Skip to content

Baseplates: a magnet pocket and a screw bore of nearly the same width can leave bad edges #77

Open

Description

@Oliver-Johnson

Found by the review of #69 and the screw-pocket work for #73 and #74. These are the cases those fixes leave. The review found them on main, with no joint involved.

What happens. When a magnet pocket and a screw bore share an axis and their walls nearly coincide, the two polygons' walls weave through each other and the plate builds with bad edges. Checks gives no warning and Download stays on.

Cases.

  • A magnet from above no wider than the shank under it. A 2.2 mm magnet from above over a 2.4 mm shank gives 24 bad edges a piece (35.93 mm, half cells, loose fit). On the screw-pocket branch, the same bores at 35.93 mm leak for shanks from 2.27 to 2.33 and are clean from 2.35; at 42 mm all are clean.
  • A counterbore nearly as wide as a magnet from beneath, and deeper than it. On main, a 10.28 mm magnet under a 10.54 mm counterbore gives 435 bad edges (40 mm, tight). The screw-pocket branch no longer cuts a counterbore that the magnet pocket holds whole, which closes that exact pair. It still leaks when the counterbore is deeper than the magnet: at 42 mm with a 3 mm counterbore over a 6 x 2 magnet, a 6.04 mm head leaves 6 bad edges and 6.048 folds 5 faces.

Possible fix.

  • A magnet from above whose pocket doesn't clear the shank's corners has nothing to sit on over the hole. Refusing it in Checks ("the magnet needs to be wider than the screw hole under it") is probably right.
  • A counterbore within a small margin of the magnet pocket's corners could be cut as one prism with it, or the head held to clear the magnet pocket by a measured margin.
  • Add audit rows at the cases above.

Activity

  1. Oliver-Johnson commented on Oct 9, 2026

    @Oliver-Johnson
    OwnerAuthor

    One more case of this kind, found while checking the #74 fix on the screw-pocket branch (claude/screw-flats-clearance).

    A magnet pocket's corner on a shank's flat. With a magnet from above and a counterbore, the magnet pocket's corners stand on the same rays as the counterbore's. Where the shank's flats meet them (the #74 mechanism, on the magnet pocket instead of the counterbore), the socket floor round the magnet pocket builds with bad edges. On the branch before the shank turn (3a54fd5), with a 6 mm magnet from above:

    • under a 7.246 mm head at 49.02 mm: shanks of 4.9911 mm leave 44 bad edges, 4.9901 19, 4.9865 15, 4.6974 13;
    • under a 12.053 mm head at 55 mm: 5.8561 leaves 116, 5.8537 84, 5.8547 79, 5.6853 43, 5.6867 25;
    • under a 6.5 mm head at 42 mm: 4.9911 leaves 43, 4.6974 12.

    The branch's turn now keeps both bores' corners clear wherever it turns the shank, but it turns only where a counterbore corner calls for it, so these sizes still leak. A fix could take the turn on the magnet pocket's corners too, or measure the magnet pocket in the same check.


    Generated by Claude Code

  2. Oliver-Johnson commented on Oct 9, 2026

    @Oliver-Johnson
    OwnerAuthor

    A third case, found by the review of #78 on main (f319c83): a magnet from above over a counterbore from beneath of nearly the same width.

    Puzzle tabs at 41.24 mm, a 7.3 mm magnet from above over a 7.504 mm counterbore:

    • With the magnet 2.0 or 2.2 mm deep, the pieces build with 71 to 82 bad edges and Download on, with no warning.
    • With the head at 7.498 or 7.502 mm, 26 to 57.
    • At 2.25 mm deep, as linked, piece A1 fails to build.

    The 7.3 mm magnet's pocket is 7.5 mm across its corners and the 7.504 mm head's is 7.504, both 14-sided on one axis, so their walls stand 0.002 mm apart where the two pockets meet. That is this issue's class; I inferred it from the sizes and haven't traced the mesh.


    Generated by Claude Code

  3. Oliver-Johnson commented on Oct 10, 2026

    @Oliver-Johnson
    OwnerAuthor

    Re-checked on main after #81 (c0dc948). All three cases in this issue are closed or refused now: a shank wider than a 2.2 mm magnet pocket is held to the pocket (2.21 to 2.45 refused, the 42 taken sizes build clean), the 6 x 2 magnet under a 3 mm counterbore is clean for every head from 6.000 to 6.100 in 0.001 steps, and the 10.28 mm magnet under a 10.54 mm counterbore is clean at 40 mm. A sweep of about 43,000 near-equal sizes found a different leak (6 to 12 open edges a cell at 42 mm, in the bottom face's fan), which is filed as #92 with repro links. #93 adds audit rows for this issue's own cases and says "Refs #77", so this issue is still open; it can be closed once #92 is fixed or if you prefer to track the rest there alone.


    Generated by Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Labels

bugSomething isn't working

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions