-18T18:21:07 the average of the result will half of the original average of course 2011-09-18T18:23:03 but the range covered will decrease in size much less and my math ability to reason about random uniform distributions isn't quite good enough to say how much on average ;) 2011-09-18T18:23:32 I'm writing a python test right now ;) 2011-09-18T18:23:39 i.e. the mean will change less than the median 2011-09-18T18:23:47 err more 2011-09-18T18:24:22 or not 2011-09-18T18:24:30 hehe 2011-09-18T18:26:41 in any case I don't see how it would form any sort of a power law distribution for the opponents selected 2011-09-18T18:26:49 which is what I'm really after 2011-09-18T18:30:55 *** amstan has quit IRC (Ping timeout: 260 seconds) 2011-09-18T18:35:35 janzert: how's this? http://pastebin.com/u7xXXbJS 2011-09-18T18:35:56 gives me unknown paste ID 2011-09-18T18:36:04 http://pastebin.com/xAg5TLyQ 2011-09-18T18:36:07 weird 2011-09-18T18:36:12 wierd 2011-09-18T18:36:22 the oatmeal is going to get me... 2011-09-18T18:39:58 the top end looks rather linear but the tail looks good 2011-09-18T18:40:27 so, I think it would work 2011-09-18T18:40:44 sort by match quality * random number 2011-09-18T18:42:01 you still have a 1 in 20 chance if your match quality is 20% lower than the best quality to be considered at the top 2011-09-18T18:42:26 I almost think 10% might be too large but yeah this will probably work 2011-09-18T18:42:38 let me replot my graph in percentages 2011-09-18T18:42:41 I get that from the 80 line having 49 picks 2011-09-18T18:42:45 oops, 1/200 2011-09-18T18:43:48 with limiting the field, then choosing the least played, if a low quality happened to get in the top X, then it will probably be chosen 2011-09-18T18:46:51 ok, I guess 10% looks fine 2011-09-18T18:46:54 given 4000 submissions, in 1000 games you would probably only play within 100 ranks up or down 2011-09-18T18:47:59 hmm, really? 2011-09-18T18:48:10 don't forget the surprise some people had with planet wars when they had to play some lower ranked players 2011-09-18T18:48:10 well, no... 2011-09-18T18:48:23 it's match quality, it's not linear 2011-09-18T18:48:34 *** mathis has quit IRC (Ping timeout: 252 seconds) 2011-09-18T18:48:35 although i don't imagine we'll have the same problems with ants, it's harder to go too hard 2011-09-18T18:48:36 it seems to me 95% of games will have the first opponent with 15% of match quality 2011-09-18T18:48:43 if I'm looking at this right 2011-09-18T18:48:46 antimatroid: that's why were discussing this, to include a small percent of those games 2011-09-18T18:49:17 yes, but that may be 25% of the field 2011-09-18T18:49:40 *** FireFly has quit IRC (Quit: FireFly) 2011-09-18T18:50:01 I think match quality falls off quite fast 2011-09-18T18:50:31 anyway I'd say go ahead and give it a try and see how it looks 2011-09-18T18:50:59 ok 2011-09-18T18:51:08 my feeling is that 10% may be too large but we can look at it 2011-09-18T18:52:52 also, to see if it's actually working well I think we need to increase turn limits again so that games are actually giving reasonable results 2011-09-18T18:53:49 I don't think that matters for matchmaking, does it? 2011-09-18T18:53:54 and to do that I've got the automatic adjuster almost ready :) 2011-09-18T18:53:58 are you thinking the mu's should be more spread out? 2011-09-18T18:55:02 well the ranking system doesn't really have anything to work with when the results aren't very good indicators of skill 2011-09-18T18:55:16 btw, run the opponent sp on the server to see picking to X percent and sorting by least played 2011-09-18T18:55:23 call opponent(); 2011-09-18T18:55:34 and we can't look at the game quality the matchmaker is giving us when the rankings