Running a tournament of Medusa only long lived games (300,000 moves) to see if it shows up any disqualification potential (it won't actually result in exclusion if it does - just for curiosity). It took about 70,000 moves to join up with itself vertically. It grows a lot bigger without other players present...

Haven't seen any orthogonal or diagonal growth yet though - only that in between angle

I love Medusa's non deterministic ant mills. Rings where a worker will get caught in a loop running round and round until eventually its randomised path takes it to just the right place to escape. Really not efficient though...

In the first game, with 300,000 moves and no competitors, Medusa amassed 165 workers, and ... 2 food. No disqualification though

Long term as the gaps between the tendrils fill in, the environment around the queen becomes a hilly landscape with local minima just out of reach of the queen, which fill up with laden workers. Food still reaches the queen because there are enough workers to fill up these valleys preventing any more workers falling into them...

@trichoplax I think we determined it was an accidental result of the modified controller I was using that had modifications I forgot about

Ah OK

Ha, vampire found forensic

Failed to leech, but still, amusing

(Game finished and Pierce won 1st, Ziggurat 2nd, followed quickly by Lone Wolf)

I have now had a crash in Firefox too, so it may not be a Chromium related problem after all. I'm going to add extra logging locally (which will slow things down) and run in both Firefox and Chromium to see if I can narrow it down to a specific player causing the problem. It may not be, but when I forced the script to stop it appeared to be in an ant Function, so it's a possibility to explore

Huh

The controller has no way of stopping an ant function. It can only time it and disqualify if it takes too long. If it never returns then the whole controller is crashed

Yeah

If one of the players has a rare condition that leads to an infinite loop then I would expect this behaviour.

Should have thought to check for it sooner, but I tend to assume problems must be in my code...

I tried figuring out which ant was at fault too, and the BEST I could define was that it was a Glider Worker 2...

But if I took that input and plugged it into my dev controller

So using webworkers allows terminating a game?

it returned a move in zero time

@trichoplax Dunno. Would depend on what his code does. It might just GC the webworker and create a new one.

I suppose even if I track it down to a particular player then the problem could still be due to my code, so unless I can pinpoint why it crashes, I can't disqualify that player

mhm

which is why I was confused when I tracked it down to Glider. As far as I could tell, Glider was behaving properly. Just that the returned move was never processed.

(Glider uses no loops that could ever infinite loop, they're all for(x=0;x<9;x++)

Maybe if I can get it to identify a player several times then I can see if it's a different player each time. Some kind of hint of where to look next

Logging doesn't seem to be slowing things down too much. It doesn't need to log anything for cached moves so it's not as bad as I feared

Mmm