0
$\begingroup$

I noticed something weird that nobody seems to have reported yet (but I am still using Blender 3.3.12, checked that the result was identical with 3.4.1 but can’t run newer versions on my current config): it looks like (at least these version of) Blender can’t return reliable results when you want to check the scale of an instance after you applied to it a negative scale in order to mirror it. The 3 attached screenshots show that

  • if I apply a scale [-1, 1, 1] without any rotation [0°, 0°, 0°], the spreadsheet returns a scale [1, 1, 1] and a rotation [0, -0, π],
  • if I revert the scale to default [1, 1, 1] and apply a rotation [0°, 0°, -180°], the spreadsheet returns exactly the same scale and rotation values than with the previous setting whereas we can check in the 3D Viewport that the transformation is obviously not equivalent to the previous one.

initial

mirror X

half turn Z

If I add two nodes Instance Scale and Instance Rotation linked to two respective Store Named Attribute, they will also return the same wrong values than the spreadsheet.

$\endgroup$
2
  • 1
    $\begingroup$ I reproduce your disturbing results with Blender 3.6.5, starting from a fresh file also (scaling along X by -1 shows as rotation of 180° around Z axis in Spreadsheet). $\endgroup$ Jan 22 at 20:35
  • $\begingroup$ Also a full "point symmetry", that is negative scale on all 3 axes [-1, -1, -1] shows as a simple rotation of 180° around Y axis $\endgroup$ Jan 22 at 20:56

1 Answer 1

1
$\begingroup$

This issue has already been reported before. See:

As you see, task is marked by solved, but 2 weeks ago. Fix is not in official versions yet.

You can see fix in alpha version of 4.1 you can grab it from daily builds to check:

enter image description here

Unfortunately, it is not implemented for 3.3.15

was simply wondering if the current alpha version of 4.1 could possibly be not totally fixed yet and returning a wrong mix between two possible ways in terms of Rotation + Scale to describe a same transformation matrix.

I made some tests, and it looks to me that it is fixed. The reason why it looks weird is that all transforms are applied to a transform matrix, and in spreadsheet you will see a decomposed version from this matrix. In this case rotation 180 on Y and scale (-1,1,-1) will result in the same transform matrix, that's decomposed in the same way:

enter image description here

enter image description here

$\endgroup$
2
  • $\begingroup$ Which was the transformation used while capturing your screenshot? If it was the point symmetry that I mentioned above, the scale [-1, -1, -1] is right but there should not be any rotation (I recently discovered using the "Matrix World" add-on, that the point symmetry can be obtained either by a negative scale on all 3 axes [-1, -1, -1] without any rotation, or by a negative scale only on Z axis [1, 1, -1] combined with a half turn around Z axis [0, 0, π], both possibilities lead to the same transformation matrix but your screenshot looks like a wrong mix of them). $\endgroup$ Jan 22 at 23:08
  • $\begingroup$ Of course thanks anyway, this information answers to my question even if I can’t run the 4.1 alpha version since my system don’t have the required GLIBC_2.28 ; I was simply wondering if the current alpha version of 4.1 could possibly be not totally fixed yet and returning a wrong mix between two possible ways in terms of Rotation + Scale to describe a same transformation matrix. $\endgroup$ Jan 22 at 23:43

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .