Part::checkIntersection returns false negatives preventing functioning of PDN Mirror transformation when faces just touch

[FreeCAD-Bug-Importer]

Issue imported from https://tracker.freecad.org/view.php?id=3065

• Reporter: @abdullahtahiriyo
• Date submitted: 6/1/2017
• FreeCAD version: 0.17
• Category: Bug
• Status: assigned
• Tags: Mirror transformation

Original report text

Problem 1

Part::checkIntersection first uses first_bb.IsOut(second_bb) for a quick test. However, this function is known to fail in cases of touching (there is even a previous comment in the code). Mirror transformation explicitly asks that touching to be considered intersection. Therefore it is wrong to rely on this test.

Problem 2

The Extrema method implemented in Part::checkIntersection fails:

Case 1: When mirroring a feature that just touches, it sometimes fails because BRepExtrema_DistShapeShape returns BRepExtrema_IsVertex instead of BRepExtrema_IsInFace, so the manyfold detection mechanism fails to identify a non-manyfold case. In one example, instead of InFace it returns 4 Vertex: [0, 5.5, -2.5], [0,5.5,2.5], [1E-15, 13, -2.5], [1E-15, 13, 2.5] for Support1 and [0, 5.5, -2.5], [0,5.5,2.5], [-1E-15, 13, -2.5], [-1E-15, 13, 2.5] for Support2. I am unsure if this can be somehow reused.

Case 2: When doing a multitransformation to create a part out of a quarter like here:
https://forum.freecadweb.org/viewtopic.php?f=19&t=22747#p176513

The extrema method provides a separation (notable, in the order of milimeters) in operations with intersection parts. This may be a consequence of which parameters we are passing the to function.

OTHER ISSUES

Some related problems not directly related to the extrema method, but that seem to be related:
https://forum.freecadweb.org/viewtopic.php?f=19&t=22747#p176509

Additional information

https://forum.freecadweb.org/viewtopic.php?f=19&t=22747

Steps to reproduce

EXAMPLE 1

1. Load the attached file.

2. Activate the first body

3. Select the pad and add a Mirror transformation (you can try directly the mirror, or a multi-transformation, vertical+horizontal to see the different cases)

EXAMPLE 2

1. Load the attached file

2. Activate the second body

3. Select the pad and mirror or multi-transform mirror the pad

Other bug information

• Priority: high
• Severity: major
• Category: Bug
• Updated: 2/6/2021
• Fixed in version: 0.17

Discussion from Mantis ticket

Comment by kkremitzki 2017-06-18 03:49

abdullah should this be resolved now that #816 has been merged?

---

Comment by abdullah 2017-06-19 12:49

Yes, this ticket is to be closed as fixed by that PR

---

Comment by Kunda1 2017-06-19 17:12

Thanks! Closed.

---

Comment by abdullah 2017-06-25 09:36

The boolean check method fails in some circumnstances where the extrema method success. For example:

https://forum.freecadweb.org/viewtopic.php?t=23029#p178800

The result is that "the shape does not intersect support" is generated, when the real problem is OCE/OCCT can not make a proper fusion of those valid and perfectly intersecting objects.

IMO it is not ok to inform that something that clearly intersect does not. So this is a bug.

The final outcome in these cases should be: no error because it does intersect, and a resulting bad solid because it is what OCC gives back. This can be solved by a three phase checking (instead of a two phase one):

1.Quick check
2. Extrema
3. If Extrema says it does not intersect (sometimes because of the non-manifold detection failure), boolean check.

In case of OCC>6.9.0, this checkIntersection, as well as all the fusing in PDN should probably be upgraded to use generalFuse (same as Part boolean fragments fuse).

---

Comment by DeepSOIC 2017-11-20 10:08

How about avoiding the check altogether? I.e., mirror, then fuse, then check the result of fusion (it should remain a single solid; if it doesn't -> show the "does not intersect support" message to user).

Another option is to drop the single-solid limitation altogether. Then, the check can also be removed.

---

Comment by abdullah 2018-01-02 17:56

DeepSOIC

Good to hear from you :)

What we have currently in master seems to work well. I am not sure if I "fixed" this, or what exactly went on (I can not remember).

I have just updated the target version to 0.18 to get it out of the 0.17 roadmap, as this is no urgent matter.

I do not close the ticket. I will eventually come back to this and assess it again.

---

Comment by Kunda1 2019-04-06 19:15

Changing ticket status from 'feedback' to 'assigned'

[chrisb-github]

Fixed in this version:

OS: macOS 10.16
Word size of FreeCAD: 64-bit
Version: 0.20.28751 (Git)
Build type: Release
Python 3.9.12, Qt 5.12.9, Coin 4.0.0, OCC 7.5.3
Locale: C/Default (C)