Multiplayer Woes

[Aevum]

v2167 (post v22.1)
We keep getting OOS very frequently. Sometimes they happen mid turn, flashing and go solid. When it happens mid turn I don't get an OOS log sadly. I swear one of the times the oos happened after discovering some barb sea animals. My carval sunk a cog... then moved into the darkness revealing a host of animals. BAM oos.
I compared the two oos logs and only found small naming differences. like:
Colonist --- Colonist (Thebes), on four or five units. Is that likely to do anything?

 

[AIAndy]

v2167 (post v22.1)
We keep getting OOS very frequently. Sometimes they happen mid turn, flashing and go solid. When it happens mid turn I don't get an OOS log sadly. I swear one of the times the oos happened after discovering some barb sea animals. My carval sunk a cog... then moved into the darkness revealing a host of animals. BAM oos.
I compared the two oos logs and only found small naming differences. like:
Colonist --- Colonist (Thebes), on four or five units. Is that likely to do anything?

The naming stuff is normal and does not effect the synchronization state.
No difference in the position of units?

 

[Aevum]

Out of the 5 or so naming differences i noticed this!!!!

Player 1, Unit ID: 401433, Galleon
X: 31, Y: 30
Damage: 0
Experience: 11
Level: 4
----------
Player 1, Uiit ID: 401433, Galleon
X: 39, Y: 37
Damage: 0
Experience: 11
Level: 4
---------
What would be the solution here, delete the unit, reload and solve itself? Cause' we're being plagued by oos, although i don't know if its causeing it everytime or once.

 

[AIAndy]

Out of the 5 or so naming differences i noticed this!!!!

Player 1, Unit ID: 401433, Galleon
X: 31, Y: 30
Damage: 0
Experience: 11
Level: 4
----------
Player 1, Uiit ID: 401433, Galleon
X: 39, Y: 37
Damage: 0
Experience: 11
Level: 4
---------
What would be the solution here, delete the unit, reload and solve itself? Cause' we're being plagued by oos, although i don't know if its causeing it everytime or once.

The problem is not the unit per se but something about the movement that does not happen the same on all computers in the game.
So deleting it will not solve anything but we need to find out what part exactly causes it.

So my questions:
Were the settings that influence pathing set to the same on both computers? Especially the new AI pathfinder option?
Were the units automated and if yes, were the settings that might influence automation behavior the same on both computers?
Does it only happen with sea units or also with land units?

In general, it may well be that something in the changes that Koshling made in the pathing system is not sync safe.

 

[Aevum]

I'll double check tomorrow, but I'm 99% sure that neither of us had the AI pathfinder option on. Should we be useing this? I never really tried it out so i don;t know the benefits.
No, it was not automated, all my galleons but this one are in a city, and this one is ferrying a colonist and some units.
As for right now, I'm saying its sea units. Of the most recent 3 or so OOS I remeber, they all occured after moving ships for me...

I'll try and play more tomorrow and see if I can't get more OOS logs.

 

[AIAndy]

I'll double check tomorrow, but I'm 99% sure that neither of us had the AI pathfinder option on.

The important thing is that it is the same setting on both computers or at least that I know what it was set to.

 

[Koshling]

The important thing is that it is the same setting on both computers or at least that I know what it was set to.

Until very recently, it was possible (though unlikely) for the caches to act non-deterministically (well strictly that's not true, but might as well be) if the Zobrist hashes of the plots with set differently on each game (which they may well be?? They are just random numbers generated at map initialisation time). The cache key for a movement graph edge value was the XOR of one plots zobrist hasgh with a rotation of the others, and this menat that occassionally it was possible to get a hash clash, which could result in incorrect plot daat being returned during pathing calculations.

However, I changed this about 2 days prior to v22 release when I realised that the plot indexes themselves could fit within an int (on any currently supported map size at least), and changed to using plot index of one plot + shift of plot index of the other. This would have eliminated the potential for hash clashes on the edge cache.

So (to finally get to the point), I assume this is v22 we're talking about here (or a SVN version that is no earlier than v22)?

