Missing property Object1 or Object2 of fixed joint causing crash

[Jiri-Macha]

FreeCAD is crashing if the 'Object' property of fixed joint (Assembly/Joints/Fixed/Joint Connector 1/Object1 or Assembly/Joints/Fixed/Joint Connector 2/Object2) is manually removed.

Steps to reproduce:

• make simple Assembly e.g. of two cubes with Fixed joint
• Select Fixed joint in the tree and go-to property 'Data' tab
• Select 'Object1' or 'Object2' of the 'Joint Connector 1' or 'Joint Connector 2'
  and remove this reference
• click by your pointing device (mouse) to the arbitrary other property

FreeCAD will crash here because the call App::DocumentObject* obj = getObjFromNameProp(joint, propObjName, propPartName); will return NULL pointer.

This problem is similar to the 4b5d079.

[FEA-eng]

There's no linked issue so I don't know on which OS the crash occurs but I can't reproduce it on Windows (with the AssemblyExample.FCStd file). Are you using Linux?

OS: Windows 10 build 19045
Word size of FreeCAD: 64-bit
Version: 0.22.0dev.37213 (Git)
Build type: Release
Branch: main
Hash: 20e7deb86a8c6c2cd2378f09f8313760933f3a5c
Python 3.11.9, Qt 5.15.13, Coin 4.0.2, Vtk 9.2.6, OCC 7.7.2
Locale: Polish/Poland (pl_PL)

[Jiri-Macha]

Yes please. My system is Gentoo Linux (with Sway tiling Wayland compositor) - almost HEAD.
I just rebuild FreeCAD from the 8fcaaed to test it once again,
this time I made the test with AssemblyExample.FCStd and the result was the same - crash - see also the
screenshot below. It is a bit surprise to me that it works on Windows. I would like
to know what is returned after the call of getObjFromNameProp(joint, propObjName, propPartName)
in such use case (after removal of Object of Joint Connector). I might explore it later on this
platform as well.

Based on your observation (inconsistency between platforms Windows/Linux) and the fact that the same
calls (of function getLinkObjFromProp and getObjFromNameProp) are made also in the AssemblyObject::getRackPinionMarkers function and return values (part1 and obj1) are there also not checked, I think, there are some more defects there, and we will need to patch it also in the near future.

In any case, this patch (pull request) is helping in my use case (Linux + Wayland compositor), is ensuring
validation of returned value and as such should not have negative side effects.

[FEA-eng]

@PaddleStroke Can you have a look at this?

[PaddleStroke]

I see no crash on my side either (windows) but the this PR can be merged as it just adds a protection it's totally acceptable.

[Jiri-Macha]

I just pushed one more patch for the rack pinion joint where the crash could also occur.