Station T-signals in absolute block

[Vladki]

Hello,

in pak sweden there are absolute block station signals (T-signals), even a double block station signal.
While the simple station signals seem to work (with minor glitches), the double signal is not working at all.
The question is was that intended at all? I thought they were intended only for time interval (with or without telegraph).

They are coded as:

Code Select Expand

obj=roadsign
name=Se_MSU_1900_T-sem
waytype=track
aspects=3
working_method=absolute_block
has_selective_choose=1
has_call_on=0
permissive=0
is_signal=1
#free_route=1
is_longblocksignal=1
double_block=0/1 # depends on if it is double block or not


[jamespetts]

I do have a recollection of it being intended that these should work, but I can no longer recall the details.

Can you upload a saved game in which a specific problem with these can be reproduced at a particular time and place?

[Vladki]

Looking at the documentation it says only about station signals for time interval methods, not absolute block. Also aspects=3 does not really make much sense in absolute block. And indeed I can sometimes see caution on those signals...

However the problem of double-block station signal was solved by moving it to a tile with platform. Further glitches remain common with single-block station signals. I will report later.

[Vladki]

So here are the issues. They can be seen on the swedish server game.
http://list.extended.simutrans.org/~vladki/debug-saves/t-signals.sve

A: at loading the game - T-signal at Bockara is at Caution (East) - well it is an absolute block signal, how can it possibly show caution? It will remain at caution (east) until the train reverses and departs. I think it should turn to danger when the train stops, and show clear (west) for departure. (For arrival either danger, or caution east, makes sense).

B: at 3:04 another train arives to Tun. There is another train in Fageras scheduled for Tun as well, and it will depart immediately, when train in Tun stops. There is only the T-signal at Tun station (no home/choose signals). I think the train in rear should wait until the first train departs. Otherwise there is a potential for deadlock. Interestingly the T-signal at Fageras is at danger all the time when train is departing.

C: at 3:20 two trains approach Klovertrask almost simultaneously, they both slow down to 1 km/h, crawl for 2 tiles, and then speed-up again before finally stopping and loading. The T-signal does funny dances, being clear-east for most of the time, then shortly jumping to clear-west, back to clear-east, and finally to danger. Train departing westwards, departs with signal at danger.

D: about the same time as both trains depart from Klovertrask, th 3rd train arrives to Tun. It stays there waiting for clearance, but the T-singal show clear-north (3rd train is heading north). At about 3:44 train from Klovertrask arrives, and slows down to 10 km/h, signal then changes to clear-south, train accelerates, and the other departs, turning the signal to danger. Again train to the south departs at danger.

I have tried changing the dat file to aspects=2, but it did not help, just the signal disappeared competely at some aspects. What helped was to put home (choose) signals in front of the station. This reduced the chance that there would be simultaneous reservations from both directions up to the T-signal which seem to be the cause of case (C).
I have also tried the same layout with time interval with telegraph T-signals, and it worked much better, but the trains still slowed down to 10 km/h like in case D. Maybe I was just not lucky enough to spot the exact situation as C with time interval. I also suspected that direction (W-E), is the problem but it behaves the same after rotation.

Maybe Ves should comment on how the t-signal is supposed to work, and perhaps accept some changes from real world (which we are not sure how it really worked). Just to make these signals more intuitive and predictable.
I think the T-signal should really signal only departures, and not arrivals. That should be signaled by home signal (choose signal, plate signal).

[Ves]

I'm not in position to check the servergame how it looks, but if you refer to the small line I built, I was experimenting with two block signals, to see if they would clear the signal protecting the following station.

We already had the big signal discussion some time ago. As I said then, the T-signal was used to flag the entry to the station, rather than the exit of the station.
Of corse I will accept that things maybe not can be completely as in real life, but as close as possible and feasible (and gameplay understandable) is desired.

I'm not in a position currently where I can reliably bughunt, unfortunately.

One thing is clear, I think:
T-signals of both absolute block and the time intervals methods should work the same in regards to when/if the signal is showing clear on entry.

I remember that James said that for the double block signals to be used, he reused some flags, among them the caution flags.

[Vladki]

Yes it is the single line on server, but I have changed the signals (see the above linked save).

I thought that station signals were originally implemented by James to signal departures. And that entry signalling was added later.

If T-signal was only for entry, what was then used to signal departure?

[Ves]

If T-signal was only for entry, what was then used to signal departure?

I am not entirely sure about the exact procedure, but there would definitely be a flagman signaling the departure, and maybe the T-signal was set to clear too. However, one can easily think of the risk that an approaching train thinks that the T-signal, which is set clear for departure, is meant for the approaching train, leading to dangerous situations. That conflict is what makes me doubt that the T-signal was used for departure at all.

[jamespetts]

I think that I have fixed problem A: I should be grateful if you could re-test with the next nightly build. Can I check whether the others can still be reproduced? It is much easier to deal with bug reports when they are limited to one per thread.

[Vladki]

Issues B,C,D are still there.
Issue A has changed a little - it is not showing caution anymore, but clear. Otherwise it is still bad - train arriving with clear (east), but the signal graphics is like danger (in all rotations). The signal stays so until the train departs.
Similar situation is on the other end of line, info box shows that the signal is clear (south) all the time, but when the train reverses, the visible aspect changes from danger to clear (south).