FreeCad crash using Thickness on a Loft with elliptic section

[wschildbach]

I am reporting what I believe to be a Freecad bug here to get a) confirmation this is a bug before entering into Mantis b) in the hopes of finding general advice on how to avoid trouble using the Thickness tool in non-pathological situations.

What I found is that making a loft in the part design workbench involving an ellipse, and then applying thickness on the loft, results in "surprising" behavior. The attached file contains a loft from a circle ("bottom" sketch) to an ellipse ("top" sketch). Both sketches are fully constrained.

shelltest.png (56.05 KiB) Viewed 286 times

The top ellipse can be turned (through an angular constraint) around its center. Whatever the angle, the actual body should be identical (disregarding a rotation).

If the top ellipse is constrained to be either (sketch-)horizontal (angle = 0) or vertical (angle = 90), then a thickness operation (selecting the top face first) succeeds.
If the top ellipse is turned to 45 degrees, then the thickness operation fails with the dreaded message

"0 <App> Document.cpp(3464): Failed to recompute shelltest3#Thickness: BRep_API: command not done
Recompute failed! Please check report view."

I guess this is an OCC problem, which I understand FreeCad devs cannot do anything about. Is there any general guidance ("avoid overlapping split lines") that might save the thickness operation?

Now on to the actual bug, if I turn the ellipse to 89 degrees, then FreeCad crashes. Should I report this as a bug in Mantis? I have seen https://tracker.freecadweb.org/view.php?id=4262 but that bug apparently involves pathological geometry so may or may not have the same root cause.

I ran into this while trying to make a somewhat more complex model (the body/hull of a glider plane); I don't think there is any easy way to replace thickness which is why I really need the thickness operation to succeed (rather than other workarounds).

Here is the example file:

shelltest3.FCStd

(21.21 KiB) Downloaded 11 times

OS: Windows 10 (10.0)
Word size of OS: 64-bit
Word size of FreeCAD: 64-bit
Version: 0.19.19694 (Git)
Build type: Release
Branch: master
Hash: 288456ee4beb9e8dc27d50a7984c40c45467161f
Python version: 3.7.3
Qt version: 5.12.5
Coin version: 4.0.0a
OCC version: 7.3.0
Locale: English/United States (en_US)

[wschildbach]

PS: I understand that the Thickness tool is rather buggy so don't really expect to get a solution

This post https://forum.freecadweb.org/viewtopic.php?t=27708 has also reported crashes on the thickness operation.

[chrisb]

Thicknes fails with a value of 1.0, it works with 0.9 and 0.8, and fails again with 0.7. If it fails it doesn't crash, it shows a message in report view instead.

OS: macOS High Sierra (10.13)
Word size of OS: 64-bit
Word size of FreeCAD: 64-bit
Version: 0.19.19717 (Git)
Build type: Release
Branch: master
Hash: 43e6be9020199e6453653e1b041679679efee574
Python version: 3.8.1
Qt version: 5.12.5
Coin version: 4.0.0
OCC version: 7.4.0
Locale: C/Default (C)

[papyblaise]

Why you turn the elips at 45° : it generate 2 generatrices
pass at 0 (parallel at X) , all it's Ok

[wschildbach]

Why you turn the elips at 45° : it generate 2 generatrices
pass at 0 (parallel at X) , all it's Ok

Yes indeed, 0 and 90 degrees work, 45 throws an error message and 89 crashes Freecad.

This is just an example file, and there is indeed no purpose of the angle=45 other than showing that (seemingly) identical geometrical bodies can make the thickness command react in very different ways.

I am interested though in what you write about generatrices --
a) is there anywhere where I can read up on what a generatrix is, and what role it plays in constructions and the loft in particular?
b) specifically, it seems that when generatrices coincide in a construction, that is asking for trouble (read this in a different post that I can't find right now). Is there any more wisdom on this?

Thanks for your help.

- Wolfgang

[l3VGV]

Basically - everything involving bsplines or curved surfaces in any way, will produce strange behavior or crash.

Here is my experience with thickness https://forum.freecadweb.org/viewtopic.php?f=3&t=43458

[papyblaise]

There are no general rules,
you can read in the forum that you must have the same number of segments between a base and the other with connection points in parallel planes (which will generate common generators), and there are exceptions,
you can also read that the loft generates B-spline that Fc manages very badly, and there are exceptions
you can read that shells go better inside than outside, and there are exceptions
with all these exceptions, only a try will tell you if it's ok
otherwise you have to find a workaround: make a 1/2 shell, or make an extrusion of 2 faces with an outside outline and an inside, or do a boolean subtraction of the outside volume minus a scaled clone, or do a reduced inside volume of the thickness to keep and booleen subtraction
you have a lot of solution, you have to find the right one and there are no general rules, they all have advantages and disadvantages, and vice versa

[wschildbach]

Thanks @chrisb for the bug confirmation and @papyblaise for the general advice / hints. What I take away from this is

• stay away from Thickness when the Part has any curved surfaces (i.e., Lofts in particular)
• The problem is in the OCC kernel; and the crash happens there, too. Besides catching/masking the fault and exiting gracefully, FreeCad can 't do much about it
• I won't be reporting a bug as there are a number of other bug reports in Mantis already that have that same solution (catching the fault)

Cheers
- Wolfgang

[chrisb]

I won't be reporting a bug as there are a number of other bug reports in Mantis already that have that same solution (catching the fault)

Good idea not to spoil the tracker. The real problem, that it doesn't work can't be fixed anyway by the FreeCAD team.