View source for User talk:Pa3k

Hey dude, are you seeing any errors in your rumble client for DrussGT 2.4.9? I see a few weird results from your client: [1] [2] [3] [4]

I know you've been running clients for a while with no problems, so maybe it's something wrong with DrussGT. I think he logs errors to disk, so if you could check roborumble_dir/robots/.data/jk/mega/DrussGT.data (or something like that) for any exceptions, I'm sure it would be appreciated. =)

Also, belated welcome to the RoboWiki! Feel free to make pages to tell us about your bots sometime. =)

Hey,

I am not sure what happened, I also saw that result being uploaded and it looked weird. There are no data.

I am running roborumble twice, with 2 different directory trees to utilize 2 cores. pb@pb-P52F:~/workspace/Robocode$ ls roborumble/robots/.data/jk/mega/DrussGT.data pb@pb-P52F:~/workspace/Robocode$ ls roborumble2/robots/.data/jk/mega/DrussGT.data pb@pb-P52F:~/workspace/Robocode$

I'll do some pages for my bots, when there is anything interesting to say :-), so far just imitating others' work with bad results.

On unrelated topic, do you know darkcanuck? I am trying to get a flag, but his not responding on wiki or email.

Regards

re: "so far just imitating others' work with bad results."

Don't sweat it. My bot is in the top 40 now, but a year and a half ago, it was about where Viper is now. Just keep thinking and trying.. and don't be afraid to rewrite large sections of code, or the entire bot if you realize a better way to put it all together! As it stands now, deBroglie is a rewrite of a rewrite of a bot I hacked together from disparate pieces I'd done as demos for students. There's no trace of that original code, but that's the lineage.

Keep at it! :)

Nice link. =) That definitely describes Dookious. But it also describes Tron, whose major versions are left in as separate RoboRumble entries. I guess Robocoders can't agree on it either!

That's my other degree (philosophy) showing there. *grins* It is a fascinating issue, even in Robocode. The only thing that connects deBroglie now to the bot I originally named Newton is just that I've decided to consider it the same project in my head.

BTW. cg on the first place

As far as I can tell, the bad results only come up for your client. I'll do a re-release and see what happens.

And if there were any errors they should have been logged. It would come up as <opponent name>.dat with the contents being a stack trace. The only way for there to be a 0/0 on bullet damage with the other bot getting 35 survival as well as data files not being written is for DrussGT to fail to load or being kicked out by the security manager. So maybe try re-installing your robocode client? Or see if you can run DrussGT in the regular Robocode on each client?

Try running the roborumble.sh in each of your client folders and manually run a battle against DrussGT. If it gets banned or doesn't run it should be fairly obvious, and you can test which one it is before deleting and re-downloading. It will also help confirm what the problem is.

How do I force roborumble to execute DrussGT battles?

I think he means robocode.sh, as in just run Robocode from your RoboRumble directory and run a battle with DrussGT, maybe against one of the bots we saw a zero score for. With any luck, you'll see the problem in action and get an idea what's going on.

doh... yeah, what Voidious said. Run Robocode from the rumble directory and see if DrussGT gives errors.

I'm going to be without internet for the next few days so if someone could re-release 2.4.10 as 2.4.10a once this is sorted out I'd be much obliged :-) (2.4.10 is just some minor bugfixes, I'll add details when I'm back)

You could also add jk.mega.DrussGT* to the EXCLUDE field in your roborumble.txt, if that's the only bot you're having problems with. But it would also be good to try some manual battles and see what's going on. After that I can repost DrussGT 2.4.10.

jk.mega.DrussGT* was excluded for now, I'll post a reply when I find out why Druss was having issues.

Thanks for the repost while I was away =)

Hmm sorry mates.. looks like something is wrong with my client .. Druss get allways 0 battles ... maybe i can check what it is going on :(. Maybe you should rerelease the last version too. Sorry for the inconvenience.