it is using aren't any good 2011-09-18T18:55:41 255 is enough to know who can gather well 2011-09-18T18:56:25 but, it is fast enought, so 500 may still give us enough results 2011-09-18T18:56:40 6.8 gpm 2011-09-18T18:57:02 don't adjust it manually yet 2011-09-18T18:57:47 I need to test my auto adjuster ;) 2011-09-18T19:05:57 aichallenge: janzert epsilon * redb69de / manager/adjust_turnlimits.py : Add script to adjust map turn limits - http://git.io/xProuw 2011-09-18T19:09:15 *** amstan has joined #aichallenge 2011-09-18T19:09:15 *** ChanServ sets mode: +o amstan 2011-09-18T19:25:53 McLeopold: I have several hours of free time right now that I could use to help resolve the Go compile problem. Let me know how I can help. 2011-09-18T19:35:16 janzert: I have several hours of free time right now that I could use to help resolve the Go compile problem. Let me know how I can help. 2011-09-18T19:35:22 *** eashoka has quit IRC (Quit: Page closed) 2011-09-18T19:52:30 *** olexs has quit IRC (Quit: Leaving.) 2011-09-18T20:07:26 ChrisH: do you use ubuntu? 2011-09-18T20:07:51 Yes, in a VM, I'm running 10.10 currently. 2011-09-18T20:08:11 okay, if you get the repo you can use worker.py to compile in a directory 2011-09-18T20:08:16 I've been looking at compiler.py more and I think I see the problem you found. 2011-09-18T20:08:22 but, I think the issue is with our jail users 2011-09-18T20:08:27 I agree. 2011-09-18T20:08:51 Is it possible we can setup ExternalCompiler with 2 commands separated by a semicolon? 2011-09-18T20:09:07 I think the calls to check_path(vglob, cmd_errors) are looking in bot_dir instead of in the jail dir 2011-09-18T20:09:08 we are passing is a list of tokens to _run_cmd 2011-09-18T20:09:15 hmm... 2011-09-18T20:09:25 could be, but it works without the jail users 2011-09-18T20:09:27 before the call to box.retrieve 2011-09-18T20:09:39 or did it? 2011-09-18T20:09:51 yes, because House just does it's work in bot_dir directly 2011-09-18T20:11:10 so, the check_path could be with CD(bot_dir): check_path() 2011-09-18T20:11:15 it is touchy though, because if ExternalCompiler is running with separate=True, then it is doing multiple commands in a for loop and checking the vglob after each, and I don't think you want to do a box.retrieve after each call 2011-09-18T20:11:57 I think the check_path should be passed a directory to look in, and ExternalCompiler should get that directory from the sandbox root. 2011-09-18T20:12:10 go would not use the separate flag 2011-09-18T20:12:30 no it doesn't, so this only matters if another language starts using vglobs 2011-09-18T20:12:49 btw, it stands for verify globs 2011-09-18T20:12:56 ok, 2011-09-18T20:13:18 But do you really need to do that check? 2011-09-18T20:13:39 I mean, when the next call is made for the linker can't it just report a file not found error 2011-09-18T20:14:04 I would want the stdout of the first command to be reported 2011-09-18T20:14:08 so just take out the verify glob check and let the intermediate files get pulled back (if they exist) by box.retrieve 2011-09-18T20:14:17 it has the info we really want to send to the user 2011-09-18T20:14:26 fair enough 2011-09-18T20:14:52 the vglobs was to prevent stdout from being captured unless we really knew we wanted it 2011-09-18T20:14:53 in that case fix check_path to look in the jail root dir instead of the bot_dir 2011-09-18T20:15:16 because the compiler is working in the jail root dir 2011-09-18T20:15:31 that's what with CD() would do 2011-09-18T20:17:10 but ExternalCompiler calls CD(bot_dir) and then uses _run_cmd() to spawn a subprocess in the jail dir, the child process doesn't affect the PWD of the parent, so I would think ExternalCompiler is still in bot_dir when it makes the call to check_path() 2011-09-18T20:18:58 maybe sandbox could have a check_path? 2011-09-18T20:19:05 Yes. 2011-09-18T20:19:10 That should work. 2011-09-18T20:19:45 There is a member of sandbox holding its working dir 2011-09-18T20:21:16 Would it be easy for me to configure my Ubuntu to use the secure Jail implementation of sandbox to verify and fix this issue. 