[Sketcher] Part of sketch flips when changing dimensions in fully constrained sketch

[gntech]

OS: KDE neon User Edition 5.16 (KDE//usr/share/xsessions/plasma)
Word size of OS: 64-bit
Word size of FreeCAD: 64-bit
Version: 0.19.17140 (Git) AppImage
Build type: Release
Branch: master
Hash: 961ffce93e0fff6f1bf8326af248fda3f608215f
Python version: 3.7.3
Qt version: 5.9.7
Coin version: 4.0.0a
OCC version: 7.3.0
Locale: English/UnitedStates (en_US)

I have noticed a possible bug in the way the sketcher solves constraints on a sketch when a dimension is changed. See pictures below and their subtitles for an explanation.

[juless]

Looking at it briefly, although the sketch is fully constrained, you may want to change the dimension constraint for another symmetry constraint. As you can see in your example, its ambiguous where changed line should go - it's definitely not symmetrically around the y axis as I think you intend.

[chrisb]

This behaviour is well known and it is not considered to be a bug: All constraints are fulfilled in both cases.
Nevertheless this could be improved to work more intuitively and there is ongoing work on it from time to time.
Levenberg-Marquardt solver (use it after switching on the expert settings in Sketcher's preferences) usually handles this better.

You can improve things by using different constraints. Symmetry was already mentioned. Using angles instead of distance dimensions often helps as well.

[bejant]

Here is a method using just a few different Constraints, following the suggestion that chrisb wrote, making Sketch become more robust:

20190626a.png (14.26 KiB) Viewed 1106 times

The most important new Constraint is probably the Symmetry Constraint that was applied between the bottom outer vertices and the Y-axis.

20190626a-sketch_example.FCStd

(5.42 KiB) Downloaded 39 times

[gntech]

Sorry for replying so late

Levenberg-Marquardt solver (use it after switching on the expert settings in Sketcher's preferences) usually handles this better.

Thanks for the suggestion, this solved my issue! Levenberg-Marquardt behaves the way I am used to from commercial CAD software. It maintains the topology of the sketch when editing a fully defined sketch.

Can this be set as the default for all sketches? Is there some known drawback with it?

[chrisb]

I remember that LM-solver on some occasions was not able to solve a mathematically solvable sketch.