bridgewater-brunel.me.uk - Simutrans-Experimental - Pak128.Britain-Ex 0.9.0

[jamespetts]

Ahh, but are both "A"s exactly the same stop? If not, one is not an "A" but a "C" that happens to be near an "A".

[chicken]

It seemed to be doing Blackchapel -> Axpool -> Blackchapel, so I believe that is the same stop.

[jamespetts]

Did you check the display of the passengers' ultimate origins and destinations? Presumably, they were not ultimately from Nestinghall Blackchapel Stop going only to there? Are any of the passengers at Nestinghall Blackchapel stop ultimately bound for that stop?

[chicken]

They all seemed to be going to various stops in Manford, a faraway city. I didn't discern any pattern in the origins.

[jamespetts]

The probability is that the routes were re-calculated whilst they were en route, so as to make their actual routes circular even if their planned route at any given time is not.

[ӔO]

I seem to have crashed the server when I tried to join.

something about a string being too long.

http://dl.dropbox.com/u/17111233/client2-network.sve

Code Select Expand

fatal error: loadsave_t::rdwr_str()
string longer (511) than allowed size (256)


The last save, that I have, from that is from 16h before.

[asaphxiix]

I have saves from 22:07 UTC, 21:03, and another from 20:27 (I believe they are stored on the server as well but I can upload if needed). Hurray new backup system!!


On a side note, I did see pax despaired from waiting to kirkmouth in coatsand (only one ship line goes there), starting to make loops on the map and eventually go back to coatsand. They shouldn't have left CS in the first place, but they were so excited to see a train going, they just boarded it!

[AP]

On a side note, I did see pax despaired from waiting to kirkmouth in coatsand (only one ship line goes there), starting to make loops on the map and eventually go back to coatsand. They shouldn't have left CS in the first place, but they were so excited to see a train going, they just boarded it!

Oh joy, the pax are getting too intelligent!  Maybe we need a rule against doubling-back... </can of worms>

[jamespetts]

The server is back again, restored from the 0100h backup. Apologies for the difficulty. It seems that the saved game was corrupted somehow - perhaps the server crashed during saving. It is not clear why this should be, as the server left no trace of the crash in the logfile:

Code Select Expand


Jan  2 01:52:41 438242 brit[25354]: Warning: nwc_sync_t::do_command:#011sync_steps 1765910
Jan  2 01:52:49 438242 brit[25354]: Message: network_command_t::rdwr:#011write packet_id=4, client_id=0
Jan  2 01:52:49 438242 brit[25354]: Message: packet_t::send:#011sent 14 bytes to socket[6]; id=4, size=14
Jan  2 01:54:13 438242 brit[25354]: Message: network_command_t::rdwr:#011write packet_id=12, client_id=0
Jan  2 01:54:13 438242 brit[25354]: Warning: nwc_routesearch_t::rdwr:#011rdwr limits=(1464, 10968, 12592, 6351264, 458383) apply_limits=1
Jan  2 01:54:13 438242 brit[25354]: Message: packet_t::send:#011sent 43 bytes to socket[6]; id=12, size=43
Jan  2 01:54:13 438242 brit[25354]: Warning: nwc_routesearch_t::transmit_active_limit_set:#011transmit succeeded sync_step=1765910 map_counter=146499501 limits=(1464, 10968, 12592, 6351264, 458383)
Jan  2 01:54:13 438242 brit[25354]: Message: karte_t::laden():#011Prepare for loading
Jan  2 01:54:15 438242 brit[25354]: Warning: karte_t::laden:#011Fileversion: 111002
Jan  2 01:54:15 438242 brit[25354]: Calculating textures ...done
Jan  2 01:54:15 438242 brit[25354]: World destroyed.


The critical part, as will be seen, is:

Code Select Expand


Jan  2 01:52:49 438242 brit[25354]: Message: packet_t::send:#011sent 14 bytes to socket[6]; id=4, size=14
Jan  2 01:54:13 438242 brit[25354]: Message: network_command_t::rdwr:#011write packet_id=12, client_id=0


The server skips two minutes in logging those two messages, so it crashed after sending the first, and the second is the first after the game starts to reload when the script to check whether it is running starts again.

Because the crash/corruption occurred at 0152h, the 0200 backup was of the bad version, and therefore I have had to restore the 0100h backup. I hope that people have not lost too much work.

AEO, did anything that happened when you were joining give you any clue as to what might have caused the difficulty?

[ӔO]

Nope, not a clue as to why.

The game was loading, and then it simply crashed once the transfer was done.

[jamespetts]

Very odd.

[asaphxiix]

server is down, it seems

I was lagged, made a command that never went through (at least not as far as I could see), then quit. I rejoined, and got the fatal error from above.

[jamespetts]

Hmm, odd. I managed to get a backtrace on this one: see here. It's not the best backtrace, as I didn't have debugging symbols enabled. It does suggest, however, that the issue is a double free error - something that does not directly originate from Simutrans-Experimental specific code, as I have not used malloc/free for any of the code that I have written. The saved game appears to be corrupted in much the same way as the previous games were - it is not clear how this is happening, I am afraid. The latest clean save was at 1544h (the corrupt save being at 1617h), so I have reverted to that for the time being. Apologies for the difficulties - this is a rather difficult issue to track down, I am afraid. Any assistance would be appreciated!