2011-09-18T20:21:20 ? 2011-09-18T20:21:42 *** bhasker has joined #aichallenge 2011-09-18T20:21:53 morning 2011-09-18T20:22:14 It doesn't have to actually be secure, just emulate the behavior by using separate directories and subprocesses. 2011-09-18T20:23:23 ChrisH: if you can get a fix, I'll test it 2011-09-18T20:23:39 janzert does the sandboxing, so he has to approve 2011-09-18T20:23:54 I'm willing to do that, but I don't know what I would need to do to set up a proper test enivonment. 2011-09-18T20:24:08 Oh, I see, I give you the code and you test. 2011-09-18T20:24:09 ? 2011-09-18T20:24:50 I could probably send you a patch file, would that work? 2011-09-18T20:25:28 *** onensora has quit IRC () 2011-09-18T20:38:47 ChrisH: we are on github, so you can fork and issue a pull request 2011-09-18T20:39:04 to set up an environment, you can setup another ubuntu vm and run setup_worker 2011-09-18T20:39:12 we are on 11.x 2011-09-18T20:39:32 mcleopold are we closer to launching the contest? 2011-09-18T20:39:39 McLeopold: Ok, I may need some hand holding on the git stuff, I'm mostly a subversion guy. 2011-09-18T20:39:39 bhasker: yes, very close 2011-09-18T20:39:49 McLeopold: how's the death star plans going? 2011-09-18T20:40:19 ChrisH: i can do git handholding 2011-09-18T20:40:31 well, there's this small hole directly into the core, but it's probably not noticable, and we've gone out to bid for a contractor to fix it 2011-09-18T20:40:32 @amstan: Great! 2011-09-18T20:40:33 ChrisH: User error, it's not my fault. 2011-09-18T20:40:50 btw, @ stuff is commands for the bot, just call me amstan 2011-09-18T20:41:11 amstan: ok, that's confusing 2011-09-18T20:41:40 I agree, I would prefer ~ 2011-09-18T20:42:22 amstan: don't know if you are up to date, but I think McLeopold and I have a fix for the Go compiler problem, but we want to test it before submitting a change request to janzert 2011-09-18T20:43:25 So, I have an edit to two python source files that I need to hand off to McLeopold. 2011-09-18T20:45:04 McLeopold: it's going to take me some time to set up an Ubuntu 11 VM, do you think that is required, or would giving you my code changes be faster? 2011-09-18T20:45:17 I can setup one up to test 2011-09-18T20:45:50 McLeopold: so you want ~? 2011-09-18T20:45:55 for @config reply.whenAddressedBy.chars 2011-09-18T20:46:03 yep 2011-09-18T20:46:15 some people like doing @amstan stuff 2011-09-18T20:46:16 McLeopold: Ok, so what is the best way to get you the code? 2011-09-18T20:46:20 it's happened more than once 2011-09-18T20:46:26 what about other ppl who are used to @later tell? 2011-09-18T20:46:42 McLeopold: i could also solve it by deopping myself 2011-09-18T20:46:48 ChrisH: fork the repo on github, then clone locally 2011-09-18T20:47:02 McLeopold: Ok, sounds easy. 2011-09-18T20:47:04 amstan: probably not worth changing now 2011-09-18T20:47:25 or, setup contest bot to respond with "use ~" for a time 2011-09-18T20:47:42 yeah, that's hard 2011-09-18T20:49:05 amstan: are the ec2 instances 32 or 64? 2011-09-18T20:49:20 McLeopold: 64 2011-09-18T20:49:21 McLeopold: why? 2011-09-18T20:49:28 I'm setting up a test worker 2011-09-18T20:49:44 11.04? 2011-09-18T20:49:48 yes 2011-09-18T20:50:25 I've become fond of arch lately 2011-09-18T20:57:16 *** McLeopold has left #aichallenge 2011-09-18T20:58:52 thats cool mcleopold looking forward to it 2011-09-18T21:04:34 amstan: I need a little git help. I've forked on github, cloned my fork on my machine, and done the required code changes. 