main is red: test_spec_page tie-break test predates #120's p-suffix rendering

[espg]

🤖 from Claude

Pre-existing failure on main (@ 4b57a1f), surfaced during PR #122's fold (evidence: Tests run 29787103143 red on 3.10 + 3.12 with the identical failure; #122's own diff is disjoint and its 23 new tests pass).

mortie/tests/test_spec_page.py::TestDecimalParseTieBreak::test_order29_string_parses_to_area_word fails with ValueError: invalid literal for int() with base 10: 'p' — issue #120's "delegate point suffix to kernel" made decimal_repr() render the terminal p kind-suffix on point words (spec §2: the render form MAY carry p on point ids), but the test at line ~87 still constructs its probe via int(dec[-1]) on a repr that is now p-terminated.

Likely fix is test-only (the behavior matches the spec): build the unmarked order-29 probe string deliberately — e.g. dec.rstrip('p') (or render an area order-29 word instead of a point) — so the test exercises what it means to pin: the unmarked string parses to the AREA word per the §4 tie-break. Worth one check while there: that the tie-break assertion itself still passes once the probe is well-formed, and that a p-marked string round-trips to the point word (which #120's own tests may already cover).

Why it matters beyond hygiene: it blocks PR #122's test matrix from going green (and therefore the release moczarr#8 adoption waits on). Not fixed inside #122 per the pre-existing-failure discipline; flagged there with the same diagnosis.

Refs: #120, #62 (the spec page + drift-pin), PR #122 (where it surfaced), espg/moczarr#8 (downstream).

[espg]

post plan here and implement

[espg]

🤖 from Claude

Diagnosis confirmed: TestDecimalParseTieBreak::test_order29_string_parses_to_area_word derives t28, t29 from dec[-2]/dec[-1], but after #120 decimal_repr() terminates a point word with the p kind suffix, so dec[-1] == 'p' and int('p') raises. The runtime is correct per spec (§2 render carries p; §4 unmarked order-29 string ties to the AREA word) — the pin, not the behavior, is stale.

Fix (test-only): strip the marker to build the unmarked probe (dec = marked.rstrip('p')), then keep the existing tie-break assertions on that well-formed string. While there I'll (1) assert marked.endswith('p'), (2) keep the tie-break parsed & 0x3F == 28 + t28*5 + t29 + 1 / same-path / non-injectivity checks, and (3) add that the p-marked string round-trips back to the POINT word (_decimal_to_word(marked) == word). No kernel / decimal_repr change.

Then green gates (maturin develop --release, flake8 --select=E9,F63,F7,F82, pytest -v) and a draft PR labeled implement.

---

Generated by Claude Code