2011-03-11T00:00:28 https://github.com/sigh/aichallenge/blob/epsilon/ants/game.py 2011-03-11T00:00:32 look at that stub class 2011-03-11T00:00:42 That's the "interface" 2011-03-11T00:01:02 *** antimatroid has left #aichallenge 2011-03-11T00:01:05 The engine excepts all of those function to exist, regardless of the game being played 2011-03-11T00:01:17 ah, Ants is supposed to inherit from this? 2011-03-11T00:01:30 Well, it could, but in python there is no point. 2011-03-11T00:01:45 I would do it for documentation purposes, if nothing else 2011-03-11T00:01:45 I could put a raise error in the base class. 2011-03-11T00:01:56 so you have to override them all 2011-03-11T00:02:01 why is this in the ants folder then? should be outside somewhere.. 2011-03-11T00:02:12 amstan_: geeze, give me time 2011-03-11T00:02:17 haha 2011-03-11T00:02:34 when I see something like that Game, I generally search the repo for places where it is used 2011-03-11T00:03:00 so if I didn't find it, I would be confused... or afraid of some scary meta-programming 2011-03-11T00:03:11 :P 2011-03-11T00:03:20 So I'm liking the work you did now, so let's go with it. 2011-03-11T00:03:56 cool :) 2011-03-11T00:04:00 But, I think it would be appropriate to nit-pick the do_attack and do_spawn functions, because everyone will be referring to them 2011-03-11T00:04:38 *** antimatroid has joined #aichallenge 2011-03-11T00:04:43 oh... before that... one nit-pick about code 2011-03-11T00:04:58 If we could get the data structure to it's own class, so the attack, spawn, food and score are the only thing in the Ants class, that would be great for helping people understand it 2011-03-11T00:05:01 why do you prefer to keep row, col separate rather than in a tuple? 2011-03-11T00:05:19 I don't know. 2011-03-11T00:05:34 Fear that unpacking would be a performance hit? 2011-03-11T00:05:45 hehe 2011-03-11T00:06:11 it's better to pack it into a tuple, because then most places in the code don't need to know about the internals 2011-03-11T00:06:14 I would make sense to just use a tuple, so people that use languages with vectors and point classes can read it better 2011-03-11T00:06:37 for example, switching form (col,row) to (row,col) would have been easier :P 2011-03-11T00:06:58 I think I went back and forth, and my distaste for [0] at the end of variables all the time won. :) 2011-03-11T00:07:29 oh, that's easy to fix... just do row,col = location if you need it explicitly 2011-03-11T00:07:41 better yet reduce the number of places where it is needed explicitly 2011-03-11T00:07:50 yeah, and I do that anyway in the code all over the place, so I really have no excuse 2011-03-11T00:07:56 :) 2011-03-11T00:08:29 and it makes thing that much easier if we decide to go 3d or something :D 2011-03-11T00:08:42 So, the new food spawning makes my test runs a lot slower. 2011-03-11T00:09:01 I can't tell if it is just extra ants, or the engine. And I haven't really looked at it yet. 2011-03-11T00:09:03 yeah... it's hard to get much faster than random 2011-03-11T00:09:20 I shouldn't be 2011-03-11T00:10:13 in general we are keeping more data now, but I don't think I did anything terribly inefficient 2011-03-11T00:10:36 in the food code, what is bfs? 2011-03-11T00:10:49 breadth-first search 2011-03-11T00:10:59 really, the offset-code should not be using it 2011-03-11T00:11:17 but it's required for the section-code 2011-03-11T00:11:37 so in sections, you just create lists for each player that are cached, then food spawning should be just as fast as random 2011-03-11T00:11:49 hmm...yeah 2011-03-11T00:12:16 We need a mechanism to control the food spawned per turn 2011-03-11T00:12:16 meh, need to profile the thing 2011-03-11T00:12:29 control in what way? 