2011-09-18T21:04:53 amstan: now I think I need to do an add or commit or something, but not sure 2011-09-18T21:04:53 ChrisH: ok? 2011-09-18T21:05:09 ChrisH: git status gives you the changes 2011-09-18T21:05:15 you do git add" 2011-09-18T21:05:17 ChrisH: "git add [files]", then "git commit", then "git push" 2011-09-18T21:05:17 amstan: yes 2011-09-18T21:05:21 git add for each file 2011-09-18T21:05:21 *** McLeopold has joined #aichallenge 2011-09-18T21:05:23 then git commit 2011-09-18T21:05:24 and git push 2011-09-18T21:05:43 do git status just before you commit to make sure you don't commit things you don't want to 2011-09-18T21:09:46 amstan: ok git push completed. 2011-09-18T21:09:52 janzert: you around? 2011-09-18T21:09:52 link? 2011-09-18T21:10:05 McLeopold: just got back 2011-09-18T21:10:15 I'm setting up a new ubuntu worker locally 2011-09-18T21:10:21 ChrisH: Once you git-push to GitHub, you can issue a pull request to the original project. 2011-09-18T21:10:28 should I create a contest user, or does the script do that? 2011-09-18T21:10:58 which script? ;) 2011-09-18T21:11:02 I'm in the setup portion of ubuntu 2011-09-18T21:11:08 worker_setup.py does not 2011-09-18T21:11:09 McLeopold: i think you do, then sudo from inside the user when you start the install 2011-09-18T21:11:10 setup_worker 2011-09-18T21:11:13 So, I browse back to the main aichallenge project on github and do a pull request? 2011-09-18T21:11:29 ChrisH: yes, something like that 2011-09-18T21:11:50 ChrisH: i think you can do it from your repo actually 2011-09-18T21:11:57 on top there should be a pull request button 2011-09-18T21:12:05 yes, it looks like I have to 2011-09-18T21:12:34 Ok, pull request submitted. 2011-09-18T21:12:47 worker_setup.py expects to be run as root, but the contest user and the contest files should already be in place 2011-09-18T21:13:05 contest files? you mean just cloning the repo? 2011-09-18T21:13:14 or pulling the tarball 2011-09-18T21:13:15 Now, let's hope it works. :) 2011-09-18T21:13:41 McLeopold: either would work 2011-09-18T21:14:00 the actual workers pull and unpack the tarball 2011-09-18T21:14:10 but that is just an exported repository 2011-09-18T21:14:15 how do you setup a blank password? 2011-09-18T21:14:39 adduser --disabled-password --gecos "" contest 2011-09-18T21:14:40 is the tarball the whole thing, or just a piece? 2011-09-18T21:14:53 the tarball is the whole repository 2011-09-18T21:15:24 website/worker_init.py is the script that sets up the worker to the point of handing it over to setup/worker_setup.py 2011-09-18T21:15:50 so I can curl | sh that? 2011-09-18T21:15:56 basically all it does is create the contest user and grab and unpack the tarball 2011-09-18T21:16:04 yes, you could 2011-09-18T21:16:15 so, on ec2, you have the initial user called ubuntu? 2011-09-18T21:16:36 that's the way the ubuntu images are setup, yes 2011-09-18T21:16:54 janzert: hmm, is there no way to have common code for the 2 classes(House and Jail) merged? maybe make them inheret another class? 2011-09-18T21:16:55 i.e. it's not something I configure 2011-09-18T21:17:19 amstan: they used to but it was more of a mess 2011-09-18T21:17:36 there isn't much duplicated code, just a common api 2011-09-18T21:19:04 McLeopold: oh, worker_init.py expects the worker_api_key as an argument 2011-09-18T21:19:24 amstan: how a about a base class that has a bunch of raise "Not Implemented" that they both inherited? 2011-09-18T21:19:38 McLeopold: well, it's kinda late now, but i'm pondering 2011-09-18T21:19:41 aka, python does not do abstract 2011-09-18T21:19:45 McLeopold: why bother? 