Part of the #241 epic. Registers C-82.
Problem
This arc introduced or left behind claims that are now false. Each is small; the pattern is the point, and it is the same pattern the arc was fixing.
Every item below was verified in this repository.
| # |
Location |
The claim |
Reality |
| 1 |
tests/test_env_declaration.py:511-517 |
A #: comment block |
Garbled edit-splice. Opens mid-sentence — "#: direction is deliberately unchecked." — with the antecedent gone, and a stranded line "#: silent while v1.5.0 adding [contract] is a red build ..." sits below a parenthetical. Reads as one block attached to _ROLES. |
| 2 |
tests/test_env_declaration.py:676-677 |
The partition docstring |
Duplicated clause: "An IGNORED table disappearing is not our business, which is an IGNORED table disappearing is silent while ..." |
| 3 |
.github/workflows/run_pytest.yml:43 |
"...test_env_declaration read the working tree..." |
Zero do. All four were moved to origin/main; this is the founding-incident file again. |
| 4 |
tests/test_ci_sibling_coverage.py:253 |
G7's docstring, "Reading the working tree from development..." |
Same stale premise. |
| 5 |
docs/ADRs/017_*.md:145 |
"None of those three is in place yet" |
D3 landed (registry read, both partners); D4 landed for FAO. |
| 6 |
docs/ADRs/016_*.md §4 |
The public? column |
An unverifiable cross-repo claim with no ADR-014 §4 trigger or owner — in the ADR whose §3 is about exactly that. |
| 7 |
reports/technical_risk_register.md C-86 |
"re-pinning is hygiene with no safety consequence" |
Contradicted by measurement: the previous pin fcf32c9 fails the new drift check for both partners. |
| 8 |
views_postprocessing/{unfao,crafd}/appwrite_env.py:8-9 |
blob/c7b597e/... in both docstrings |
A second, unguarded copy of the sha. The docstring test parses only the version. An annotated-tag pin already shipped a 404 from this exact URL on this branch. |
Work
Fix each. Where a claim was load-bearing, say what replaced it rather than deleting it silently — the register's house style keeps corrections visible.
For #8, add the guard, not just the fix: something must assert that the sha in the URL equals SEAM_CONTRACT_COMMIT and that it resolves to a commit. That is the third time this sha class has bitten.
#6 is a judgement call: either give the public? column a trigger and an owner, or remove the column and say why the fact is not tracked. Do not leave it as an unowned assertion about another repository in the ADR that forbids them.
Acceptance criteria
Implementation notes
This story is deliberately last among the code stories: several of these sentences describe behaviour that #245 and #247 change, and correcting them first would mean correcting them twice.
Items 3 and 4 are the same claim in two files, which is why nobody caught it — worth one grep for the phrase across the tree rather than two edits.
Labels
story, documentation
Part of the #241 epic. Registers C-82.
Problem
This arc introduced or left behind claims that are now false. Each is small; the pattern is the point, and it is the same pattern the arc was fixing.
Every item below was verified in this repository.
tests/test_env_declaration.py:511-517#:comment block[contract]is a red build ..." sits below a parenthetical. Reads as one block attached to_ROLES.tests/test_env_declaration.py:676-677.github/workflows/run_pytest.yml:43origin/main; this is the founding-incident file again.tests/test_ci_sibling_coverage.py:253development..."docs/ADRs/017_*.md:145docs/ADRs/016_*.md§4public?columnreports/technical_risk_register.mdC-86fcf32c9fails the new drift check for both partners.views_postprocessing/{unfao,crafd}/appwrite_env.py:8-9blob/c7b597e/...in both docstringsWork
Fix each. Where a claim was load-bearing, say what replaced it rather than deleting it silently — the register's house style keeps corrections visible.
For #8, add the guard, not just the fix: something must assert that the sha in the URL equals
SEAM_CONTRACT_COMMITand that it resolves to a commit. That is the third time this sha class has bitten.#6 is a judgement call: either give the
public?column a trigger and an owner, or remove the column and say why the fact is not tracked. Do not leave it as an unowned assertion about another repository in the ADR that forbids them.Acceptance criteria
pytest -q tests/test_doc_accuracy.py tests/test_register_integrity.pygreen.Implementation notes
This story is deliberately last among the code stories: several of these sentences describe behaviour that #245 and #247 change, and correcting them first would mean correcting them twice.
Items 3 and 4 are the same claim in two files, which is why nobody caught it — worth one grep for the phrase across the tree rather than two edits.
Labels
story,documentation