2011-03-11T00:12:56 so, if it is 3 per turn, and there are 5 bots, then spawn 5 every so often 2011-03-11T00:13:03 oh... I just realised 2011-03-11T00:13:17 McLeopold: i get this when I try to get the visualiser thing going 2011-03-11T00:13:18 http://pastebin.com/7NgZnmww 2011-03-11T00:13:19 which means, we don't need to spawn every turn 2011-03-11T00:13:25 my food spawning will spawn a lot more food by default than before I messed with it 2011-03-11T00:13:31 like num_players times more 2011-03-11T00:13:34 the browser shows the map with start positions, but i get nothing after that 2011-03-11T00:13:40 which means a lot more ants => more run times 2011-03-11T00:14:20 antimatroid: the replay file is being loaded by javascript, which is why you need the python http server in the first place 2011-03-11T00:14:35 look at the script in firebug to see what the file is that it is requesting 2011-03-11T00:14:45 antimatroid: press arrow keys in browser 2011-03-11T00:14:48 you should be able to hit it in your browser 2011-03-11T00:14:52 oh, that :P 2011-03-11T00:14:58 that'll do it :) 2011-03-11T00:15:08 cheers :) 2011-03-11T00:15:39 I should try to see if Frontier's works 2011-03-11T00:16:04 has Frontier got something that conforms to the new format? 2011-03-11T00:16:10 maybe 2011-03-11T00:16:29 I purposely kept the current output as is, and just wrote to *.replay.new 2011-03-11T00:16:42 That's fine for now 2011-03-11T00:16:50 I'll add parser options later 2011-03-11T00:17:04 yeah 2011-03-11T00:17:24 also, I made some changes to my version of ants, so I'm going to need to merge 2011-03-11T00:17:33 and I'm scared. 2011-03-11T00:17:36 haha 2011-03-11T00:18:30 actually, I pushed it 2011-03-11T00:19:53 btw, is it easy to make the bots take a --seed argument or something? I added one to playgame in a hope to make reproducible tests, but failed because the bot are still random 2011-03-11T00:19:59 ? 2011-03-11T00:20:40 well, that's up to the bot owner 2011-03-11T00:20:48 but our example ones can 2011-03-11T00:20:58 I was going to change Hunter to not be random 2011-03-11T00:20:59 that should have been set before we started making the tournament engine and the starter packages 2011-03-11T00:21:24 oh... can we pass seed as a parameter? 2011-03-11T00:21:34 at the start of the game? 2011-03-11T00:21:35 yes 2011-03-11T00:21:41 it can be ignored by the bots if they like 2011-03-11T00:21:52 but it is great for testing for those who want it 2011-03-11T00:22:00 let's add it 2011-03-11T00:22:05 ok, I'll do it 2011-03-11T00:22:24 playerSeed n seed, for the replay 2011-03-11T00:22:39 meh, give them all the same seed 2011-03-11T00:22:43 or, 1 seed to rule them all? 2011-03-11T00:22:53 yeah, no reason not to 2011-03-11T00:24:14 okay, I'm heads down in worker.py right now. 2011-03-11T00:31:43 janzert: you here? 2011-03-11T00:32:29 I'm curious about the compiling sandbox/jail 2011-03-11T00:42:54 can there be a specified seed that means no food is to be spawned at all? 2011-03-11T00:43:27 that way people can test on a map they set up without worrying about food spawning 2011-03-11T00:50:13 *** Palmik has joined #aichallenge 2011-03-11T00:51:26 *** chemuduguntar has quit IRC (Ping timeout: 276 seconds) 2011-03-11T00:54:13 *** delt0r_ has joined #aichallenge 2011-03-11T00:55:12 *** delt0r___ has quit IRC (Ping timeout: 255 seconds) 2011-03-11T00:55:25 antimatroid: that doesn't make sense, that functionality seems orthogonal to what the seed does 2011-03-11T00:55:29 antimatroid: why? 2011-03-11T00:55:55 like, a map with food already there? 