Signal light disappear since revision #2322

[BB25187]

As explained in a topic of the Elvas Tower forum, we noticed that the light of some signal disappear when one get closer from them. After some back-tracking, it seems that this happens since revision #2322 (Experimental signal light glows). The issue happens when the experimental option is not set. It doesn't affect all signals.
 To reproduce, you may want to (sorry, I couldn't find a simpler setup :-( ):
- Load this route at http://www.activitysimulatorworld.net/chargements_temporaires/Lignes/ASW_Chargements_Ligne_PLM_V2.html. You should install the three files of the version 2 of the route using 7-Zip (PLM_V2.7z.001 to 003), and the version 3 (PLM V3_07.exe
 - Use the path attached to the topic pointed above to create an activity. The first signal you will meet exposes the problem.

I attache a possible patch, which restores the z-bias that was applied until rev #2321 to Out.Position.z. This z-bias was removed at rev #2322, and replaced by non proportional offset, which doesn't work when distance is small. The behavior when glow option is enabled isn't modified by this patch.

[James Ross (twpol)]

It is worth noting that the old positioning would cause the light to incorrectly cut through bits of the signal shape, so I'm not sure we should just go straight back to it.

[BB25187 (bb25187)]

My experience (with french, german, austrian and swiss) signals is that most of the troubles with old positioning are caused by the (much) larger light size in OR than in MSTS. Reducing the size of the light is often sufficient to avoid nasty interferences between signal shape and lights. This may be different with signals from other countries though. From this stand point, revision #2322 didn't improve things significantly. But clearly the only purpose of this patch was to avoid a regression when the new glow option isn't set.

[James Ross (twpol)]

I am not aware of this "(much) larger light size in OR than in MSTS" and if that is correct you better file a bug with details, as it should be fixed.

[BB25187 (bb25187)]

OK James. Sorry for not having reported it already :-(
I will prepare a small setup to expose this in the coming days.

[BB25187 (bb25187)]

I think my statement on signal light size was wrong, and things are more subtil than I first thought.
I compared the rendering in the two simulators, and in fact the size looks nearly similar. What I interpreted as a larger size was in fact linked to the point you raised on the parts of the signal incorrectly hidden by the lights. This seems to be highly dépendent on the distance and angle of view.
I confirm that from this standpoint, the revision #2322 represents a significant improvement.
As for the light which disappears, I could also resolve it by increasing the light offset to 2.5 cm (instead of 1). This preserves the improvement on light/shape ordering. Nevertheless, this remains somewhat empiric and may not work for other signals (for instance, the wrong ordering reappears when the offset is larger than 4 cm).
Just in case, I attach the tricial patch which corresponds to this last attempt.

[James Ross (twpol)]

I've had a look in MSTS and as best I can work out, it puts the signal light 1.5cm in front of the location specified in the file (at least when the camera is close - it's hard to know what it does further away).

[James Ross (twpol)]

Based on the MSTS attachment, I've changed Open Rails to use 1.5cm in X2560. This does appear to resolve the problem for me on the PLM route mentioned above.