Large map with 193 convoys "freezes" at month start

[jeffatsqi]

I am trying to get play experience with the Britain/UK large map and Simutrans Experimental. Specifically using the SR 1750 V2 map created by AP and the Experimental 7.1.

I have been playing to game "forward" with just passenger service until goods transportation modes that are economical show up. I now have 193 convoys in year 1815.

The game now "freezes" frequently at the start of the month. The game is working very hard - consuming larger amounts of CPU - but never return control to the user interface.

The game file is upload here:

http://simutrans-germany.com/files/upload/SR+rivers_1750_102.2.1_v2%20tial%20v16.sve

Also, I tried modifying some of the config files settings to see if they had any effect. The only change that made a difference was  changing

max_route_steps = 1000000

to

max_route_steps = 10000

With this change the game freezes less frequently - 4 to 6 months - as opposed to a month or 2 with the original.


This problem became more frequent as the number of convoys incresed.

[Junna]

Save game freezes for me in June 1815.

I don't think it is the number of convoys per se, but more likely a combination of issues such as the sheer size of the map (sure lags pretty bad whenever it tries to place industry), or a problem relating to pak Britain-ex, as I've had over 500 lines and a 1000 convoys with pak128 and only once did it freeze up like this save game does.

[AP]

Interesting. Glad to see the map is being used/tested, that's what it was made for.

It was certainly speculated that the larger maps, whilst fun, might be too much for some older PCs, don't know if that's an issue here. Alternatively they may just be too much for the game to handle, then. If you identify anything with any certainty I'd be interested to know, if we have to limit map size in future we can of course.

PS -am away for the next week or so but back thereafter

[Bernd Gabriel]

While freezed simutrans spends a lot of time calculating new season and snow line dependend images.
15,416,848 tile images have to be re-calculated!

Some questions:

Which kind of processor works in your computer and at which speed?

Which executable do you use? Self-compiled or official one?
A debug version is extremly slower compared to an optimized release version.

At which window size do you play?
A smaller window or less frames per second reduce the redraw effort.

[jeffatsqi]

My intuition - developed over years of experience as a software executive - tells me I am reporting a bug and not a condition of slow responsiveness. Also, I think the bug is a "race" condition that shows up from time-to-time.

Here is why:

1. When the game freeze/failure occurs it NEVER returns. I have let it run for 10 hours on a quad core system, using 100% of one core, and it does not return. This is a loop condition and not a workload issue.

2. If you start with the file I uploaded and run it a number of times you will see that the freeze/failure will occur at the start of different months. Thus it sometimes completes the workload for a given month and on the next run will fail at the same workload. This is a code issue.

3. The freeze/failure issue shows up sooner on slower CPUs than fast. On my old laptop it now fails in a month or two. On the quad core deskside in average time-to-failure is in the 5 to 7 month period. In fast forward both systems will fail in the first month almost always.

I am using the Ex 7.1 download with 128.Britan Ex 0.4. I think AP used that to build the game map but we need to ask him specifically.

I will get the hardware details when I get home.

[jamespetts]

Jeff,

thank you very much for your bug report: apologies for not replying sooner. I am having some trouble in loading your saved game, I am afraid: it tells me that there is an invalid halt co-ordinate. The same error appears for me in both 7.1 and the current incomplete -devel branch of 7.2. Can you try re-saving and re-uploading to see whether this fixes the problem? Incidentally, what platform are you using? Thank you very much indeed for the detailed bug report.

Edit: The invalid halt always seems to be at -1, 511, 0. Do you have any idea what that might be on your map?

[Dwachs]

This coordinate is outside any map I guess. So there is something wrong before.

[Bernd Gabriel]

Where can I get this error?

I've downloaded the savegame again and Simutrans Experimental 7.1 (19.DEC.2009) and my current 7.2 branch still start without errors with pak128.Britain (29.JUL.2009) and pak128.Britain.Exp (08.Nov.2009).

James,
did you try to download the savegam again? It might got broken during download.
did you use the "official" Release 7.1 and an "official" pak128.Britain or pak128.Britain.Exp?

[jamespetts]

Bernd,

thank you for the pointers. I tried loading it with an older version of Pak128.Britain-Standard, and it worked: I can only assume that Jeff was using an older version of Pak128.Britian, as a new goods type is added in later versions (of both Standard and Experimental), which, because of a bug in both Simutrans-Standard and Simutrans-Experimental, breaks saved games with a different number of goods types.

As to the original problem, I have run the game on fast forward for about an hour and a half: it is now 1817 and I have experienced no freezes. I shall keep running it until I go to bed this evening and report back any findings. For now, however, I cannot reproduce this issue.

[jeffatsqi]

@jamespetts

I am providing what I can of my environment because we are getting different results form the same game file.

The gaem engine is the Simutrans-Experimental 7.1. My specific environment is uploaded below (with the Pak set removed because of upload size limits).

http://simutrans-germany.com/files/upload/British-71.zip

And the Pak set is the Pak128.Britain-Ex 0.4. I can not upload because even zipped file it exceeds the upload limit. But the Pak was just downloaded on installed without any changes.

Hope this helps reproduce the failure.

Jefrf

[jamespetts]

Jeff,

thank you for your response. In relation to the pakset, the addition of the new goods type must have been after the release of 0.4; in any event, using the standard Pak128.Britain seems to allow me to open your saved game. The additional things that you have uploaded will not, however, affect whether the game freezes; in particular, the settings in simuconf.tab are saved with the game, so changing these after the game is created will have no effect. In particular, your changing of "max_route_steps" will not register on your existing saved game, because the original max_route_steps value will have been saved with the game.

When you run that game on your fast computer, at what game year does it first freeze, on average?

[ӔO]

I've managed to recreate freezing at start of the month using current 8.2ex with pakbrit-ex0.6
I haven't changed any of the route calculation parameters from default.

It's hit or miss with game speed set to 1, but seems to freeze quite reliably with fast forward.
This is on a Core 2 2.66ghz, 3GB RAM, Geforce 7900GS system.

http://files.[ simutrans [dot] us (site down, do not visit) ]/files/get/5xF-Jl5zjZ/test.sve

[jamespetts]

AEO,

I think that I have found what was causing the fault in your game, which may or may not be related to the original issue: when a factory was deleted, the code for checking the electricity consumption of that factory was called after code which might have deleted it, resulting in an access violation. This has now been fixed on the -devel branch. Thank you very much for your report :-)

[ӔO]

This is back again in 9.2.

After deleting some factories and tourist attractions, and only started happening after 1920ish.
only happens at the beginning of month, but seems to be quite random.

At least the game doesn't crash now and just sort of hangs for around 1~2mins before returning to normal.

http://dl.dropbox.com/u/17111233/acetate%20193x%20freeze.sve

[jamespetts]

Aha, thank you for spotting that! This was actually a completely different bug, which is now fixed on my 9.x branch. The fix will be incorporated into the next (9.3) release. Thank you for your report!