2011-03-11T00:56:16 I'd rather have a food_spawn option of none 2011-03-11T00:58:39 anyway, I've implemented the seed in the engine and the python bots if someone wants to pull: https://github.com/sigh/aichallenge 2011-03-11T00:58:49 you can now run deterministic games :) 2011-03-11T00:59:07 I'll update the spec, I suppose 2011-03-11T01:01:31 sigh: I'll pull sometime tomorrow 2011-03-11T01:03:13 McLeopold: a food_spawn option of none would do 2011-03-11T01:03:33 the map may/may not have food, but it'd be useful for testing to see if your bot is behaving as you want 2011-03-11T01:03:40 without food spawning mucking it up 2011-03-11T01:11:20 hmmm... I have a bug 2011-03-11T01:19:24 *** Frontier has joined #aichallenge 2011-03-11T01:24:11 sigh: good job, i'll see if i can catch up 2011-03-11T01:26:01 haha 2011-03-11T01:26:30 Frontier: I know you are responding to my @later tell, but the last thing I said in the channel was "hmmm... I have a bug" 2011-03-11T01:27:30 lol 2011-03-11T01:27:49 let me read the logs :p 2011-03-11T01:32:54 *** boegel has quit IRC (Quit: Leaving) 2011-03-11T01:33:13 sigh: What kind of bug is it? 2011-03-11T01:33:40 oh I fixed it, it was in my conflict resolution 2011-03-11T01:34:05 I suppose mostly because the bots I was testing with never issued conflicts :P 2011-03-11T01:35:35 my bot on there will give you conflicts :P 2011-03-11T01:37:03 i whish we had let the starter bot be the random bot, not the top-right bot always gets crushed by the others :p 2011-03-11T01:37:21 *now 2011-03-11T01:40:44 RαηdσmBστ ιs αςτμαllγ α ρrεττγ dεςεητ bστ fσr τhιs gαmε. Iτ σςςμριεs ιτs τεrrιτσrγ αηd dσεsη'τ sτrαγ τσσ mμςh. 2011-03-11T01:42:16 Bτω. JαναSςrιρτ sμρρσrτs Uηιςσdε ιdεητιfιεrs, bμτ σηlγ Chrσmε sεεms τσ ςαrε. Fιrεfσχ ιs fιηε ωιτh Grεεκ lεττεrs, αηd θρεrα sριτs σμτ αηγτhιηg τhατ ιs ηστ ASCII. 2011-03-11T01:43:51 meh, let people do random bot themselves 2011-03-11T02:23:49 *** mega1 has joined #aichallenge 2011-03-11T02:58:14 *** smellyhippy has quit IRC (Quit: Take your stinking clause off me you damn dirty predicate!) 2011-03-11T03:04:19 *** Kingpin13 has joined #aichallenge 2011-03-11T03:09:21 *** McLeopold has left #aichallenge 2011-03-11T03:14:58 *** boegel has joined #aichallenge 2011-03-11T03:39:57 *** aerique has joined #aichallenge 2011-03-11T03:46:30 *** smellyhippy has joined #aichallenge 2011-03-11T03:53:20 *** mceier has joined #aichallenge 2011-03-11T04:17:57 *** Accoun has quit IRC () 2011-03-11T04:20:30 *** Accoun has joined #aichallenge 2011-03-11T04:35:03 *** Naktibalda has joined #aichallenge 2011-03-11T04:38:33 *** amstan_ has quit IRC (Remote host closed the connection) 2011-03-11T05:03:49 *** delt0r___ has joined #aichallenge 2011-03-11T05:04:47 *** delt0r_ has quit IRC (Ping timeout: 264 seconds) 2011-03-11T05:18:24 *tumbleweeds* 2011-03-11T07:13:39 *** Frontier has quit IRC (Quit: Leaving.) 2011-03-11T07:44:06 *** antimatroid has quit IRC (Read error: Connection reset by peer) 2011-03-11T07:45:21 *** antimatroid has joined #aichallenge 2011-03-11T08:12:05 *** iFire` has joined #aichallenge 2011-03-11T08:19:56 *** iFire has quit IRC (Ping timeout: 246 seconds) 2011-03-11T08:28:04 *** sigh has quit IRC (Remote host closed the connection) 2011-03-11T08:47:02 *** UncleVasya has joined #aichallenge 2011-03-11T08:50:38 *** |UncleVasya| <|UncleVasya|!kvirc@88-214-119-72.dialup.umc.net.ua> has joined #aichallenge 2011-03-11T08:52:43 *** UncleVasya has quit IRC (Ping timeout: 260 seconds) 2011-03-11T08:54:36 *** delt0r_ has joined #aichallenge 2011-03-11T08:56:03 *** delt0r___ has quit IRC (Ping timeout: 252 seconds) 2011-03-11T09:07:27 are we still having a maximum of 26 players? 2011-03-11T09:07:33 otherwise we need to get rid of the char maps 2011-03-11T09:25:25 i've written my bot to account for 99 players currently :) 2011-03-11T09:26:06 yeah, i'm trying to work out how to change mine 2011-03-11T09:26:11 it gets a bit messier 2011-03-11T09:31:34 internally i'm using an array of integers for representing the map, chars didn't seem ideal 2011-03-11T09:32:27 why? 2011-03-11T09:33:16 Naktibalda: for one, the problem antimatroid just mentioned 2011-03-11T09:33:59 ah, a player code 2011-03-11T09:52:44 aerique: what is your code for dead ants? 2011-03-11T09:53:00 and how are you planning to handle colliding ants? 2011-03-11T09:53:20 i can't think of a way that isn't messy as hell 2011-03-11T09:53:21 when is the official start of a challenge? 2011-03-11T09:53:32 when everything is ready :P 2011-03-11T09:53:44 I just followed the wiki and am using lowercase for a live ant and uppercase for a dead one when printing the map to terminal. 2011-03-11T09:53:54 you have unfair advantage :) 2011-03-11T09:54:18 I haven't even thought about colliding ants. Just finished the most basic sample bot :) 2011-03-11T09:54:54 I'm hoping to work on some other (non-challenge) projects before the challenge starts. 2011-03-11T09:57:08 <|UncleVasya|> Naktibalda: I asked about this yesterday and answers were from 3 to 4 weeks. 2011-03-11T09:57:27 thanks 2011-03-11T09:57:47 aerique: i'm thinking about how i will represent stuff myself 2011-03-11T10:12:38 *** |UncleVasya| <|UncleVasya|!kvirc@88-214-119-72.dialup.umc.net.ua> has quit IRC (Ping timeout: 260 seconds) 2011-03-11T10:16:18 *** _flag <_flag!~flag@69-165-173-172.dsl.teksavvy.com> has joined #aichallenge 2011-03-11T10:16:57 *** UncleVasya has joined #aichallenge 2011-03-11T10:47:56 *** aerique has quit IRC (Quit: ...) 2011-03-11T10:55:44 *** amstan has joined #aichallenge 2011-03-11T10:55:44 *** ChanServ sets mode: +o amstan 2011-03-11T10:56:20 *** Accoun has quit IRC () 2011-03-11T11:01:55 *** Accoun has joined #aichallenge 2011-03-11T11:17:11 *** McLeopold has joined #aichallenge 2011-03-11T11:27:44 *** dr`away has joined #aichallenge 2011-03-11T11:30:02 McLeopold: if we're to be passed info for ants that have collided, can the line identifier be different to those who died from battle? 2011-03-11T11:33:07 hi 2011-03-11T11:33:24 hi :) 2011-03-11T11:33:26 antimatroid: it doesn't matter 2011-03-11T11:33:28 *** boegel has quit IRC (Ping timeout: 248 seconds) 2011-03-11T11:33:36 it does for the programmer 2011-03-11T11:33:37 the dead ants will be at the same location, if you need to know 2011-03-11T11:34:02 yes, the point is having to deal with knowing if more than one dead ant line was given for the same square 2011-03-11T11:34:05 and they will have the same owner 2011-03-11T11:34:09 and tehn accounting for a potential 3rd etc. 2011-03-11T11:34:31 there will be multiple dead ant lines for the same square 2011-03-11T11:34:36 and assuming one can play around with the attack radius, it could not be from the same owner 2011-03-11T11:34:54 so were good then with what we have 2011-03-11T11:34:54 yes, and it's extra code to work out whether you got those extra lines 2011-03-11T11:35:10 it'd be easy to just ignore that information if it was given as a conflict death 2011-03-11T11:35:17 is it possible that attack range > visibility range? 2011-03-11T11:35:19 as conflict deaths are useless information imo 2011-03-11T11:35:30 Naktibalda: no, we won't do that 2011-03-11T11:35:49 McLeopold: my issue isn't that it isn't possible to work the information out, i just don't like having so much implicit information 2011-03-11T11:36:16 i'd rather everything was told explicitly, i can compromise on food information being semi implicit in order to reduce input size, but this doesn't change input size? 2011-03-11T11:36:27 It is explicit info, dead ants that died on the same square died from conflict 2011-03-11T11:36:48 yes, but it's implicit in that to