I'm not sure if I'm required to create a test for this fix.
The commit which caused the issue didn't have one, so I also don't need one?
In case I need one I don't really know how to create such test since I'm not familiar with in dept programming and appreciate every help to get the issue fixed.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.09%. Comparing base (0859b35) to head (92c5f44).
⚠️ Report is 77 commits behind head on main.

@@           Coverage Diff           @@
##             main    #3489   +/-   ##
=======================================
  Coverage   97.09%   97.09%           
=======================================
  Files          56       56           
  Lines        9664     9664           
  Branches     1750     1750           
=======================================
  Hits         9383     9383           
  Misses        168      168           
  Partials      113      113           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

I'm not sure if I'm required to create a test for this fix.
The commit which caused the issue didn't have one, so I also don't need one?

I have not been aware that the faulty commit would not have proper tests, thus I did not require a new one. With the current regression, we should add one to ensure this does not break again.

In case I need one I don't really know how to create such test since I'm not familiar with in dept programming and appreciate every help to get the issue fixed.

This is a very strange side effect, as the boxes all look correct:

@pytest.mark.enable_socket
def test_scale_by():
    """Tests for #3487"""
    url = "https://github.com/user-attachments/files/22685841/input.pdf"
    name = "issue3487.pdf"
    reader = PdfReader(BytesIO(get_data_from_url(url, name=name)))

    original_box = RectangleObject((0, 0, 595.275604, 841.88974))
    expected_box = RectangleObject((0.0, 0.0, 297.637802, 420.94487))
    for page in reader.pages:
        assert page.artbox == original_box
        assert page.bleedbox == original_box
        assert page.cropbox == original_box
        assert page.mediabox == original_box
        assert page.trimbox == original_box

        page.scale_by(0.5)
        assert page.artbox == expected_box
        assert page.bleedbox == expected_box
        assert page.cropbox == expected_box
        assert page.mediabox == expected_box
        assert page.trimbox == expected_box

Thus, we need a rendering test with Ghostscript here to actually detect this bug (which has already been fixed in the past in #1314 - without adding a test).

If I am not mistaken, the added test is not sufficient, as running it against main is successful as well, not providing any benefits (as outlined in my comment as well). We have to add a test similar to the discussions in the issue by comparing the rendering results with Ghostscript, as mentioned in my comment as well (examples are already available for watermarking for example).

If I am not mistaken, the added test is not sufficient, as running it against main is successful as well, not providing any benefits (as outlined in my comment as well). We have to add a test similar to the discussions in the issue by comparing the rendering results with Ghostscript, as mentioned in my comment as well (examples are already available for watermarking for example).

Ah, the usage of Ghostscript exceeds my current skill level for now and I don't know, when I have the time available to look into that for a proper test. 🤔

I have added a Ghostscript-based test. After resolving the open discussion, this should be ready for merging.