[BUG] 0.20.26556 sketch / padding visual glitch

[MRx]

OS: macOS 10.15
Word size of FreeCAD: 64-bit
Version: 0.20.26556 (Git)
Build type: Release
Branch: adaptive_dicretize_bugfix
Hash: 4357816c94ff4938b778eb9f8aba8fe0e1c917dc
Python version: 3.9.0
Qt version: 5.15.2
Coin version: 4.0.0
OCC version: 7.5.0
Locale: English/Germany (en_DE)


it doesn't happen with an installed LS3 version
ok 26202 is also ok
26556 shows up that this issue after padding
I also reset 26556 configuration again .. and it still happens

Without construction lines this issue doesn't happen.

To reproduce this issue remove the pad and just pad it again.

Can someone reproduce this issue, and if which version. Anything below 26202 no need to test..

[chrisb]

Removing and recreating the Pad shows the artefactual plane. However, it vanishes on a full recompute. It's quite sure not the construction line, a change of the sketch only triggers the recompute.

Besides. it is another argument to not attach things to faces.

OS: macOS High Sierra (10.13)
Word size of FreeCAD: 64-bit
Version: 0.20.27078 (Git)
Build type: Release
Python version: 3.9.9
Qt version: 5.12.9
Coin version: 4.0.0
OCC version: 7.5.3
Locale: C/Default (C)

[kisolre]

Happens also here but you can expect all kinds of weirdness with not properly defined geometry. Like almost but not quite tangent edges for example. If that sketch is properly constrained there are no problems with that second pad:

constructionlines_kiss.FCStd

(20.04 KiB) Downloaded 24 times

And all works well even with only proper tangents and one equality between the big arcs:

constructionlines_OnlyTangents.FCStd

(22.17 KiB) Downloaded 23 times

[MRx]

thanks for testing & confirming.

I did some more investigation here to isolate the problem -- it is indeed not an issue with the construction lines.

it does not happen with 26202, neither with an older LS3 version.
I just fear that another issue might have slipped in with some change between 26202 - 26556

I'll set up another notebook with FreeCAD next month for bisecting (I'm just waiting for some replacement parts to fix it), I'm almost running out of space here on my notebook.

attached I have disconnected the sketch from the face and removed all the construction lines, my initial assumption was wrong.
Just change the Pad001 length and it will appear, and even a recompute on that object will not make it go away.
Purpose of this is to trigger the glitch so it can be bisected / fixed later.

[chrisb]

From kisolre's observations I would strongly assume that it's an OCC issue. The branch is a completely diferent game which should be considered separately.

[MRx]

This one works, no glitches. It uses the same OCC version as your FreeCAD version but you are getting this glitch.
This is a downloaded freecad build snapshot.

OS: macOS 10.15
Word size of FreeCAD: 64-bit
Version: 0.20.26202 (Git)
Build type: Release
Branch: master
Hash: ca6d49d080dea0abc23d954743eca7c46f33469b
Python version: 3.9.7
Qt version: 5.12.9
Coin version: 4.0.0
OCC version: 7.5.3
Locale: C/Default (C)

There seems to be a face which is crossing lines - this shouldn't happen, it should rather be 2 faces or not cross.
I see nothing special on the tolerance side.

CE0BUb.png (10.45 KiB) Viewed 798 times

Now if you select the edges with one of the supposed working versions it will select the entire face of either the circle or the shape where the circle is sitting on - but nothing merged.