[0.12.2] Train sometimes fails to take longer, but open path

[linux-user]

I might not understand how chain signals work, but it seems that my trains are sometimes choosing to take a shorter, but obstructed route, rather than a longer, but unobstructed route. I would expect the train to always prefer an unobstructed route over an obstructed one.

This is a portion of my rail network: http://i.imgur.com/qnhLV4p.jpg. You can see the whole network in the minimap. The north/south rails are two-way rails, and a small portion of the southernmost east/west rail is also two-way.

A small train is coming from the north. Its destination is to the south, and has been called out on the minimap. Once activated, the train will turn onto the diagonal rail and wait -forever- at the chain signal.

This network is not broken: when the Blocking Car is not present on the shorter path, the train takes the shorter path for both Northbound and Southbound travel.

When the Blocking Car is present on the shorter path, the train takes the longer path for Northbound travel -as expected-, but tries to take the shorter path for Southbound travel, and ends up sitting forever.

Does my explanation make sense? Am I misunderstanding something? I can provide a savegame and/or more information if required.

[sillyfly]

You have an uncoupled signal on the longer route, making it basically a one-way in the wrong direction, so the train has no choice.

[linux-user]

You have an uncoupled signal on the longer route, making it basically a one-way in the wrong direction, so the train has no choice.

Yes, that was exactly it! Pairing a chain signal with the block signal on the turn on the longer path now makes the train behave as expected. Thanks! So, this is not a pathing bug.

So, I guess this is a UI issue? Maybe a chain signal that controls the wrong way on a one-way section should display the red signal, or (maybe) display the test pattern?

[Vin]

The problem with that is it would break all kinds of things that chain signals are useful for. The singleton signal could be closing off an intersection and you'd want the chain signal to work.

As far as I'm concerned, this is working perfectly. The chain is green because the block it controls is empty, and there's no signal for it to chain to.

[linux-user]

The problem with that is it would break all kinds of things that chain signals are useful for. The singleton signal could be closing off an intersection and you'd want the chain signal to work.

I agree that chain signals should work across intersections. Consider this scenario, which is analogous to the one in my original report: http://i.imgur.com/5W0NUDr.jpg. Each train is attempting to get to the station directly in front of it. Notice that neither train has a valid path to its destination station, but each chain signal is lit green.

I argue that the chain signal should either be red or display the test pattern in this case. Can you show me a valid, useful rail intersection that this behavior change would break?

[ssilk]

One of those cases, where the game can detect obviously wrong signal setting, but not a bug.