I know there's been some discussion of disappearing goods in recent versions -- I've found a reproducible version involving passengers in 11.5. (I noticed this in 11.1, but it happens in 11.5 too.)
https://dl.dropboxusercontent.com/u/61716/disappearing.rar (https://dl.dropboxusercontent.com/u/61716/disappearing.rar) (the pak folder contains a few new files from the last one, so you'll want that too)
Follow the train in the frame, on the 'Merseyrail - Chester 1' line. Note that it is carrying a number of passengers heading to Waterloo via Bromborough. (This strange routing anyway, but I think this can be put down to the fact that all lines in the area are new and have not posted any point-to-point journey times.)
When the convoy gets to Bromborough, the Waterloo passengers disemark -- but they simply disappear immediately, and cannot be found at Bromborough or any other station within walking distance. There is no record of them leaving as "unhappy" or as having departed.
Tests suggest that this appears to happen because, on arriving at Bromsborough, the passengers find that there is no route to their ultimate destination. Could this be correct?
No, there is a route from Bromborough to Waterloo, albeit not a direct one*. I'm not sure why the passengers would have planned a route via Bromborough in the first place if they had thought there was no way to reach their destination. However, all available routes are new, having just been built in the last in-game hour or so -- if that makes a difference. The fact that this bug seems to stop after the month changes and everything is recalculated makes me think that the "newness" of the route might somehow be relevant.
* Edit: the most obvious route is in fact by going back the way they came, i.e. via Chester -- I don't know whether passengers would be prohibited from doing this?
Hmm, that's an odd one. It suggests that the ultimate issue is the path explorer not being refreshed at the right times such as to detect new routes; or it might simply be that, because your network is so enormous, the performance-saving features of the path explorer dictate that the refreshing of some routes should be held back for a while (I know that when Knightly wrote the feature, he included something that did something like that, but I do not know the details).
The problem is at least not a fundamental one.
Thanks James -- I agree that the issue is not fundamental, especially since it irons itself out eventually.
Here's quite a striking instance of a similar phenomenon to that described above -- i.e. bizarre passenger routing on opening of a new route.
https://dl.dropboxusercontent.com/u/61716/Hadfield.rar
In this save, the first train on the Manchester-Hadfield branch has just left the depot. Fastforward and you'll notice that passengers at Piccadilly immediately start to route via Hadfield. In fact, all passengers headed to the south east of England will, bizarrely route via Hadfield. But Hadfield is on a branch and does not lead to anywhere else. In order for the passengers to think decide that the best way to (e.g.) Basildon was to go via Hadfield, they would presumably have to think that the Hadfield trip would take "negative minutes" -- i.e. that it would gain them time over just waiting at Piccadilly for a London train -- since any routing via Hadfield would have to involve coming back to Piccadilly, presumably.
This suggests perhaps a severe anomaly in the "default" journey times, i.e. those that kick in prior to any point-to-point times being registered. One would expect some anomalies given the method of estimation, but quite on this level, I think.
Quote from: jamespetts on August 15, 2013, 08:41:20 PM
it might simply be that, because your network is so enormous, the performance-saving features of the path explorer dictate that the refreshing of some routes should be held back for a while (I know that when Knightly wrote the feature, he included something that did something like that, but I do not know the details).
I don't know if the path explorer has undergone major overhaul since I first wrote it. My original design was to recalculate all new routes first, then applying the new routes to all existing and future ware packets. But before recalculation of all new routes are completed, the old set of routes will still be in force.
Thank you both for your input - that is most helpful. I shall have to investigate this further when I have a moment (I have been somewhat busy this week with a tea party that I had yesterday, so there is something of a backlog of Simutrans related matters to deal with).
Morelook/Knightly - welcome back! I thought that it might have been you from your previous post in relation to the "scaled_pax_demand". I have not undertaken, nor has anyone else, any major overhaul of the path explorer (I cannot imagine that I should be able to re-write it as well as it was written in the first place), although some aspects of it have been altered to the extent necessary to accommodate new features, such as counting the walking time to/from a stop, and allowing walking between stops, as well as the point-to-point journey times, so the description that you give in your previous post should still hold true for the current system. (One other peripheral change that I have made is to increase the frequency with which the periodic recalculations of routes occur - but the same method for that recalculation is used as before, subject only to the changes needed to accommodate the features described above).
Carl - your reference to this post in the other post reminded me of this, and I realised that I forgot to look into this. Can you reproduce this in 11.27? I note that the problem is not immediately apparent on loading a saved game. Is it the case that, on loading a saved game, after a particular period of time, the problem will eventually recur in a particular place? What would be extremely helpful is a saved game coupled with when and where the problem manifests itself - without that, I will just be running it for an indefinite period hoping to find it by chance, which I am afraid is not a feasible way for me to find and fix bugs.
Hi James,
This is one of those very rare bugs whose cause is entirely opaque and so I can give no advice on how to reproduce it. The only solid correlation is that it occurs only with new or recent additions to the network. I can't say for certain when I last saw it, but it was substantially later than 11.5. If and when it comes up again, I'll post another savegame here.
Thank you - that is helpful.