differentiate it from an ant that died from a battle, you have to work it out yourself 2011-03-11T11:36:50 Food is given every turn, btw 2011-03-11T11:36:56 you don't know that information when you receive the line 2011-03-11T11:37:22 food is given every turn, but if you go out of range of a food square and it's land when you come back in view, you only know implicitly that it's not there 2011-03-11T11:37:28 hence there is implicit information there as well 2011-03-11T11:38:21 couldn't one just receive 'c' instead of 'd' for a collided ant death line? 2011-03-11T11:39:52 antimatroid: I'd rather not change it. Maybe wait for jeff to come around again, then spend your effort convincing him of battle option 1? 2011-03-11T11:40:46 i will, i was just thinking about how to represent stuff, and it's already quite messy 2011-03-11T11:40:55 dealing with all the implicit information makes it even messier 2011-03-11T11:41:21 i want to move away from using char maps even in the starter packs 2011-03-11T11:41:33 because that limits anything to not breaking the starter bots to 26 players 2011-03-11T11:41:41 Yeah, for input parsing, just make a little temp hash table of death locations, then flip your "attack/conflict" bit when you get a collision. It's not too messy. 2011-03-11T11:42:23 Hey, it's an ant collision *and* a hash table collision! 2011-03-11T11:42:24 hmm, i might have to create a separate structure just for passing input : 2011-03-11T11:42:54 I don't think we will get above 26 bots. 2011-03-11T11:43:07 not even on crazy tcp? :( 2011-03-11T11:43:29 Not if I want to surf the net in the next 2 months. :) 2011-03-11T11:43:50 nah, have an official crazy maps tcp on the official servers 2011-03-11T11:44:26 it'd be an effort to get that many people in a game on the other tcp servers 2011-03-11T11:44:53 so, lower and upper case, 52 players? 2011-03-11T11:45:02 upper case is dead ant 2011-03-11T11:45:50 if you move away from char maps, people can pretty much use any number of players 2011-03-11T11:46:02 i guess we're restriced by map files 2011-03-11T11:46:11 so it's safe to assume we'll stay <=26 players 2011-03-11T11:46:26 where is it described? 2011-03-11T11:46:33 the game? 2011-03-11T11:46:38 I thought that owner is integer, not character 2011-03-11T11:46:56 it is for input 2011-03-11T11:47:02 but the map files are char maps 2011-03-11T11:47:14 https://github.com/aichallenge/aichallenge/wiki/Ants-Game-Specification 2011-03-11T11:47:19 the specs page is there 2011-03-11T11:48:00 McLeopold: did you figure out whatever you were wondering last night? 2011-03-11T11:48:44 McLeopold: how hard would it be to change it so that dead ant info was explicitly different when reading? 2011-03-11T11:48:57 surely you agree that explicit information is better than implicit where possible? 2011-03-11T11:49:32 no 2011-03-11T11:51:43 janzert: how easy is the zeta sandboxing to use? 2011-03-11T11:51:57 no idea, haven't even looked at it 2011-03-11T11:56:25 *** superflit_ has joined #aichallenge 2011-03-11T12:02:09 *** delt0r_ has quit IRC (Read error: Connection reset by peer) 2011-03-11T12:02:31 *** delt0r_ has joined #aichallenge 2011-03-11T12:19:06 *** amstan has quit IRC (Ping timeout: 250 seconds) 2011-03-11T12:23:19 *** UncleVasya has quit IRC (Ping timeout: 260 seconds) 2011-03-11T12:29:29 *** UncleVasya has joined #aichallenge 2011-03-11T12:36:45 *** UncleVasya has quit IRC (Ping timeout: 260 seconds) 2011-03-11T12:37:00 *** UncleVasya has joined #aichallenge 2011-03-11T12:51:52 *** boegel has joined #aichallenge 2011-03-11T13:16:09 janzert: you wanted to redo the submission statuses, right? 