The following screenshot shows small rail triangles that can contain multiple signals in the middle. The versions in the first row contain only straight rails, the lower versions have a curve at the top of the left diagonal. The columns show the different angles.
In 8 of 9 cases, the signals can be placed on all three lanes wihtin the inner triangle. In one version (column 1 row 2), however, this is not possible, as the signal cannot be placed on the right diagonal within the red circle. This is probably because of the curve, because in the same version without a curve (column 1 row 1) it is possible, even though the inner triangle is visually the same. Further, it is also possible in the even smaller triangle (column 1 row 3).
Hence, the position should be available for (column 1 row 2) as well.
This problem is the same for the mirrored version, or when both diagonals have curves at the top.
[2.0.28] Missing position for signal
-
- Manual Inserter
- Posts: 2
- Joined: Mon Jan 06, 2025 6:04 pm
- Contact:
Re: [2.0.28] Missing position for signal
Welcome to the forum, and thank you for the report.
A view with show-rail-segment-collision-boxes debug setting: That curved rail overlaps the diagonal rail.
Works correctly.
Duplicate of 124952 Rail signal fails to separate section.
A view with show-rail-segment-collision-boxes debug setting: That curved rail overlaps the diagonal rail.
Works correctly.
Duplicate of 124952 Rail signal fails to separate section.
-
- Manual Inserter
- Posts: 2
- Joined: Mon Jan 06, 2025 6:04 pm
- Contact:
Re: [2.0.28] Missing position for signal
Overlapping collision boxes doesn't to be an exclusion criterion/the criterion is not consistent. Below, in row 1 is the "broken" version, below is the wider variation, once without (row 2) and twice with signals (row 3 and 4). The curve completely overlaps with the highest segment and partly overlaps with the middle segment. Still, signals are possible.
The criterion seems to neither be whether the connection point of two rails lies within another bounding box. This could be infered from my examples but would not explain the problem in 124952.
For issue 124952 one can argue that the crossings are too near to each other and thus the position is intentionally not enabled. Here, however, there is a lot of space and no visual difference to the straight version exists. Thus, I would consider this a special case not tested with the new rails, and ask to not mark this as "Duplicate" or "Not a bug"
The criterion seems to neither be whether the connection point of two rails lies within another bounding box. This could be infered from my examples but would not explain the problem in 124952.
For issue 124952 one can argue that the crossings are too near to each other and thus the position is intentionally not enabled. Here, however, there is a lot of space and no visual difference to the straight version exists. Thus, I would consider this a special case not tested with the new rails, and ask to not mark this as "Duplicate" or "Not a bug"
Re: [2.0.28] Missing position for signal
Note that show-collision-rectangles is different from show-rail-segment-collision-boxes.