Graphical glitches with new signals

[Vladki]

hi, i'm trying the latest devel version of simutrans experimental, and I have noticed that the semaphore pre-signals change their aspect to caution when the map is rotated, or in other situation they stau at clear even if the main signal is at danger. However the aspect changes on rotation. Saves are here:

https://uran.webstep.net/~vladki/simutrans/glitch1.sve
https://uran.webstep.net/~vladki/simutrans/glitch2.sve

Edit:

I have prepared circuit-block signals for pak.sweden, but either I do not understand exactly how they should work, or there is something broken. It seems that these glithes are not pakset specific.

Savegeme: https://uran.webstep.net/~vladki/simutrans/test.sve
Pakset: https://uran.webstep.net/~vladki/simutrans/pak128-sweden.zip
Just let it run and watch signals and reservations.

- Pre signals sometimes stay at clear even when main signal is at danger. (Is there any difference if I put both is_singal and is_presignal in dat file, or should it be only is_presignal ?)
- The reservations on the double track with 3-state signals are really weird. Usually whole track is reserved, throught the junction and through the choose signal, many blocks ahead.
- The choose signal for the train with clay does not choose alternate platform (maybe it is too far?)
- Sometimes you can see signal at clear with no reservation going to danger on trains apporach
- Or you can see a signal at danger with reserved block behind it.
- Different rotaitions seem to have effect.
- When the double track line is empty, sometimes the signals are all clear, sometime all at danger.
- I tried to use cab-signalling and it was even worse.

[jamespetts]

Thank you for your various reports. I have moved this from the pakset board because it clearly relates to the code. I should be grateful if you could in future post one thread per bug, as that does make it very much easier for me to keep track of these things (as I know that all the discussion in any given thread relates to a specific bug; it is easy to forget about one of several bugs if they are reported all in the same thread).

I do like the progress on the Swedish pakset, incidentally: this is looking splendid so far.

Can you upload the .dat files for the signals that you are using in this pakset so that I can see whether they are set up correctly? The ones at the Github repository seem to be out of date and do not contain, e.g., the "Modern_ThreeStateSignals_right" that I can see in the saved game.

To answer your question, as in Standard, is_signal should not be defined where is_presignal is defined (Edit: except for a combined signal). A train does not stop at any signal with is_presignal defined (Edit: except for a combined signal), as this does nothing other than alert the driver to the state of the next signal.

For the various issues that you describe, can you identify where and when they occur in the saved game so that I do not have to spend a long time searching for instances of them? Can you also describe in detail how cab signalling was worse?

Thank you again for your feedback - it is most helpful.

Edit: I have now fixed the rotation issue in "glitch1".

Edit 2: I have now also fixed the issue identified in "glitch2".

[Vladki]

I have commited the new signals to GIT - see the signals.new file. Whole pakset is uploaded. I have also modified the signals to show lights on the backside. It's ugly but needed for testing. I'm about to check again in game, and give some more details.

I have made a set of screenshots to document. Some things may be effects of the same bug.
https://uran.webstep.net/~vladki/simutrans/experimental-signal-screenshots/

00, 04: 3-state signal showing clear (green+white) with next signal showing danger (red)
02: presignal showing clear (white), with following choose signal on danger (red)

01: reservation through choose signal (entry to York station) is too early. It blocks the junction and platform for too long time, while it could be used by another train
03: similar to 01, the train from west reserves all way to York, blocking the junction with track to coal mine. IMHO the reservation should stop at the junction, unless the train is closer (aspects-1 blocks ahead)

05: again similar long reservation, this time the train has reserved path to alternate platform, with some blocks not displayed as reserved. Chosse signal shows clear for alternate path.
06: a few seconds after 05, blocks are properly reserved, but the choose signal shows clear for main route.
05+06+07: The train is scheduled to reverse at the station after the choose signal. Singals at the end of platform are correctly at danger, but the choose signal shows clear (with or without choose) instead of caution. The only place where it shows caution (and correct aspects for main/alternate route) is the station at coal mine with dead end tracks. (03)
07: again the train goes to alternate platform, but the signal shows clear for main route. This was the choose signal that was too far (that is now fixed).
08: Train from east is taking an alternative route through the station (not stopping there), still it gets clear for main route - should be clear for alt. route. IMHO.
09: train goes to scheduled platform, and reversing. There's end of signalling sing at the end of platfor. Signal shows clear instead of caution.
10: train goes to alt platform, choose signal shows clear for alt.route, but after a while changes to clear for main. (11). Should show caution for alt. route IMHO. Perhaps same as 05+06

[jamespetts]

Thank you very much for these: the signal .dat files that I have looked at so far seem correct for the 3 aspect track circuit block signal and the track circuit block repeater (pre-signal).

I have so far pushed fixes for issues 00, 01 and 02 (and the duplicates thereof). As to 06, real life practice has varied here: sometimes, a clear aspect was shown when approaching a terminus station, at other times a caution aspect. I decided to go with the older practice of showing a clear aspect, as this was easier to code. This is the same issue as 09 and 11.

I have yet to look into 08.

In the meantime, however, can you test and confirm that the issues have been fixed?

[Vladki]

It seems that issues 00, 01, 02 are fixed. Reservations are not so long any more. However there are some left.

Issue 08 seems to behave differently now. The train that has to take alt. route has danger on choose signal, all the time.
Have a look at 10+11. Even if I take it that it should show clear, there is still an issue that at 10 it shows correctly clear to alt. route, but after a while it changes to clear to main rote. It should not do that.

And there's one more issue. Similar to 02, but it happens on the passing loop shown at 08. The presignal shows clear, choose signal show danger. The difference is that there is an end-of-choose in front of choose signal.

[jamespetts]

Thank you for that. I am having some trouble reproducing the conditions for no. 008 just by letting test.sve run; at what time after loading does this occur?

Edit: Never mind: I think that I have found and fixed 008 now; could you re-test? I should be most grateful.

Edit 2: And I can't reproduce the issue that you describe with the choose signal being at danger when the pre-signal is at clear, I'm afraid.

Edit 3: I also think that my latest commit has fixed situations in which a choose signal correctly shows clear to a subsidiary route but then changes incorrectly to show clear to the main route; can you test that, too? I should be very grateful.