Cura: (4.6.0) Z Seam Starting in middle of a wall, even with Sharpest Corner and Hide Seam

[googlematt]

Description
Either Z Seam alignment or Corner Preference does not seem to behave correctly. With these settings on Sharpest Corner and Hide Seam, the seam ends up in the middle of a wall.

Application version
4.6.0 (worked correctly in 4.5.0)

Platform
Win 10, 5700 XT GPU

Printer
Custom Printer (Tevo Tornado)

Reproduction steps

1. Open Cura 4.5.0 and 4.6.0 (perform all subsequent steps in both)
2. Import a Benchy
3. Choose the Draft Profile
4. Expose the Z Seam alignment and Seam Corner Preference settings
5. Ensure they are on "Sharpest Corner" and "Hide Seam", respectively
6. Slice
7. Move down to a layer in the middle of the boat hull (I used 108 for my examples)
8. Pick a wall on the boat hull and note where the seam starts in both versions

Screenshot(s)
(Image showing the problem, perhaps before/after images.)

Actual results
Seam in 4.5 starts on a corner. Seam in 4.6 starts in the middle of a wall.

Expected results
Both behave like 4.5.

Project file
TT_3DBenchy_ProjectFiles.zip

Log file
CuraLogs.zip

Additional information
If you enable coasting, it becomes really apparent in the preview where all the seams are

[googlematt]

I'm glad I started with a Benchy. Would have sucked to discover this on a bigger print

[smartavionics]

I think the different behaviour could be due to the change in how the polygon simplification is done. However, IMHO, using sharpest corner is pretty much a dead loss anyway, it very rarely does what you want. Far better, IMHO, is to use a user-specified z-seam hint.

[googlematt]

I think the different behaviour could be due to the change in how the polygon simplification is done. However, IMHO, using sharpest corner is pretty much a dead loss anyway, it very rarely does what you want. Far better, IMHO, is to use a user-specified z-seam hint.

I've never actually used the User Specified setting. However, I've found that sharpest corner has been rather successful so far.

Either way, it should still work as intended, which it doesn't appear to be. I only used the one layer as an example, but it's the same across almost every layer. In 4.5.0 it starts on a corner for each shell, but 4.6.0 starts in the middle of a wall.

[googlematt]

This is the difference it makes for me. Granted I'm still dialing things in (new printer, new filament), but this is the difference between 4.5 and 4.6. This is with my tweaked profile, but it's the same profile across both.

[smartavionics]

The problem with sharpest corner has always been that the sharpest corner can occur anywhere so when you have a wall like this...

````

                                                 |
                                                  -------------------------------------------------

````

(Not sure if my AA looks right for you, I think it depends on the font used)

The sharpest corner can occur at that little line segment.

I haven't actually tested this hypothesis but I think that in 4.5 and earlier the polygon simplification would have removed that little line segment but in the 4.6 version (due to https://github.com/Ultimaker/CuraEngine/pull/1238/commits/ad1dbf395c98ef1826c47826305854cbe938615d) that line segment won't get removed. Arguably, that's a regression.

[googlematt]

Maybe I should have put "Regression" in my title haha. Been a while since I worked Software QA

[Ghostkeeper]

In current master the behaviour is yet again different from what you're seeing (we've made more changes to the Simplify algorithm since then), but indeed I can see the bug happening there too. It's indeed caused by a tiny hook in the contour that causes the sharpest corner to be found there.

[Ghostkeeper]

I think I had 4.6 checked out by accident. This bug is fixed by https://github.com/Ultimaker/CuraEngine/pull/1214/commits/1afa4507c180caea34dd0e08d94546749842760c. If I repeat the slice with master I don't see this happening any more.

We're looking into what we can do for 4.6 now with this bug.

[Sophist-UK]

It seems to me that the obvious thing is to disregard sharpest corners where one or both walls are shorter than xmm.

[rburema]

We're taking this very seriously. Look for an update on this around mid-next week.

[googlematt]

We're taking this very seriously. Look for an update on this around mid-next week.

Thanks, I look forward to it! Glad I switched to 4.6 in the middle of tuning a new printer and had clear before and after results.

[nallath]

This will be fixed in 4.6.1

[googlematt]

@nallath @rburema thanks for the quick turnaround on this :)

[brunoosti]

This problem still there in 4.8