2011-09-18T21:19:56 janzert: exactly 2011-09-18T21:20:05 think of it as an example of duck typing 2011-09-18T21:20:40 jbroman: to what extent does your zeta worker code cover what's done with the new epsilon worker? 2011-09-18T21:20:56 is zeta been touched? 2011-09-18T21:20:59 at all? 2011-09-18T21:21:10 not since march 2011-09-18T21:21:23 epsilon is the new beta 2011-09-18T21:21:24 jbroman did some worker code on it 2011-09-18T21:21:25 zeta 2011-09-18T21:21:30 and really not since february 2011-09-18T21:21:49 I think zeta should be an epsilon to django port, the worker stuff is solid 2011-09-18T21:22:25 McLeopold: he seems to have documented it nicely too: https://github.com/aichallenge/aichallenge/tree/zeta/worker 2011-09-18T21:24:09 looks like what was used for planetwars? 2011-09-18T21:25:15 i.e. it doesn't use a chroot root environment, just uses locked down users on the system 2011-09-18T21:25:41 the s3 stuff might be nice 2011-09-18T21:25:49 err, chroot sandbox environment 2011-09-18T21:26:13 but then we don't need amazon if we don't want 2011-09-18T21:27:05 McLeopold: how's your worker going? is it for testing Chris's stuff? 2011-09-18T21:27:10 yes 2011-09-18T21:27:11 close 2011-09-18T21:28:02 McLeopold: link if you need it: https://github.com/aichallenge/aichallenge/pull/254 2011-09-18T21:31:35 how does the check_path(vglob stuff in ExternalCompiler ever work? 2011-09-18T21:32:12 janzert: I believe never, but since it is only used by Go lang.... never means not for Go 2011-09-18T21:32:13 it looks like it's checking for the vglobs in whatever directory the compiler is run from 2011-09-18T21:32:29 amstan: Haven't tried to integrate anything with epsilon. 2011-09-18T21:32:41 somehow I got it to report the stdout errors, maybe I was tired 2011-09-18T21:32:42 ChrisH: ahh, that would explain it 2011-09-18T21:32:45 jbroman: i was just wondering how it compares 2011-09-18T21:32:59 janzer: Well, it works in on a non-secured machine, because House works in the original directory 2011-09-18T21:33:14 janzert: do I just make up an api key from the worker? 2011-09-18T21:33:20 Haven't touched zeta in quite a while, so it's state will be unchanged for some time, then may enter a state of rapid flux at some point. 2011-09-18T21:33:39 If you're not ever connecting it to the beta server 2011-09-18T21:33:49 but I am 2011-09-18T21:34:26 what stage is your worker vm at? 2011-09-18T21:34:28 pull the api_create_key from server_info.php? 2011-09-18T21:34:44 contest user create, cloned repo, writing the worker_setup cmd 2011-09-18T21:34:49 yes, that's the api_create_key though not the worker_api_key ;) 2011-09-18T21:34:59 McLeopold: yes, and then put that into a get request to one of the php files 2011-09-18T21:35:18 normally api_server_setup.php will create the worker api key 2011-09-18T21:35:23 yes! that ^ 2011-09-18T21:36:06 so from the worker run curl 'http://aichallengebeta.hypertriangle.com/api_server_setup.php?api_create_key= 2011-09-18T21:36:22 but that's not in worker_init.py? 2011-09-18T21:36:43 that feeds you back worker_init.py with the api key 2011-09-18T21:36:59 ok 2011-09-18T21:37:08 but I already created the contest user 2011-09-18T21:37:20 specifically the ec2 instances start with "curl 'http://aichallengebeta.hypertriangle.com/api_server_setup.php?api_create_key= | sh" 2011-09-18T21:37:41 from the ubuntu user? 2011-09-18T21:37:48 from root 2011-09-18T21:37:54 but just a sec 2011-09-18T21:38:09 since you already have the user and repository in place 2011-09-18T21:38:26 just manually copy down the api key given to you by the above call 2011-09-18T21:40:12 then from the setup directory run "./worker_setup.py -y --username contest --api-url http://aichallengebeta.hypertriangle.com --api-key --install-cronjob --start" 2011-09-18T21:40:38 oops that url needs a trailing slash I believe 2011-09-18T21:43:17 and perhaps a sudo 2011-09-18T21:44:02 sorry, thought I mentioned before it needed to be run as root 2011-09-18T21:44:53 you might want also leave --start off the end 2011-09-18T21:45:07 and may or may not want the --install-cronjob part as well 2011-09-18T22:01:57 btw, so long as ChrisH changes actually work ;) they look fine to me 2011-09-18T22:02:53 also ExternalCompiler.vglobs seems to be completely misnamed since I don't believe it can contain globs at all but must be actual file names 2011-09-18T22:05:00 *** eashoka has joined #aichallenge 2011-09-18T22:11:20 *** bhasker has quit IRC (Ping timeout: 252 seconds) 2011-09-18T22:14:16 *** nux67 has quit IRC (Ping timeout: 252 seconds) 2011-09-18T22:15:08 aichallenge: janzert epsilon * rbd13abc / manager/adjust_turnlimits.py : Add option to run adjust_turnlimits without storing the results in the database - http://git.io/tmg0Rg 2011-09-18T22:20:18 janzert: what's a glob, then? 2011-09-18T22:20:41 a glob is a wildcard pattern, like *.go. 2011-09-18T22:20:43 *.jp 2011-09-18T22:20:47 yes 2011-09-18T22:20:59 http://en.wikipedia.org/wiki/Glob_%28programming%29 2011-09-18T22:21:25 well, I was close :) 2011-09-18T22:21:53 my worker is still installing... 2011-09-18T22:22:14 we support a lot of language :} 2011-09-18T22:23:24 ec2 probably has faster network than me 2011-09-18T22:23:26 yes, it's impressive 2011-09-18T22:23:43 (# of langs that is) 2011-09-18T22:24:31 McLeopold: yeah, ec2 sets up way faster than I can locally 2011-09-18T22:26:14 hmm, beta server time is about 14 hours off? 2011-09-18T22:26:48 suspiciously almost exactly 14 hours off 2011-09-18T22:27:18 oh I guess not 2011-09-18T22:27:40 it's just set to moscow daylight time? 2011-09-18T22:27:51 *** McLeopold1 has joined #aichallenge 2011-09-18T22:28:01 which doesn't really make sense either though 2011-09-18T22:28:08 *** McLeopold has quit IRC (Ping timeout: 260 seconds) 2011-09-18T22:28:15 *** McLeopold1 is now known as McLeopold 2011-09-18T22:29:02 McLeopold: any idea how or why the beta server is set to moscow time? 2011-09-18T22:29:17 it's in moscow? 2011-09-18T22:29:44 where is demerzel from? 2011-09-18T22:30:10 looks like it's hosted in an he.net datacenter 2011-09-18T22:30:28 pretty sure they're chicago and maybe east coast 2011-09-18T22:31:19 I thought it used to be set to central time, but I may not be remembering correctly 2011-09-18T22:38:16 McLeopold: hmm, speaking of demerzel though any idea why he isn't getting any games since the ranking reset? other C# bots seem to be fine 2011-09-18T22:41:32 looks like he's the seed in a matchup that has a NULL max_turns; 2011-09-18T22:44:30 there are 5 matchups with null max_turns, guessing maybe they were stuck in the transition? 2011-09-18T22:48:23 going to just update all null max_turns to 300 2011-09-18T22:51:31 ok 2011-09-18T22:54:35 hmm, that doesn't seem to have unstuck them. What's the best way to just get rid of them? 2011-09-18T22:54:52 matchups? 2011-09-18T22:54:58 yes 2011-09-18T22:55:04 well, i'd rather keep them and solve the problem 2011-09-18T22:55:45 ok, 266777, 26670 and 266781 or just select * from matchup where worker_id = 0 2011-09-18T22:57:27 do you know what that utf-8 error is? 2011-09-18T22:57:33 looks like demerzel's the only seed of those three games that is still an active submisssion 2011-09-18T22:57:38 no 2011-09-18T22:58:22 i'll set the worker_id to null 2011-09-18T22:58:59 hard to tell on most of those errors without a traceback 2011-09-18T22:59:17 ahh, it needs a null worker_id to pick it up 2011-09-18T22:59:18 yeah, I need to send them back 2011-09-18T22:59:21 yes 2011-09-18T22:59:30 but it should not have set to zero... 2011-09-18T23:07:09 McLeopold: is the test environment still configuring? 