Multi Player bugs and crashes - After the 13th of August 2013

Please report all multi player bugs and crashes from the 13th of August 2013 (release v32 = SVN Revision 5949)

When posting a an OOS, a bug or a crash please include

Deleted OOS Report

Toffer90 said:

Only played 128 turns this time. wanted to make a quick check if multi-threading issues were fixed, seems pretty stable.

Chaser (player 0) attacks crocodile, end turn, immediate OOS.

View attachment 359404

Click to expand...

The random logs math perfectly, but the player OOS logs show a different amount of damage on a unit at one point. That suggests that an async rand is being used somewhere to determine damage, either in combat or possibly something like terrain damage, or damage meted out via Python (if that ever happens).

This is not really my area, so apart from the above diagnosis I am not planning to work on this in the short term, but TB may do...

Koshling said:

The random logs math perfectly, but the player OOS logs show a different amount of damage on a unit at one point. That suggests that an async rand is being used somewhere to determine damage, either in combat or possibly something like terrain damage, or damage meted out via Python (if that ever happens).

This is not really my area, so apart from the above diagnosis I am not planning to work on this in the short term, but TB may do...

Click to expand...

So correct me if I'm wrong but ALL of combat takes place within a synched environment right? Terrain damage never involves randoms. Auto Promotes - not quite sure when they fire but they wouldn't be async as there's no UI interface to them where a rand would be called. And I'm unaware of any combat python - nothing that would influence damage anyhow - capture of ships though, which I'm pretty sure has nothing to do with this.

So that leaves events - particularly an event with an option, as the primary suspect I think. Perhaps the event that heals a unit? Even that's a little odd for it to have that effect though.

Another region it could be - differing Bug options such as Defensive Withdrawal, Stack Attack, etc...

Thunderbrd said:

So correct me if I'm wrong but ALL of combat takes place within a synched environment right? Terrain damage never involves randoms. Auto Promotes - not quite sure when they fire but they wouldn't be async as there's no UI interface to them where a rand would be called. And I'm unaware of any combat python - nothing that would influence damage anyhow - capture of ships though, which I'm pretty sure has nothing to do with this.

So that leaves events - particularly an event with an option, as the primary suspect I think. Perhaps the event that heals a unit? Even that's a little odd for it to have that effect though.

Another region it could be - differing Bug options such as Defensive Withdrawal, Stack Attack, etc...

Click to expand...

I think differing options would show up as a differing random number sequence, since one side would roll for (say) withdrawal, where the other side would make no roll.

Good point. Baffling basically...

Hi,

i'm new here (about posting stuff not reading) but i tried C2C since 2 days and i have a probleme with Multiplayer game.

It's easy, i can't generate a map.

I launch the game with the mod, i create a game in multiplayer, i tried with different maps, i click Launch and after 30s, back to the desktop.

I play BTS 3.19, C2C v32, Last patch for C2C found in this forum, Blue marble graphic mod, in French, with default map option but custom game options.

Everything is ok in solo mode.

I tried lots of different mods without any problem, in solo or multiplayer mode.

I know i need to give you logs but which one ? (what files, located where ?) Sorry but the first post explaining is not clear for a beginner -OOS & standard ?

lolcivf said:

Hi,

i'm new here (about posting stuff not reading) but i tried C2C since 2 days and i have a probleme with Multiplayer game.

It's easy, i can't generate a map.

I launch the game with the mod, i create a game in multiplayer, i tried with different maps, i click Launch and after 30s, back to the desktop.

I play BTS 3.19, C2C v32, Last patch for C2C found in this forum, Blue marble graphic mod, in French, with default map option but custom game options.

Everything is ok in solo mode.

I tried lots of different mods without any problem, in solo or multiplayer mode.

I know i need to give you logs but which one ? (what files, located where ?) Sorry but the first post explaining is not clear for a beginner -OOS & standard ?

Click to expand...

Blue Marble probably wont work with C2C as we have a different set of terrains.

Why everything is ok in solo and not in multiplayer ?
Blue marble is not a multiplayer mod so i don't think that's could be a problem for creating a multiplayer game.

B;ue marble wont work right in single player either. We have different terrain sets.

Ok for solo game but i have no issue in solo game. I will disable BM.

How to fix multiplayer ?

Thunderbrd said:

So that leaves events - particularly an event with an option, as the primary suspect I think. Perhaps the event that heals a unit? Even that's a little odd for it to have that effect though.

Another region it could be - differing Bug options such as Defensive Withdrawal, Stack Attack, etc...

Click to expand...

We played with identical BUG options, same SVN revision (ca. 5990) and no events option.

Continued the same save as last report is from, but updated to rev. 6061 to try out the new memory management system (turned off viewport). The new system seems very stable and effective, thank you Koshling.

View attachment 360007
View attachment 360008

Both OOS events seems to have happened during AI turn.

Toffer90 said:

We played with identical BUG options, same SVN revision (ca. 5990) and no events option.

Click to expand...

Huh... I suppose that means we're left with something in the combat mechanism going async... I'm really surprised that could even be happening. I'll keep an eye out for something that might be able to explain this as I go about further programming in that dept. There's some caching in the commander structure but I don't suspect that commanders have anything to do with it because if I'm not mistaken, it happens with battles that don't involve them. hmmm... withdrawal becomes a little suspect but again I don't think anything is taking place async there. I wonder if these battles that cause OOS errors are taking place in the same plot on both systems... differing plot modifiers would account for a varied hp result at the end of combat, which could point to differences in ai processing regarding where one of the units had moved to and thus differences in the defensive modifiers at that plot. Such an OOS error would be rooted in the ai random.

Now another thing is... are there differing ways randoms can be generated, one of which causing a potential oos??? Not sure on that...


Hey... also... since you're gathering these OOS reports, let me ask you, are you running with the multi-core processing engaged or have you turned the multi-core mechanism down to 1 core processing? Last I played with 4 we had OOS errors every round but I have lost the OOS reports and haven't had focus on the situation enough to gather another set.

We are indeed testing the MP stability of multithreading; the OOS every turn issue was fixed by Koshling during the short period it was configured in SVN by default as single thread in A_New_Dawn_GlobalDefines.

btw
My hunch, about the last three OOS we had, is that it is related to combat outcome.

We have been busy bees.
All except OOS-12 happened during AI turn.
OOS 12 seems to be related to critical hit or/and withdrawal in combat.

View attachment 360079


Edit: Wanted to try a new game with simultaneous turns, but was wondering if that may cause more OOS problems. Anyone who knows anything about this?

EditEdit: Simultaneous turns did not seem to have any impact on OOS frequency, we played 504 turns before the first OOS happened.