2.9 Alpha1 - Avoid crossing perimeters - better, but not fully fixed

[marsmanp]

Description of the bug

The same as seen here: #13445

The Avoid crossing perimeters being enabled does not always force travels to be within the object even if it is possible to do so. In the attached file you can get great/good/poor/bad results depending on other settings like number of perimeters, seam location, and infill type. There may be other settings that help/hurt the behavior, but these are the ones I know of

Project file & How to reproduce

ACP.zip
A few examples of different settings for perimeters/infill causing changes in travels crossing/not crossing perimeters.

Checklist of files included above

• Project file
• Screenshot

Version of PrusaSlicer

2.9.0 alpha1

Operating system

Windows 10

Printer model

Anycubic MegaPro

[SachCZ]

Hello @marsmanp, thank you for the report. We know about this and it will be fixed in the next iteration.

[Jan-Soustruznik]

SPE-2578

[dattas]

Just want to verify things since the other tickets got closed incorrectly as the issue persists. This is still a problem and is causing problems with my Prusa XL causing stringing by crossing the perimeters. Is this still being looked at improving further? I know it isn't just me as others are also indicating there is still an issue: #12432 (comment)
Attached is a few examples using built in primitives. The first two have wipe tower, which does quite a bit and the last one does not have wipe tower and still is not honoring the avoid crossing option

crossing-perimeters.zip

[pmarks-net]

Here is a test case where "Avoid crossing perimeters" fails: https://www.thingiverse.com/thing:6948297

PrusaSlicer 2.9.0:

Cura 5.9.0 for comparison:

[SachCZ]

Hello, thank you for your feedback. I just checked and the original issue is fixed. When it comes to @pmarks-net example, I am not aware of PrusaSlicer ever working like cura in this case. It is a difference in the algorithm used. You can open a feature request if this is something you want us to implement.

The @dattas case is similar in that I do not think this is a regression. When moving to other objects (wipe tower including) the move is usually direct.
There is no way around it, for example, imagine you are going from object infill to a wipe tower. Because the perimeter is a closed loop you have to cross it one way or another. In that case it is better to just go directly.
The same holds for the multi material islands. Each has its own perimeters, meaning if you go from within the island elsewhere, you will have to cross it. Is is better to just go directly.

The direct move mentioned tries to avoid "holes" in the islands (islands are single material, bounded by perimeters), but in the example @dattas provided, there are no holes, because of the multi material setup, so nothing to really avoid when going from object to object.

If you still believe there is a regression, please provide a PrusaSlicer version where this works as expected and version where it is broken.

If this is not a bug but something you would want to work, please open a feature request where you describe in detail how it should work.

Thank you for your time and effort.