2011-09-18T23:09:24 hmm, the worker shows that it got the matchup, ran it and reported the result but I don't see that on the server at all 2011-09-18T23:10:04 matchup has errors 2011-09-18T23:10:19 error is null 2011-09-18T23:10:31 I see some NoneType errors 2011-09-18T23:10:33 and no errors in the worker log 2011-09-18T23:10:45 for matchup 266777 I mean 2011-09-18T23:11:12 the other had the NoneType errors, I assume related to the seed no longer being an active submission 2011-09-18T23:11:39 maybe we should report than nicely? in get submission? 2011-09-18T23:11:57 http://pastebin.com/CTSUtH88 2011-09-18T23:12:02 relevant section of the worker log 2011-09-18T23:14:02 that's just a guess on the NoneType error since there isn't anything in the worker log about it at all 2011-09-18T23:15:54 the php_api.log shows handing out the task, but nothing after that 2011-09-18T23:16:00 no response written 2011-09-18T23:16:18 if the worker got the hash back, then api_game_result must have crashed 2011-09-18T23:16:52 yeah, shows it as getting the hash back 2011-09-18T23:18:26 php did crash 2011-09-18T23:18:37 I must have another double $$ 2011-09-18T23:19:38 grep doesn't see it ;) 2011-09-18T23:20:57 check ~/logs/error.log 2011-09-18T23:21:15 the php error is at the end, and the time corresponds with matchup 266777 2011-09-18T23:21:33 I should wrap phpshills in a try 2011-09-18T23:21:36 it must have crashed before before line 138 since the matchup didn't get deleted 2011-09-18T23:21:38 phpskills 2011-09-18T23:21:40 haha 2011-09-18T23:21:52 113 2011-09-18T23:23:57 the player_id's in matchup_player are all -1 2011-09-18T23:24:42 weird 2011-09-18T23:25:02 and explains the crash 2011-09-18T23:25:20 *** delt0r_ has quit IRC (Ping timeout: 260 seconds) 2011-09-18T23:25:26 but not how it got to that 2011-09-18T23:25:47 those id's are set to -1 until the last part of the generate matchup sp 2011-09-18T23:25:54 so, the stored procedure crashed? 2011-09-18T23:26:44 maybe it was out of sync with the database schema for a little bit and couldn't handle the new max_turn fields? 2011-09-18T23:27:05 although I don't know what in it would have been effected by that 2011-09-18T23:29:38 so, let's ignore that one 2011-09-18T23:29:48 although we should have failed earlier 2011-09-18T23:30:08 ok :) 2011-09-18T23:30:08 get_task can check for that and delete the matchup? 2011-09-18T23:30:21 probably makes sense 2011-09-18T23:30:24 I really need to get transactions working... 2011-09-18T23:30:31 but php sucks 2011-09-18T23:31:58 user 22, submission_id 1170 is causing the utf8 issues 2011-09-18T23:32:53 the submission is active and playing games... 2011-09-18T23:33:35 yep, the log on worker 203 has several games with it that appear normal 2011-09-18T23:34:56 across 3 different maps, 3 different workers, that submission is the only constant 2011-09-18T23:35:44 we need the worker to log a more detailed error when it sends an error back to the server 2011-09-18T23:37:21 *** delt0r_ has joined #aichallenge 2011-09-18T23:37:33 the traceback is at the debug level 2011-09-18T23:37:40 I can bump that up to error :) 2011-09-18T23:37:50 sure 2011-09-18T23:38:21 verbose logs on errors are good ;) 2011-09-18T23:38:49 you want the trace back to the database too? 2011-09-18T23:39:09 doesn't matter so much to me 2011-09-18T23:39:20 pulling it out will be more painful 2011-09-18T23:39:42 of the database? at least we will see something 2011-09-18T23:40:29 I've looked at worker logs quite a bit, often