[asaphxiix]

Thanks James! I wish I could assist, maybe someday...  I'm sure you will be able to nail it soon.

Perhaps until fixing save frequency should be increased to 30 minutes? Funny thing, I haven't noticed any auto-saving going on, maybe it just looks like someone is joining. If it does run in the background without interrupting, perhaps even save every 10 minutes?

[jamespetts]

According to my logs, auto-saving is very rarely necessary, as the game has always been saved within the last hour. I could alter that to 30 minutes if it would assist? That would mean more regular force-syncing, however (which appears in game just as if somebody new has joined, but without any new players actually joining).

[asaphxiix]

I've lost more than an hour before... Personally I get really bummed from losing even a little work, so I'd be happy to pay the price to increase save frequency from 60 to 30 minutes.

[asaphxiix]

I wanted to save just now, so I saved the game locally and rejoined. Upon joining got the error message (didn't notice what message this time). Then I tried to load my local saved game, but it said:

FATAL ERROR: quickstone<T>::quickstone_tpl(T*,uint16)
slot (2) already taken

could this be related to the new backup system?


saved game

[jamespetts]

The backup system would not itself cause errors in the game. I have modified it now to check for a save within the last half hour.

I am sorry that there seem to be so many difficulties at present - there is not much that I can do until I get home where I can run things in a debugger. There seems to be some corruption happening on occasions when the game is saved, the reason for which is presently far from clear.

Does anyone notice any sort of pattern in these problems - any state of affairs or activity in the game world that makes it more likely that these problems will occur?

[asaphxiix]

Dear James

I hope you are enjoying the holidays! Forget about the problems, we are having a hell of a time in the game, if I may.

Perhaps a hiatus may benefit us, as well as our relations with our surroundings, and general nutrition.

[jamespetts]

I'm glad that you are enjoying it! Hopefully your enjoyment will soon be more uninterrupted than at present...

[jamespetts]

Oops - had incorrectly renamed the saved game when restarting the server. Should be restarted properly now.

[sdog]

the games on bridgewater appear to be quite involved and complex. Good examples of what simutrans can provide. If anyone of the players here is on g+ and in the simutrans community, would you be so kind to give a few glimpses of the game there?

There's a gameservers' category at the community where this would fit perfectly.

Showing interesting aspects, boasting with your network, giving sumaries of aeras, and anything else is appropriate i think. (i'm always looking for content that might draw in more people or increase their ties to simutrans)

[jamespetts]

Sadly, this seems to be down again. I shall have to leave it down for the present, as the frequency of failures is presently too high to be sustainable. I shall have to look into the various bugs when I get back home. I am most sorry for the trouble. For reference, the most recent intact save is available here.

Until I am able to fix the problem, might I suggest some forum threads in which addicts dedicated players might wish to participate in the time that they might otherwise be spending on expanding their vast and ever growing railway empires? The list is below.

• Reconsidering the speed bonus
• Server rules for online play
• Calibrating the passenger factor
• Modelling private car traffic without breaking the CPU bank
• [New release candidate] Simutrans 112.0 Experimental 10.9018
• 18th and early 19th century road transport
• Canal Overhaul
• A snippet of relative pricing information

I do hope to be able to put the server back online before too long. I do hope that not too much of your work has been lost.

[ӔO]

I think I know what is causing it.

Some of my ships are in replacement mode and they are not finding proper paths or they are getting stuck in depots due to corrupt schedules.

It seems to work fine after removing the offending ships.

[jamespetts]

Aha - thank you very much for that lead! That is most helpful. I shall look into that when I get home.

[chicken]

Thanks for all the work so far. I just wanted to mention an oddity I observed and I wonder if anyone else saw it. Sometimes when I was laying my railroad line I would get everything working and it would run fine. Then I would log out, go away, come back and and login again, and find that my railroad line was turned into a nasty spaghetti line -- although still connected on both ends -- instead of a straight line it was zig-zagging back and forth. So I would remove the tracks and relay them into a straight line and then it wouldn't happen again to that section. But it did happen again the next time I created a new segment of track.

Very strange bug, not sure how to reproduce. I'll continue trying next time. Has anyone seen anything like this?

[jamespetts]

I think that this is a known issue in Standard with the click and drag tools which I suspect might well have been fixed in the latest versions of Standard on which the current RC is based.

[prissi]

I have never heard of this for standard.

[jamespetts]

Ahh, perhaps I misremembered, then: but the code in question (the one click drag tool for ways) has not been altered for Experimental, and I do have a vague recollection of this being discussed for Standard somewhere - I think along the lines that the one click tool is built on the two click tool and that it is equivalent to clicking at one end and at the other, and occasionally one does not get the result shown by the drag preview.

[prissi]

For parallel diagonals one should always use the CNTRL-key, not matter if dragging or not. Should work in networkgames as well.

[chicken]

Well, again, the track was laid out correctly and I even watched trains run over it successfully. It only became spaghetti when I left and rejoined the game.

[jamespetts]

Thank you - that is helpful. Chicken - was it a parallel diagonal that you were trying to lay?

[ӔO]

maybe the server crashed and reverted?


I was thinking of a large joint station at Ingleminster. I can add more platforms if needed.

[chicken]

Just a single track, diagonal yes.

[jamespetts]

Hmm, if it was single (not next to another diagonal track), there is no need to hold down CTRL, but AEO might well be right that you suffered from a crash and revert, or even some sort of corruption of the saved game.