Take Care

Well it turns out that the same is happening on my mac client. No errors or anything it just drops the battles while all other battles are fine. Not sure how but looks like an issue with unix/linux? Maybe i can find out more - and will have a look at your bot in the visual client.

Hmm very strange ... this is the output:

Round 1 of 35
=========================
My hitrate: NaN
Actual hits: 0
DC gun score: 0
DCAS gun score: 0
gun: DC
SYSTEM: DrussGT 2.4.11 skipped turn 3
SYSTEM: DrussGT 2.4.11 skipped turn 4
...
...
SYSTEM: DrussGT 2.4.11 skipped turn 33
SYSTEM: DrussGT 2.4.11 has not performed any actions in a reasonable amount of time.
SYSTEM: No score will be generated.
=========================
Round 2 of 35
=========================
SYSTEM: This robot has been stopped.  No score will be generated.
Enemy damage: 0.0
My damage:    0.0
Accumulated, weighted enemy hitrate % : NaN
Flattener enabled: false
My hitrate: NaN
Actual hits: 0
DC gun score: 0
DCAS gun score: 0
gun: DC

If i start the match with "paused" game and make a single step, everything is ok and he finishes the match like he should. But as far as i start the match just normal it gets instantly disabled. Do you have some heavy object loading at the start or something?

I figured it out. I was also getting 4-6 skipped turns on start, but not enough to disable the bot. My FastTrig class wasn't being initialized until the first call to it. This might be somehow related to Java's lazy class loading. I've uncommented the hard initializing the FastTrig now in the main DrussGT class and I'm not getting any more skipped turns. Could you try with the old 2.4.8 to confirm that the problem doesn't happen? Thanks

Hi mate. 2.4.12 works fine for me.

Hmm looks like your bot give me a hart time :) ....

Fighting battle 2 ... jk.mega.DrussGT 2.4.12,et.Predator 1.8
Robot jk.mega.DrussGT 2.4.12 is not stopping.  Forcing a stop.
java.lang.ThreadDeath
        at java.lang.Thread.stop(Thread.java:758)
        at net.sf.robocode.host.security.RobotThreadManager.stop(RobotThreadManager.java:205)
        at net.sf.robocode.host.security.RobotThreadManager.stopSteps(RobotThreadManager.java:187)
        at net.sf.robocode.host.security.RobotThreadManager.forceStop(RobotThreadManager.java:158)
        at net.sf.robocode.host.security.RobotThreadManager.waitForStop(RobotThreadManager.java:143)
        at net.sf.robocode.host.proxies.HostingRobotProxy.waitForStopThread(HostingRobotProxy.java:148)
        at net.sf.robocode.battle.peer.RobotPeer.waitForStop(RobotPeer.java:1577)
        at net.sf.robocode.battle.Battle.finalizeRound(Battle.java:476)
        at net.sf.robocode.battle.BaseBattle.run(BaseBattle.java:217)
        at java.lang.Thread.run(Thread.java:662)
RESULT = et.Predator 1.8 wins 2220 to 0
Fighting battle 3 ... jk.mega.DrussGT 2.4.12,zch.David 0.21
Can't load jk.mega.DrussGT 2.4.12, because it is invalid robot or team.
Skipping battle because can't load robots: jk.mega.DrussGT 2.4.12,zch.David 0.21

Maybe you can find the issue with this little error snippet. Before this break i had a couple of fights where DrussGT did well, so it looks the original issue is solved.

Every time I get the "invalid robot or team" error, I do the following:

1. run RoboCode in my rumble directory. (That opening splash screen will take forever with so many bots. be patient.)
2. do Options -> Clean robot cache (this one takes a while too.. watch the terminal screen and wait for "cleaning done")
3. close

That clears up lots of weird rumble issues for me.

I guess you can just remove the "robot.database" file in the robots/ directory and the next time you start your client it will be rebuild. At least this is what i do from time to time.

I'm not sure what that is, however I did add some things to the static initialization, maybe it's too much now? I've gone back to 2.4.10 with just the FastTrig fix, let's see what happens.