gmane.comp.games.tp.devel

Re: AI battle

Victor Ivri — 2008-08-21T17:02:14

I'm living in ET, which is -5 UTC. Tuesday 1PM UTC is fine for me. That's fine! I want to try and ping you prior to that, to see approx. how long a turn should take. Now I'm afraid, you see ;). Mine takes no longer than 3-4 seconds on a local server. Are you using TP04? I switched to it, but then I had problems with getting OrderDescs (server broke off connection with client for some reason, and might've crashed, too.. don't recall exactly), and I had to switch back to TP03. So mine doesn't send an EOT frame. like in Monthy Python's Killer Joke (http://tinyurl.com/4fhx7p). Cheers,

Re: Starmapper and wormholes

Krzysztof Sobolewski — 2008-08-21T16:04:22

Re: Starmapper and wormholes

Vincent Verhoeven — 2008-08-21T13:31:13

2008/8/19 Krzysztof Sobolewski interia.pl>: I think I'm making some progress. You are using the sizes of the galaxies to determine the map bounds (TParsecDataModel:312 and onwards). Risk has a size of 0 and a position of (0,0,0) on all galaxies it seems. I suppose it only uses them as a container for other stuff, and not something meaningful to be displayed. This means that the mapbounds are (0,0) to (0,0), thus width/height is 0 and division by zero later on. I'm going to let others fight on whether Risk or Starmapper needs to change. What do you even do when there are no galaxies, NullPointerExceptions? Or is that situation just not allowed by the protocol? Iwanowitch

Re: Mini-milestone on texture to globe projection

Jeffrey Miller — 2008-08-21T10:03:17

Re: Mini-milestone on texture to globe projection

Jeffrey Miller — 2008-08-21T09:53:15

Re: AI battle

Krzysztof Sobolewski — 2008-08-21T05:39:31

Re: AI battle

Victor Ivri — 2008-08-21T05:37:28

One suggestion, if I may. I noticed that it's very hard to conquer planets, and the game can be somewhat static. There's only "dramatic action" towards the end, where the party with a slight territorial advantage gains considerably more troops. For example, when my bots fight against each other, the game usually lasts over 100 turns, even though they try to be aggressive. In my mind, it could be a bit more dynamic.

Re: AI battle

Ryan Neufeld — 2008-08-21T03:35:33

I would love to watch as well. Keep us informed. I have to say also, I'll probably create a secondary Risk branch to make a bunch of improvements to the ruleset on, since many of them will probably break your AIs, and I'll let you know when I've done (and what it IS i have done.) Ryan 'jphr' Neufeld --------------------------- visit me at hammerofcode.com

Re: AI battle

Vincent Verhoeven — 2008-08-21T01:04:40

Can't we wait until after evaluation? ;) Nah, fine to me. I'm in UTC +2, so preferably not between 1AM and 6AM UTC. :) I propose either Tuesday somewhere beteen 8AM and 3PM UTC (preferable not after that) or sometime Wednesday, I have no plans yet for the whole day I think. Pick something that goes well for you then. I can host, though I'm fine if you'd rather have someone else. Starmapper doesn't work yet with Risk, I've spend a few hours today trying to fix it but I don't have the debugging skills for multithreaded java-fu. Well, I could do it in Eclipse perhaps, but it refuses to boot with a JVM segfault in libpthreads. Hooray. Mmm, longer turn length please. 120 should be plenty I think, but 60 is a bit close. My bot is pretty slow. :) Though it also sends a request end of turn frame at the end of its calculations, so it won't keep you if it actually isn't thinking. And laugh. :) Iwanowitch

AI battle

Victor Ivri — 2008-08-21T00:31:09

Hey guys, Well, my bot's been polished and thoroughly debugged, and is ready for conquest ;). Iwanowitch, what do you say we give a show to remember? I'm pretty flexible on date and time, as long as it's not the weekend, and as long as I get ample notice. Perhaps early next week? I'm thinking perhaps someone besides me and Iwanowitch should host the game. Anyone? Preferably, this should be someone who knows how to set up "Starmapper" for recording the game. As to the type of game, it's "Risk", with the "random" start. I suggest using quickstart-risk-random.conf with the following adjustments: turn_length_under_threshold = 60 (too long will be boring; too short and perhaps not all orders will go through) max_players = 2 Iwanowitch, you ok with that? As far as I can tell, any number of people can log in as "guest" (pwd: "guest") and watch. Let me know.

