[adamjohnwright]

Spec 001, T008 (research R4), the low findings of #106's review. Each was reproduced as a failing test first, and each fix has a mutant that the tests catch.

• Letter holes. The counters of B, O, d… are left out before the baseline vote. Before, a short label's holes could set the baseline ("BBB" read 62% small). Release 98's smallest labels are long names and were unaffected: R4's numbers, recounted with this reader, stand.
• Exponents (1e-3, 1E-3) are read as numbers, not as commands.
• A subpath begun after a close without a move starts where the last one began, instead of joining it.
• Each command's count of numbers is checked (C takes six, not any even count).
• A transformed label, or one inside a transformed group other than the render service's viewport, is unread, so the figure is reported as unmeasured. The groups are followed through the markup in one pass of its tags.
• An ancestor's illustration standing in for an event links with the event selected.
• Stale wording of the 5 pt floor is gone from the code's comments, from R4, and from T018's note.

All 150 tests pass against the fixture graph.

🤖 Generated with Claude Code