have an ongoing tail of each one in the background 2011-09-18T23:40:40 this is the first time I've ever looked at the database 2011-09-18T23:41:29 aichallenge: McLeopold epsilon * rfa0ee5c / worker/worker.py : increase log level of game errors - http://git.io/UpWRug 2011-09-18T23:41:29 aichallenge: McLeopold epsilon * r8cf4e8f / (7 files in 3 dirs): Merge branch 'epsilon' of github.com:aichallenge/aichallenge into epsilon - http://git.io/zpiOTQ 2011-09-18T23:41:44 you'll need to recycle the workers now 2011-09-18T23:41:51 if you trust my code... 2011-09-18T23:41:59 I have a bad track record 2011-09-18T23:42:03 *** dvladim has joined #aichallenge 2011-09-18T23:42:13 is that the only change that needs updating? 2011-09-18T23:42:22 yep 2011-09-18T23:42:59 crap, it is buggy :( 2011-09-18T23:43:20 :P 2011-09-18T23:43:53 aichallenge: McLeopold epsilon * r0aeafb9 / worker/worker.py : increase log level of game errors 2 - http://git.io/jHsVYg 2011-09-18T23:44:02 this is why I need that worker setup locally... 2011-09-18T23:45:32 ok, 203 update manually 2011-09-18T23:45:34 *** eashoka has quit IRC (Ping timeout: 252 seconds) 2011-09-18T23:49:35 whoa, my vm network is really slow, but I'm getting a task 2011-09-18T23:50:23 janzert: my worker reported, "Could not find an unloacked jail" 2011-09-18T23:50:24 both ec2 workers are updated btw 2011-09-18T23:50:35 :( 2011-09-18T23:50:57 but the traceback is in the database :) 2011-09-18T23:50:58 the setup seemed to complete without any errors? 2011-09-18T23:51:19 yea, it did 2011-09-18T23:51:29 I left off --setup and --install-cron 2011-09-18T23:52:05 I have 32 jail users 2011-09-18T23:52:11 do I run worker as root? 2011-09-18T23:52:18 nope 2011-09-18T23:52:31 do I add contest user to a group? 2011-09-18T23:52:39 you mean you left off --start right? 2011-09-18T23:52:46 yes 2011-09-18T23:53:32 contest is a member of: contest adm dialout cdrom plugdev lpadmin sambashare admin 2011-09-18T23:54:02 jailkeeper? 2011-09-18T23:54:18 under /srv/chroot should be a bunch of jailuser directories 2011-09-18T23:54:19 oh, log out, log in 2011-09-18T23:54:41 McLeopold: According to sandbox.py that error is produced if the results of os.listdir(jail_base) come back empty. 2011-09-18T23:55:04 jail_base is /srv/chroot 2011-09-18T23:55:16 yeah, contest didn't have jailkeeper rights because I needed to logout/login 2011-09-18T23:55:25 ChrisH: well or there is a "locked" directory within each jailuser directory 2011-09-18T23:55:42 janzer: yes 2011-09-18T23:55:57 much better 2011-09-18T23:56:01 janzert: I keep leaving the 't' off your handle for some reason. 2011-09-18T23:56:08 now I need to manufacture a go game 2011-09-18T23:56:13 heh 2011-09-18T23:56:25 I can send you something. 2011-09-18T23:56:27 ChrisH: usually you can tab complete, depending on your chat client 2011-09-18T23:56:40 cool, didn't know that. 2011-09-18T23:56:51 ChrisH: no, I need a matchup for my worker, don't resubmit 2011-09-18T23:57:08 yep, and ChrisH keeps throwing me cause there is also Chris_... :} 2011-09-18T23:57:10 McLeopold: I know. 2011-09-18T23:57:34 McLeopold: just saying, I have a zip file on my machine I could send YOU for testing. 2011-09-18T23:57:54 McLeopold: you could run the compiler manually 2011-09-18T23:57:58 well, I want to run as a proper worker to test the new jail code you wrote 2011-09-18T23:58:12 so long as it is configured to use the sandbox 2011-09-18T23:58:19 janzert: I have my methods, it won't be an issue :) 2011-09-18T23:58:44 does this go code need to pass the bot validation, or just compile? 2011-09-18T23:59:33 he's just generating a matchup with your current submission 2011-09-18T23:59:52 oh, ok then, that's what I would have sent anyway. :)