Re: Risk colonization segfault

Ryan Neufeld — 2008-08-20T16:21:31

I will most likely redo a lot of that code some time next week. As for this week I am out of town and can only do so much until next monday. Ryan 'jphr' Neufeld --------------------------- visit me at hammerofcode.com On 19-Aug-08, at 9:08 AM, Vincent Verhoeven wrote:

Re: Starmapper and wormholes

Krzysztof Sobolewski — 2008-08-19T16:51:09

Re: Starmapper and wormholes

Vincent Verhoeven — 2008-08-19T14:17:53

2008/8/18 Krzysztof Sobolewski interia.pl>: I didn't really expect Sun to be able to divide by zero, but well... [t 0] Error while drawing image! java.lang.ArithmeticException: BigInteger divide by zero at java.math.MutableBigInteger.divide(MutableBigInteger.java:780) at java.math.BigInteger.divide(BigInteger.java:1297) at jezuch.utils.starmapper3.Starmapper.calcSize(Starmapper.java:222) at jezuch.utils.starmapper3.Starmapper.drawImageConcurrent(Starmapper.java:582) at jezuch.utils.starmapper3.Starmapper$DrawTask.call(Starmapper.java:1034) at jezuch.utils.starmapper3.Starmapper$DrawTask.call(Starmapper.java:1013) at java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:303) at java.util.concurrent.FutureTask.run(FutureTask.java:138) at java.util.concurrent.ThreadPoolExecutor$Worker.runTask(ThreadPoolExecutor.java:885) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:907) at java.lang.Thread.run(Thread.java:619) Same thing, thus. This is with java-6-sun from Ubuntu: $ java -version java version "1.6.0_07" Java(TM) SE Runtime Environment (build 1.6.0_07-b06) Java HotSpot(TM) Server VM (build 10.0-b23, mixed mode) Iwanowitch

Re: Risk colonization segfault

Vincent Verhoeven — 2008-08-19T14:08:52

I believe result.first will still be NULL when numUnits == 0 for each bid, which happens when (for example) I place a Colonization order from a planet with only 1 army. You should also emphasize in the documentations that troops are moved from a planet to another planet with a colonization order, and bids have nothing to do with the reinforcement pool as I first assumed (and I believe vi1985 also thinks that - so it's not just me). Or perhaps this has changed during development or something? Anyway, I was a bit confused. Also, general learning moment... If I upload a core file, shouldn't I also upload the executable? Or can I assume that, when build on another machine, gdb will correctly find the symbols? Anyway, I don't think it's necessary in this case, but I still don't fully comprehend core files I think. Oh, another question on the same topic: if I have a core file from a stripped executable and I have a non-stripped version, can I somehow combine the symbols from the non-stripped version with the core file? This happened to me with my X server crashes from tpclient-pywx: I have a core file but it's from a stripped version. Is that useful in any way, either to me, the Ubuntu X maintainers or the X developers? Iwanowitch

Re: Risk "reinforce" order crashes server

Tim Ansell — 2008-08-19T04:00:14

Yes, it's always better to keep it in the thread. Just because you are no longer crashing the server (I'm assuming you are sending valid information now) doesn't mean someone else can't crash the server with this knowledge. This requires two fixes, one in your code and one in the server code. Tim 'Mithro' Ansell

Re: Risk "reinforce" order crashes server

Victor Ivri — 2008-08-19T03:48:59

Thank you, mithro. I already posted an notified the list that the problem is solved on a separate thread "SOLVED Re: Risk "reinforce" order crashes server". Perhaps it would've been a better idea to keep it in this thread!

Re: Risk colonization segfault

Tim Ansell — 2008-08-19T02:12:14

A "bt full" gives more information and can be a bit more helpful. Feel free to upload the core file to your public_html directory on git.thousandparsec.net. The following command should work, scp verhoevenv< at >git.thousandparsec.net:~/public_html/ You should then be able to access it at http://www.thousandparsec.net/~verhoevenv Tim 'Mithro' Ansell

Re: Risk "reinforce" order crashes server

Tim Ansell — 2008-08-19T02:05:56

A client should never be able to crash the server. Lee and Jphr can you please figure out what is causing the server to crash (I assume you should just be returning an "invalid operation" or similar). Tim 'Mithro' Ansell

Re: Mini-milestone on texture to globe projection

Tim Ansell — 2008-08-19T02:04:32

Hey jsmiller, That looks pretty cool. Where are the black dots coming from? Do you know what license these files are under? If they are public domain it would be good to add them to our repository. Tim On Wed, 2008-08-13 at 22:48 -0700, Jeffrey Miller wrote:

Re: Starmapper and wormholes

Krzysztof Sobolewski — 2008-08-18T17:31:36

Re: Risk colonization segfault

Ryan Neufeld — 2008-08-18T17:15:16

Strange that it is segfaulting on an IGObject::getName(). Looks like the exact place it fails is result.first->getName() in #2 , line 299 of colonize.cpp. Weirdest part is that result.first points to i- >first, and i->first is cast as a planet then asserted, so IMO it should fail at the assert and not after it; since i->first should be a valid planet IGO if it passed its assert. Ryan 'jphr' Neufeld --------------------------- visit me at hammerofcode.com On 18-Aug-08, at 11:55 AM, Vincent Verhoeven wrote: