Adjudicator bug?

[kestasjk]

I got a report of an adjudicator bug. I think it's been over a decade since we last had a confirmed report of an adjudicator bug so I always give the adjudicator the benefit of the doubt, but this does seem like a bug:

https://webdiplomacy.net/board.php?gameID=455699

Previous turn positions:

B.png (10.56 KiB) Viewed 2749 times

Current turn positions:

A.png (11.99 KiB) Viewed 2749 times

Surely this violates the no self-dislodge rule, but I can't believe this wouldn't have come up until now, the adjudicator hasn't changed in almost 15 years.

Anyone see anything I'm not?

[Chaqa]

Could it have something to do with Bulgaria having two coasts and how that interacts with Constantinople as a canal? Plus Italy's support. you could make a dummy game to test it out manually.

[JECE]

Wow! This does look like a misadjudication to me too.

Have you tried double-checking that webDip still passes all the DATC tests?

[JECE]

Kestas, did you catch this from the global chat of this game?
"2. When I put in the Pause vote a couple of days ago, the game also entered a Draw vote, which I did not enter."

It reminds me of a serious error that I experienced back in 2009:
https://webdiplomacy.net/forum.php?threadID=499708

[JECE]

"Maybe we can set a record for longest game in actual days."

The Dual Monarchy has no idea what they're wishing for.

[Yonni]

Forget dystopian Facebook Diplomacy, this is the most exciting thing to happen here in years.

[MerryBBuck]

Maybe BUL->CON was not an attack, but a special military operation.

[Bonatogether]

that's crazy

[Mark Murray]

I have wondered whether one player changed their move at the last instant, just as the deadline time approached or even nearly simultaneously. I don't know if that happened, or it that would affect anything.

[kestasjk]

It's quite something this bug has been undetected for so long. The core adjudicator hasn't changed in over a decade now!

I've added a test case to check this:

And I've changed the adjudicator so that a unit has 0 support strength against their own country when doing a country check. All the other tests are still passing.

The adjudicator does check that you aren't attacking / supporting the attack of your own unit, but because there is also a paradox / loop that ends at just the right spot it broke out before that clause.
When it detects a paradox it needs to find the minimum and maximum amount of support possible to see if the move succeeds/fails without solving the paradox, as solving the paradox (by failing all moves in the loop) should only happen if there's no other way.


I recently got back in touch with the DATC author to try and put him in touch with the company working on the new Diplomacy board, so he could make sure the rulebook is solid, so I might show him this and see what he thinks & whether it should be added as a test etc.

[kestasjk]

Regarding the game I fixed the bad adjudication so things can continue, thanks for reporting

[kestasjk]

Kestas, did you catch this from the global chat of this game?
"2. When I put in the Pause vote a couple of days ago, the game also entered a Draw vote, which I did not enter."

It reminds me of a serious error that I experienced back in 2009:
https://webdiplomacy.net/forum.php?threadID=499708

Damn, that must've left quite a scar for you to remember it after 13 years :| Votes are now logged when you place them now so we can see here:

Code: Select all

+--------+-------------------------+-------------------+------+-------------+
| gameID | FROM_UNIXTIME(timeSent) | message           | turn | phaseMarker |
+--------+-------------------------+-------------------+------+-------------+
| 455699 | 2022-11-22 14:14:57     | Voted for Pause   |    7 | Builds      |
| 455699 | 2022-11-24 05:00:04     | Voted for Draw    |    8 | Diplomacy   |
| 455699 | 2022-11-25 19:35:30     | Un-Voted for Draw |    9 | Diplomacy   |
| 455699 | 2022-12-01 23:12:50     | Voted for Pause   |    9 | Retreats    |
+--------+-------------------------+-------------------+------+-------------+

The pause/draw votes were a long time apart so I'd say this is more likely a mis-click