2011-03-11T13:17:32 Anywhere a check for an error status is done would be trivial if all error statuses were strictly larger than non-error statuses :} 2011-03-11T13:17:46 with the current codes that is *almost* already the case 2011-03-11T13:18:42 and before ready to run? 2011-03-11T13:19:07 looks at status list, basically if 30 and 40 were transposed it would be fine I think 2011-03-11T13:20:13 I don't think we need 10, 15, 27, 30 or 60 2011-03-11T13:20:44 probably don't need all of them anyway 2011-03-11T13:20:44 there is no need to create a database entry if we don't get a successful upload 2011-03-11T13:21:26 you do need a unique submission id before you can attempt to move the upload into its final place though right? 2011-03-11T13:21:47 yes, but at that point, we've got a file, so it will be status 20 2011-03-11T13:22:23 ok 2011-03-11T13:22:30 well, no 2011-03-11T13:22:59 20 is after it has been moved to the final destination on the main server and is ready for a worker to grab and compile 2011-03-11T13:23:27 yeah, but I'm removing 10 and 15 2011-03-11T13:23:29 or I suppose the very initial status could be null, but that seems rather ugly 2011-03-11T13:24:08 I'd think you want 10 initially, then 20 once you moved it into the correct location 2011-03-11T13:24:37 yeah, but I don't see stuff getting stuck in 10 unless we have serious issues with the main server. 2011-03-11T13:25:31 so, we had 2 tournament managers on the workers to keep the cpu's busy? 2011-03-11T13:25:57 it shouldn't, but you have a race if the initial database entry says the submission is in place before it really is 2011-03-11T13:26:00 yes 2011-03-11T13:26:24 *** Naktibalda has quit IRC (Quit: ChatZilla 0.9.86 [Firefox 3.6.15/20110303024726]) 2011-03-11T13:26:24 okay, I'm trying to get the worker to just use 1 script right now. It's looking nice so far. 2011-03-11T13:27:02 Then we could use threads and calc a load value 2011-03-11T13:27:13 because games will have different number of bots 2011-03-11T13:27:16 hmm, ok 2011-03-11T13:27:26 except the bots are run in sequence right? 2011-03-11T13:27:43 so you should always just assign one core per game 2011-03-11T13:27:44 not yet, :) but yes 2011-03-11T13:28:22 I forget how the jail users were used. One per worker script? 2011-03-11T13:28:35 one per bot running 2011-03-11T13:29:24 *** Chris_0076 has joined #aichallenge 2011-03-11T13:29:37 i.e. each bot runs under its own temporarily dedicated jail user 2011-03-11T13:30:07 oh yeah 2011-03-11T13:30:17 plus compiles, if we can sandbox them 2011-03-11T13:30:23 right 2011-03-11T13:30:38 so, each worker can create/verify it's own jail users, then there is no database? 2011-03-11T13:31:16 well there is a permanent pool of jail users 2011-03-11T13:31:29 that's why I said verify 2011-03-11T13:31:41 but each can have it's own pool 2011-03-11T13:31:46 there just needs to be a way of assigning them without going through a database 2011-03-11T13:31:50 yes, you could 2011-03-11T13:32:07 so we start a worker like this "worker.py -t -n0 -id 1" 2011-03-11T13:32:13 but it means each worker needs a pool as larger as the largest number of players in a game 2011-03-11T13:32:20 and it looks for jail_users assigned to id 1 2011-03-11T13:32:38 so, is a few hundred jail users a big deal? 2011-03-11T13:32:44 s/as larger as/as large as/ 2011-03-11T13:33:44 it seems uglier than running a database to me but 2011-03-11T13:33:46 * janzert shrugs 2011-03-11T13:34:08 if all the workers are going at full capacity, then you need that many anyway 2011-03-11T13:34:27 actually we'll run into memory issues before that 2011-03-11T13:34:42 I was thinking about that a bit last night 2011-03-11T13:35:05 hmm... how much can each bot allocate? do we quota them? 2011-03-11T13:35:35 Ec2 high cpu like we used last time except the larger instance, with 8 cores has 7GB 2011-03-11T13:35:55 we had a problem last time trying to limit java to use 1GB(?) 