Selection order should not matter for constraints

[Irev-Dev]

Here's a video of how to replicate the bug for the equal-length constraint, but it, likely applies to a few other's, so part of the issue is checking the other constraints.

https://github.com/KittyCAD/modeling-app/assets/29681384/c8e71d92-7c74-4151-a014-8dc08ba25a00

Tasks:

• [ ] Fix for equal length constraint
• [ ] Check which other constraints are affected and fix them
• [ ] Make sure there are tests, a unit test might suffice since it's likely an AST mod bug.

Pointers: equal length code is pretty much all in src/components/Toolbar/EqualLength.tsx so that's where I would start. It might be as simple as reordering selectionRanges.codeBasedSelections by source ranges, but I'm not sure.

[jtran]

I'm having trouble reproducing this in my development environment. I started by pasting the same initial kcl into the editor, and then I edited the sketch visually in the same way. But the resulting kcl looks like this, which is what we expect.

const sketch001 = startSketchOn('XZ')
  |> startProfileAt([0.76, 0.58], %)
  |> line([0.35, 0.83], %, $seg02)
  |> angledLine([35, segLen(seg02, %)], %)
  |> line([2.12, 0.11], %, $seg01)
  |> angledLine([-65, segLen(seg01, %)], %)

Was there some other action done that was cut from the video that might have led to an invalid state?

I also tried a few variations like sketching the initial lines, changing whether I exited sketch mode, and changing which segments I constrained.

[jtran]

We determined that the problem only arises in Chromium-based browsers. Safari, Tauri app on macOS, and Firefox all work as expected.

[Irev-Dev]

Make sure there are tests, a unit test might suffice since it's likely an AST mod bug.

We might chat about the best way to test this after we figure out what the bug actually is, because my initial hunch of AST mod bug seems less likely with the "not on webkit" revelation.

[jtran]

After investigating it some, I realized that I could only reproduce the problem when DevTools was closed. So this is likely a race condition and not browser-specific.