[0.12.1] train signal symmetric inconsistency

[LotA]

Was playing with train and stuff then i discovered this :

fac.jpg (268.33 KiB) Viewed 22774 times

with this kind of junction, train signals works only on the right side.
signals are placed just afer the curve, first allowed position.

using either signal or chain signal doesn't change anything, i believe its something about the block detection

I believe this is a bug, it should either work or not, but no depending on the axis.
(And i believe it should work in both case as the train signal is placed just after the curved line, so technically on a straight rail)

[rampelstinskin]

Another strange case.

[MF-]

I don't understand what is happening nor what should be happening there.

[LotA]

fac2.jpg (246.64 KiB) Viewed 22695 times

Interesting news :

First of all, unlike i said, the signal is not placed "just after" the curve, but on it (although it looks quite straight and the showed hitbox stops just before the signals)
Second, if you reverse the "axis" (diagonal being \ instead of / ), it will fail in both cases (see below), its only working with this "symetry"
The problem is definitely related to crossing with diagonal tracks.

Signal will work / faill to recognize blocks depending on which one you placed first.
In the picture :
if you place the "top" signal first -> signal will work
if you place the "bottom" signal first -> signal won't work

So after all i don't know if should work or not but yet it should definitely behave the same way whatever is your "axis"

[MF-]

So this is all about signals binding to / recognizing track?

Pehaps a picture with those symmetric layouts, a clear "---> mirror image ---->" label
and then captions on signals "works if 1st", "works if 2nd", "always broken", "always broken"?

Sure the devs can read the signal much better than me. But sill a good report warrants better chances I think.


EDIT: That 1st / 2nd thing is explained in your last picture, nice and clean.
Does it mean that a tiny rail segment was created between the two signals in the failing case?

[LotA]

straight rail doens't collide with curved so there is every pieces of straight rail possible in the diagonal, althought afaik only the last, overlapping one is involved. All rails were placed before any signals.

new tests set : in fact you can make signal either work / fail with both symmetry.
it can happen with any rotation of a curved rail with the diagonal end

fac.jpg (202.25 KiB) Viewed 22610 times

watching how it works with h/v end leads me to think is should simply fail in both cases.

Oh and sorry for the missing m in the title :d Only one in my native language

[LotA]

if you place the straight piece after the signals, only 1 signal gets "disconnected"

fac.jpg (93.47 KiB) Viewed 22602 times

[HanziQ]

Thank you for the report, this will be fixed in the next patch, but for newly placed rails/signals only, you need to replace the existing problematic ones.

[mooop12]

I believe fixing this introduced a new bug, in the image below chain signal in red rectangle is disconnected and chain signal in green rectangle is connecting properly, but before 0.12.5 both positions were correct. The chain signal works fine on previously built junctions until I try to replace it.

Screenshot_1.png (855.15 KiB) Viewed 22447 times

[HanziQ]

Unfortunately, this bug would require serious refactoring of the way that signals are currently attached to rails, therefore we have decided to postpone that. Moving to 'known issues' for now.

[kovarex]

Similar topic for reference: https://forums.factorio.com/forum/vie ... =7&t=17757

[kovarex]

All these problems were solved as by product of the rail signal placement indicator for 0.13 (https://www.factorio.com/blog/post/fff-134)