2011-03-11T13:36:10 with an average of 4 users, that's 32 bot processes at a time 2011-03-11T13:36:24 wait, sequential, 2011-03-11T13:36:32 but we still need the memory 2011-03-11T13:37:06 right, even if we get sequential execution we'll quickly run into memory problems :( 2011-03-11T13:38:05 well, for now, I'll make each worker instance manage it's own jail users 2011-03-11T13:38:20 we could go to a high memory instance and get 34GB with 4 cores 2011-03-11T13:38:31 that might be better 2011-03-11T13:38:40 faster cores than the high cpu instance too 2011-03-11T13:38:59 but the price per core is double 2011-03-11T13:39:16 we can get one of each and see what does better 2011-03-11T13:44:19 if we get faster cores we might be able to get away with cutting down the turn time limit a bit 2011-03-11T13:48:55 especially if the slower core machine started swapping bots out to disk :P 2011-03-11T13:50:39 *** amstan has joined #aichallenge 2011-03-11T13:50:39 *** ChanServ sets mode: +o amstan 2011-03-11T13:50:43 heh yes 2011-03-11T13:57:29 *** boegel has quit IRC (Ping timeout: 250 seconds) 2011-03-11T14:19:17 *** boegel has joined #aichallenge 2011-03-11T14:26:54 aichallenge: McLeopold epsilon-new-schema * r56f606b / (4 files): Removed make file and replaced with example scripts using playgame.py. - http://bit.ly/fSiVNE 2011-03-11T14:26:57 aichallenge: McLeopold epsilon-new-schema * r0796834 / (22 files in 4 dirs): Worker can now request compiles and report status. - http://bit.ly/eTLfno 2011-03-11T14:26:59 aichallenge: McLeopold epsilon-new-schema * r64d311d / worker/engine.py : Removed ants import from engine. - http://bit.ly/ft0hWi 2011-03-11T14:30:06 janzert: for the functional test, 10 turns on a small map with a starter bot? 2011-03-11T14:30:52 sure, that should be sufficient. 2011-03-11T14:31:04 well, multiple opponents at least 2011-03-11T14:31:14 why? 2011-03-11T14:31:58 to be sure it can handle it, although I guess with fog of war its less likely to actually get to a spot it matters 2011-03-11T14:32:01 i'd say it doesn't even need *any* opponents, really 2011-03-11T14:32:15 well, okay, i guess it'd be good to have one 2011-03-11T14:32:16 I was actually planning on not running it in a real game 2011-03-11T14:32:33 just sending it the initial state and make sure it responded with a valid move 2011-03-11T14:32:52 I think that would catch a good majority of the crashers 2011-03-11T14:32:54 right, but it needs to respond 10 times :) 2011-03-11T14:33:06 but playing out 10 turns is even better of course 2011-03-11T14:33:16 that was my idea, too, janzert 2011-03-11T14:33:32 And if I set the map up correctly, I can get the starter bot to attack it. 2011-03-11T14:33:33 but with multiple test cases, rather than one test case with multiple turns 2011-03-11T14:34:28 Zannick, it would be unfair to give them vastly different states each turn. 2011-03-11T14:34:46 it would be new games each time 2011-03-11T14:34:55 which is not the most efficient, of course 2011-03-11T14:35:10 I think a small map with 1 opponent with a senario including spawn and attack is just fine. 2011-03-11T14:35:16 also if it does send an invalid move or crash in the tests the output should be sent in the failure email 2011-03-11T14:35:54 I'm attempting to capture stderr, and logging bad moves. 2011-03-11T14:36:18 * janzert nods 2011-03-11T14:36:43 janzert: if you have time, look at worker.py, which I just pushed up 2011-03-11T14:36:54 The compile stuff is really messy, as I tried not to touch it. 2011-03-11T14:37:03 and nothing is using jail users yet 2011-03-11T14:37:10 the workers should do the emailing, in that case, rather than pass stderr and stdout logs around 2011-03-11T14:37:27 Zannick: yes 2011-03-11T14:37:45 For real games, it will pass stderr back to the database. 