 

[Aevum]

yes sir. v2167 (post v22.1) and I can confirm 100% that no one had AI pathing option on.
Also we have the exact same bug options. As the user setings folder was deleted and copied over to ensure no msitakes.

 

[Aevum]

New OOS, same game, same issue...navy ships are always our culprit. this time:

Player 1, Unit ID: 434300, Man'O'War
X: 21, Y: 51
Damage: 0
Experience: 15
Level: 4

Player 1, Unit ID: 434300, Man'O'War
X: 31, Y: 53
Damage: 0
Experience: 15
Level: 4

 

[BlueGenie]

Auto move, as in set a Go To (or right clicked some turns movement away)?

Cheers

 

[Aevum]

Nope, the ship in question was moved manully, all in one turn. Open completion of the move, OOS occured.

 

[AIAndy]

What promotions do the ships have?

 

[Aevum]

The ship that caused the above error had:

Stregth 10%
Hunter 1
Assistance of An
Navigation 1, 2 & 3.

 

[Aevum]

Tried to play another turn, OOS upon ship movement & combat:

Next Soren Rand Value: 8634
Next Soren Rand Value: 1864

Player 0 Assets: 5149
Player 0 Power: 2860
Player 0 Num Cities: 11
Player 0 Num Units: 116

Player 0 Assets: 5152
Player 0 Power: 2870
Player 0 Num Cities: 11
Player 0 Num Units: 117

Player 0, Sloop Of War, Unit class count plus training: 8
Player 0, Sloop Of War, Unit class count plus training: 9

Player 0, , Unit AI Type count: 9
Player 0, , Unit AI Type count: 10


Player 0, Unit ID: 311371, Sloop Of War 5 (Thebes)
X: 21, Y: 21
Damage: 0
Experience: 22
Level: 5

----------Blank SPACE


Player 1, Unit ID: 417895, Privateer
X: 20, Y: 22
Damage: 76
Experience: 13
Level: 4


Player 1, Unit ID: 417895, Privateer
X: 20, Y: 22
Damage: 0
Experience: 10
Level: 4

seems like there was no communication that a ship was sunk between the 2 computers.

 

[AIAndy]

While the problem is likely caused somewhere in the pathing, there has been a lot of code rewriting there recently (and still ongoing), so it is difficult to find out where in there.
Best would be if you somehow achieve to make a reproducable test case, meaning a save game that can cause the OOS when following specific instructions.

 

[Thunderbrd]

So... we got the second computer established at this point and we've been desperately working through a number of connection issues to work out how to get a LAN game started.

At this point, we're up against strange stuff that must have to do with something regarding the mod itself as we can see each other's games in the staging room of the basic BTS but cannot when attempting to establish a game on the mod itself. ??? Any theories as to what could be the issue here? I'm at a loss.

 

[AIAndy]

So... we got the second computer established at this point and we've been desperately working through a number of connection issues to work out how to get a LAN game started.

At this point, we're up against strange stuff that must have to do with something regarding the mod itself as we can see each other's games in the staging room of the basic BTS but cannot when attempting to establish a game on the mod itself. ??? Any theories as to what could be the issue here? I'm at a loss.

The staging room tends to not work properly with mods. I recommend using Direct IP.

 

[Thunderbrd]

Thanks for the advice... I'd never worked out how to do a direct IP but once I did, it seems to work just fine... funny - never had trouble with the Lan connection before with C2C so its just a bit odd. But that said, I think the connection is a better one on IP anyhow.

We had enourmous troubles at first, even on IP - horrific OOS errors out the gate, but some reading on here suggested that it's map dependant and I think it just had to do with Mongoose. We changed to a PW map and its working nicely now.

 

[Aevum]

OOS, SVN 2251

Next Soren Rand Value: 1459
Next Soren Rand Value: 6904

I'll fetch the rand logs soon, and I'm still trying to get a repeatable OOS on ships, see if I can get to that tonight.

 

[AIAndy]

OOS, SVN 2251

Next Soren Rand Value: 1459
Next Soren Rand Value: 6904

When the rand values differ, the rand log shows where they started to differ.