9.2: severe bug with waiting times

[Carl]

Hi all,

I finally decided to switch to 9.2 -- but I've had some very strange experiences with a large map created in 8.2. It ran fine for a month or two, but now almost all of the waiting times at stations have increased to implausible amounts (e.g. Peterborough, 50 mins. travelling, 2479 mins waiting). Obviously this causes chaos with the network. The problem seems only to be getting worse with each passing game-month.

I've made some in-game changes to the Settings menu -- reducing the passenger-factor and increasing the distance_per_tile -- but I don't see how that could have caused the problem.

Any idea what could be causing this trouble? I can upload a save if that would help.

Many thanks!

[jamespetts]

Jha4ceb,

thank you for your report. Firstly, can I check - are you running the 32-bit Linux binary with 64-bit Linux? If so, that is quite likely a cause of your problem (the 64-bit binary works now).

If not, then can you upload your saved game (preferably the game as saved by 8.2 if you have it as well as the new one from 9.2) so that I can investigate? Thank you for your help, and sorry that you have had problems.

[Carl]

I'm running the Windows binary on a 64-bit system.

Here's the offending file:
http://simutrans-germany.com/files/upload/UK3-9.sve

And here's a version created and saved in 8.2.
http://simutrans-germany.com/files/upload/UK3-aargh.sve

Here's my pak folder for add-ons, etc:
http://simutrans-germany.com/files/upload/jha4ceb-pak.rar

The 9.2 version is a couple of months along from the 8.2 version -- that's how long it took for the problem to appear.

I wonder if this could have something to do with the new feature to regulate high waiting times?

Many thanks!

[steffen]

I've had the same problem. I fixed it by updating from the 30dec version of the Linux-32bit build, to the 4jan version of the Linux-64bit build. Since 9.2 was released a few days before that it might be worth trying a nightly rather than the released 9.2

Edit: Well, James fixed it, and told me what I needed to do, I merely downloaded the file

[Carl]

Hi Steffen,

I saw your report in the Solved Bug Reports folder. I'm running a Windows build, not a linux build, albeit on a 64-bit system -- so maybe this is an indication that there is a similar problem with the Windows binaries on 64-bit systems too?

[jamespetts]

It's odd - I am told reliably that 64-bit issues like this do not occur on Windows, but, once I am able to get Git working on my new computer (which is 64-bit), I shall have a go at compiling a 64-bit Windows binary to see whether these problems are solved thereby.

[Carl]

Thanks -- I look forward to hearing the result!

[jamespetts]

Jha4ceb,

I have tried to reproduce this with my new 64-bit Windows 7 machine, but failed. I ran your original UK3-aargh.sve file (the original that you had saved in for many game months on fast forward, and could not find any places with a waiting time that was outside sane boundaries; indeed, there were many instances of the (new) minimum waiting time of 4 minutes.

I am wondering whether the problem might be a bug somewhere in the settings dialogue. Have you ever had any problems *without* changing the settings?

Edit: I also tried it with your uk3-9.sve file, and could likewise not reproduce the problem. Most places started with an unknown waiting time, and eventually registered the minimum waiting time of 4 minutes. However, on inspecting the settings from the "i" menu, there were a number of values that appeared perverse or out of range, which suggests a bug with the settings dialogue. For the time being, it may be better to avoid using the settings dialogue until the bugs are fixed. Thank you for your report!

[Carl]

Hi jamespetts,

Thanks so much for checking this out.

The reason I wanted to change the settings actually relates to another minor quirk of the 9.x updates: it seems that loading my 8.x saves in 9.x leads to a drastic increase in the number of passengers. As such I needed to reduce the passenger_factor. Any idea what might be causing this latter phenomenon?

[jamespetts]

Hmm, that's odd: using your UK3-aargh.sve file and running it for several months in 9.x, the number of passengers generated as shown in the city list chart remains constant. Was it on your UK3-aargh.sve that you experienced the increase?

[jamespetts]

Can I check - do you still have this problem?

[Carl]

I'm just starting a new game in 9.2, so I'll let you know about the waiting times issue.

However, having recently loaded an 8.2-created map in 9.2, I can confirm I still get the increase in passengers. If I remember correctly, volume increases by around 25%.

[jamespetts]

Ahh - 9.3 has just been released. Can you re-check with 9.3?

[Carl]

The original bug -- concerning exponential rises in waiting times after entering the 'settings' menu -- appears not to occur in 9.3.

