Sketcher: Solver Failures on Scaling Factor Changes

Is there an existing issue for this?

• I have searched the existing issues

Forums discussion

No response

Version

0.21 (Development)

Full version info

OS: Windows 10 Version 2009
Word size of FreeCAD: 64-bit
Version: 0.21.0.31847 (Git)
Build type: Release
Branch: master
Hash: 502ef8c9ba619671b5137833851a7ebbbab24bf0
Python 3.8.10, Qt 5.15.2, Coin 4.0.1, Vtk 8.2.0, OCC 7.6.3
Locale: French/Canada (fr_CA)

Subproject(s) affected?

Sketcher

Issue description

Apologies, this bug report is going to be messy on account of ongoing trial and error when these messages occurred and not noticing them immediately.

Got two different messages saying to report the issue while trying to add Spreadsheet and Expression based scaling to my model:

Important: the LevenbergMarquardt solver succeeded where the DogLeg solver had failed. and Important: the SQP solver succeeded where all single subsystem solvers have failed.

Initially had failed recomputes with little else to go on - the "Report Window" is unfortunately quiet on these things - started suspecting the way my constraints are set might break on a diagonal once it starts using scaled values.

I removed one length constraint on diagonal 18-Line and replaced it with a vertical one using expression sqrt(pow(<<Dimensions>>.DiagonaleEntreNiveaux; 2) / 2) instead.

I had inserted rows and renamed aliases along the way and it looked like the rename did not propagate correctly, which broke some constraint references whom were no longer following changes in my scaling factor, instead trying to get the original unscaled value in column C.

I tried scaling factors of 0.5 and 2, going back to one with a successful solve in between.

To see my sketch scale I would sometimes need to double-click on a constraint and cancel out of that dialog.

See attached for a version of the file taken while writing this report.

Anything else?

T-RV-7.zip

Edit:

Also related: https://github.com/FreeCAD/FreeCAD/issues/21553. A different jump likely happens there with the same effect.

Code of Conduct

• I agree to follow this project's Code of Conduct