2011-03-11T14:37:58 does one worker.py run per core then? 2011-03-11T14:37:59 you have to be careful in that case 2011-03-11T14:38:07 or per server 2011-03-11T14:38:09 repr is probably sufficient, though 2011-03-11T14:38:30 janzert: when I get the jail users going, we should be able to spin up multiples per server 2011-03-11T14:39:09 Zannick: 1024 byes only and only on a timeout/crash. 2011-03-11T14:39:24 btw for security reasons I think Jeff was opposed to giving bot output directly back to the developers before 2011-03-11T14:39:28 McLeopold: i meant in terms of injection attacks 2011-03-11T14:39:36 personally I'm not quite as worried about it 2011-03-11T14:39:39 like i said, repr ought to take care of that 2011-03-11T14:39:49 Zannick, yes, we'll stripe everything but alpha and numbers :) 2011-03-11T14:40:14 ...that works too, i guess :) 2011-03-11T14:40:18 janzert: like if they tried to get system info? 2011-03-11T14:40:23 right 2011-03-11T14:40:29 I hadn't thought of that 2011-03-11T14:40:46 huh 2011-03-11T14:41:01 we could at least capture it, then give it out if asked for? 2011-03-11T14:41:17 if they get sensitive information with their bot currently it's a bit harder for them to make use of it 2011-03-11T14:41:38 have to encoded it in the way their bot plays :) 2011-03-11T14:42:11 so if you see ants spelling out ascii characters, quickly suspend that submission 2011-03-11T14:42:15 yeah, if someone were able to grab the jail user's private key and communicate it to them... 2011-03-11T14:42:33 jail user's don't have a key anymore 2011-03-11T14:42:49 oh 2011-03-11T14:42:53 but yeah that's the general idea 2011-03-11T14:43:32 http://en.wikipedia.org/wiki/Aztec_Code 2011-03-11T14:43:56 what is the get_languages api supposed to be used for? 2011-03-11T14:45:22 to get the language info from the database 2011-03-11T14:45:27 compile_anything needs the info 2011-03-11T14:45:29 ahh, I see it's a request of the languages table basically 2011-03-11T14:45:43 http://en.wikipedia.org/wiki/Data_matrix_(computer) 2011-03-11T14:45:50 okay, that barcode is better 2011-03-11T14:46:25 I'm sure I could come up with a bar code over time system for getting info. 2011-03-11T14:46:50 hmm, what does compile_anything need of it? 2011-03-11T14:46:56 I was thinking the worker might need it for the command line to actually run currently 2011-03-11T14:47:11 and I was hoping to move that out of the database actually 2011-03-11T14:47:23 the language data should be moved out of the database 2011-03-11T14:47:37 this is the plan (and i believe is implemented) for zeta 2011-03-11T14:48:24 janzert: just the list of langauge main files, right now 2011-03-11T14:49:02 ok, should be good for now I suppose. hopefully we have time to get rid of it later 2011-03-11T14:49:12 languages are in the database so we can tag a submission, and do the per language rankings 2011-03-11T14:49:38 so, it needs a list so it can report back to the server the correct id 2011-03-11T14:49:56 which means you can't take it out, even in zeta 2011-03-11T14:49:57 just report the language name back 2011-03-11T14:50:04 let the server map that to an id 2011-03-11T14:50:19 why bother? just put the data in the database 2011-03-11T14:50:42 what's the goal of removing it? 2011-03-11T14:50:44 because it doesn't belong in the database :) 2011-03-11T14:50:51 it's not something that should change 2011-03-11T14:50:57 but it is 2011-03-11T14:51:02 we add languages 2011-03-11T14:51:04 it's constant data 2011-03-11T14:51:14 yes but each language is fixed 2011-03-11T14:51:32