However, loading an 8.2-created game in 9.3 still appears to lead to a change in passenger volume -- this time down 20% month-on-month (where the passenger volume on this large map is usually very steady).

[jamespetts]

Jha4ceb,

that might be due to the revised passenger generation algorithm in 9.3, which fixed a bug in earlier versions in which too few passengers used private cars. Can you upload a saved game?

[Carl]

I meant to say that there was also a huge increase in the "No Route" category -- by a factor of 7 or 8, from very low numbers -- at all stations. Before I upload a saved game, could this explain the issue?

[jamespetts]

The passenger generation alone oughtn't alter the number of passengers recording "no route". I haven't seen any increase in "no route" indications in any games that I've seen before, so it would indeed be helpful if you could upload.

[sdog]

... the revised passenger generation algorithm in 9.3, which fixed a bug in earlier versions in which too few passengers used private cars.

haven't mentioned, i think you might like to hear it. Ratio of pax transported dropped for me from 36% to 17% after implementation of this. This is with the massive game, with all cities connected in mid 1960s. Car ownership should be 50%ish, you certainly remember car ownership figures better.

[jamespetts]

Very interesting! Has this affected your strategy at all: have you, for example, felt inclined to close any minor railway lines and replace them with 'bus routes? I should be interested to see the latest developments in this map in saved game form...

[sdog]

I reacted by improving the local bus lines in cities. A new fast link north of the main line.
Less pax on the main line were also welcome, less overcrowding.
Also had to replace all DC catenary with AC catenary, AC trains are faster so i expect more pax in the south.

The car users are shown at the station display as "no route"? At first i though the number of hops was too low, and tested this by introducing point to point trains connecting some larger towns with each other.

All measures together i could increase ridership by 2% in two years. (Had hardly any time to play.)

[Carl]

Apologies for not getting round to uploading a relevant game yet -- it's been a busy week. Will do so ASAP.

[jamespetts]

SDog,

that sounds very interesting - I should be interested in seeing a saved game! Private car passengers are not shown at all at stations, as they don't use stations. However, if you look at the passenger destinations map, you can see which destinations that passengers from any given town reach by their private cars rather than by your transport.

[Carl]

Below is a link to a recent file, with my most recent pak-folder, plus some explanation.

http://simutrans-germany.com/files/upload/LARGE_UK_9-3.sve
http://simutrans-germany.com/files/upload/jha4ceb-pak.rar

Having played around with 9.3 some more I've isolated some of the pockets of oddness that have occurred.

>> Stops which rely wholly on Tourist Attractions for their traffic, rather than houses/shops/industry, have almost stopped generating passengers altogether -- although, strangely, they still *receive* a large volume of passengers. At such stations, thousands of 'no route' passengers are displayed. You can see this most sharply at Heathrow, where I've simulated the passenger traffic by adding attractions (since I don't plan to have any planes on this map but wanted the lines to nevertheless be operational). Nothing at all departs from Terminal 5 station anymore, though very small volumes still depart from Terminals 4 and 1,2,3. But since 1,2,3 and 5 are both served by the Paddington train, it's odd that there should be this discrepancy. So something weird seems to be going on with tourist attractions in 9.3.

>> Many stations have a large asymmetry in their arrived/departed numbers -- for many towns on the outskirts of London, their 'arrived' numbers are far, far lower than their 'departed' numbers (though note that this is not always the case). This is possibly related to the first problem...

>> Although I have entered the 'settings' menu since loading this map in 9.3, these problems were evident before I did so, and therefore appear to be independent.

I'm aware that the project I'm engaging in here (building a map from scratch) is in many ways non-standard and presumably not exactly what Simutrans was designed for -- so perhaps some of the oddness is related to that. But as far as I can tell, the strange behaviour of Attractions doesn't *seem* to be traceable back to any of the odd things I'm doing.

[Carl]

One more thing to note: the "arrived/departed" discrepancy also sometimes goes the other way, i.e. far more departed than arrived. This appears to be the case at Swanley, for example.

[jamespetts]

Ahh, thank you for that report. That was due to a bug in the new passenger generation code resulting in all return journeys being made by private car. That will be fixed in the next version.

[Carl]

Ah, I see! I'm relieved that the explanation is so simple rather than as a result of some complex thing I was doing wrong. So does that bug account for the large amounts of 'No Route' passengers at such stations?

[jamespetts]

Yes, it's possible that this will account for "no route" indications in cases where there is no suitable private car route back to the origin.

[jamespetts]

The remaining issue identified here should be fixed in the now released 9.4.

[Carl]

It is indeed -- many thanks!