Multiplayer crash (87b435c)

[Magnate]

There might be a way for us to override that, but I'm not immediately seeing it. It could be you'll need to avoid pauses in the tens of minutes-- either sending a little message every 10-20 mins or so, or else just saving/quitting and then restarting later.

It should be possible to make the server and clients send pings to eachother ever minute or so. That said, several years ago I attempted to set up an auto turn timer function on the server, but couldn't get it to work.

If I ever stop playing this wonderful game long enough to contribute to it, it will be this. There's no reason games shouldn't be able to be up 24/7.

[Vezzra]

This latest one, attached, is a timeout crash. There's a long pause between turns (like an hour or more - because one of us is cooking or something), and when the second player (me) clicks end turn, it crashes per (i) above. This time I've asked the other player for her logs too.

ok, please post them once you get the other freeorion.log file

I think he did that in the post directly above yours...

Got them - other player's logs attached, from the same crash at the same time.

Not sure if you've missed that, so I wanted to mention it just in case.

[Dilvish]

I think he did that in the post directly above yours...

Got them - other player's logs attached, from the same crash at the same time.

Not sure if you've missed that, so I wanted to mention it just in case.

I had missed those, thanks very much for pointing it out. The server log was just from an SP game (which makes sense), and the client log indicates that their client simply crashed-- the log file terminates abruptly with no warning. So that doesn't seem to me to fit with the hidden-networking-timer theory. Magnate, next time this happens, please again ask your friend to send their freeorion.log file, and also ask them if they had any kind of error message along with the crash, and just what they had done prior to the crash.

It could be that there is some instability with the multiplayer messaging that is causing memory misuse and crashes. I have often encountered this in the past month when testing AI issues in a moderated multiplayer game, either the AI will start having strange errors with its ShipDesigner cache, or in some cases the client will simply crash to desktop when I try zooming or moving the starmap around. I didn't used to encounter that problem at all, but the AI ShipDesigner code works fine in long single player games, so I don't think that is the problem. I had thought it was likely to be something about the moderated game (because at times it clearly flags crashes due to issues with moderator OrderSets and auto-save), but perhaps there is some broader instability that has crept into multiplayer messaging. I'll try experimenting a bit with Observer games.

[Vezzra]

A few years ago, when I implemented AI process priority management, I stumbled across some oddities during order submission processing / save game processing. Geoff and I had a brief discussion, but came to no conclusion, I kept a reference in case something turns up in the future that might be related to that - maybe that's the case now.

See the "[BUG] Weird sequence of messages/events in order submission" thread.

[Magnate]

Thanks all for your willingness to help diagnose this - I will keep sending log files from crashes. AFAICR I have never had a crash in a single-player game in any version, and all our multiplayer crashes happen during turn processing.

As an aside, what is a moderated game?

[Magnate]

Ok, another two crashes. First one a normal during-the-game crash - my assumption is that we'd taken too long on a turn. I clicked End Turn, it got halfway (downloading the first set of data) and paused for a long time before "The connection to the server has been lost". Logs attached: the zip file is mine, the other log file is the other player's.

Then when we started it crashed shortly after I clicked "Start Game". Again I suspect a timeout - I had loaded up the game before popping out to the shops, the other player had joined while I was out, and when I got back (about 45 mins later) I added the AIs and started the game. No logs for this one I'm afraid - I was too quick to restart without thinking.

Out of interest, am I the only person experiencing frequent crashes in multiplayer? I can't see any other recent reports.

Also, would it help to rotate log files so we keep the last five or ten (like autosaves) instead of always overwriting the last one? That way I can guarantee to provide logs for every crash without compromising my friend's gaming!

[Vezzra]

Out of interest, am I the only person experiencing frequent crashes in multiplayer? I can't see any other recent reports.

My guess is that not very many people use multiplayer.

[Magnate]

Another crash, identical to the previous one - new logs attached (mine in the zip, hers separate). This time I can confirm that I was AFK for less than 20 minutes (cooking!). Crash was again at the mid-point of turn processing (turn 171).

It seems to be more crash-prone later in the games.

[Dilvish]

As an aside, what is a moderated game?

The person setting up the multiplayer game can, instead of playing an empire, be a 'moderator' with ability to make various changes as the game progresses (give/take techs, planets, etc.) akin to the "Dungeon Master" of classic role playing.

As for these logs, it seems we keep not getting enough info in the logs to really identify the cause of the crash. I will put it on my to-do list to try to look over things and see where we could add some more crash related logging. I'll post back here if I make any changes on that front. In the meantime, it seems probably not worth the bother for you keep posting these logs, I'm afraid.

[Magnate]

As an aside, what is a moderated game?

The person setting up the multiplayer game can, instead of playing an empire, be a 'moderator' with ability to make various changes as the game progresses (give/take techs, planets, etc.) akin to the "Dungeon Master" of classic role playing.

As for these logs, it seems we keep not getting enough info in the logs to really identify the cause of the crash. I will put it on my to-do list to try to look over things and see where we could add some more crash related logging. I'll post back here if I make any changes on that front. In the meantime, it seems probably not worth the bother for you keep posting these logs, I'm afraid.

I understand. I'll try to record useful observations and patterns.