tests/test_value_pattern_line_break.py: ID_READERS says checked_value refuses a line feed in adopt-project.py's --id, which never calls it, and the docstrings say a fullmatch reader passes the guard when it must be listed (#1166 follow-up) #1229

(lane openRepoShape-5, 2026-10-23)

Goal: tests/test_value_pattern_line_break.py says only what is true of the readers of PROJECT_ID_RE: no row of ID_READERS gives a reason that does not hold for the tool it names, and the guard test_every_reader_of_the_id_pattern_is_in_the_table tells the author of a new reader, in its docstrings and in its failure message, that the one way to pass is a row in ID_READERS, whichever method the reader calls.

Problem

This issue applies once #1166 (#1139) has landed. #1166 is open at 5102b4294 (read with gh pr view 1166 on 2026-10-23, updated 18:04:21Z), and its second commit adds the guard. Its body was corrected for the two wordings below that it repeated (it says that adopt-project.py never calls checked_value on --id, and that "ID_READERS still gives it the checked_value reason"), so only the test file is left. Line numbers are tests/test_value_pattern_line_break.py at 5102b4294.

1. A false reason for adopt-project.py's argv --id (rc1166b's NIT 2).
   • The module docstring's "LEFT AS THEY ARE" bullet (:42-45) reads: "The argv readers of PROJECT_ID_RE (--id in scaffold-project.py, family.py init and adopt-project.py): each holds the id to checked_value after the pattern, which refuses a line feed as unsafe-value naming --id (tests/test_derived_id_refusals.py)."
   • The ID_READERS rows for scripts/family.py (:158-159) and scaffold-project.py (:160-161) say "argv --id: checked_value refuses a line feed after it". The adopt-project.py row (:162-166) begins "argv --id, as above", which is the same reason.
   • It holds for two of the three. scaffold-project.py:2517 calls checked_value("--id", project_id) and scripts/family.py:2626 calls checked_value("--id", family_id); tests/test_derived_id_refusals.py's TOOLS (:59) pins that answer for scaffold and family only.
   • adopt-project.py never calls checked_value on --id. Its calls (git grep 'checked_value(' 5102b4294 -- adopt-project.py) are for --project (:5401), the tracking branch, --pin-owner, the leg paths and --org, and for the plan's keys. Its reader of the id plan takes, passed or derived, is PROJECT_ID_RE.match(project_id) at :5140, which takes northwind followed by a line feed.
   • A line feed is still refused, by another door: the project's own validate-manifest.py (measured below, for plan).
2. The docstrings say fullmatch is enough (rc1166b's NIT 1).
   • The module docstring (:57-60): "A NEW READER of PROJECT_ID_RE is held to the same rule by test_every_reader_of_the_id_pattern_is_in_the_table: it asks with fullmatch, or it is listed in ID_READERS with the reason a line feed cannot reach it."
   • The test's docstring (:196-198): "A reader written later that asks PROJECT_ID_RE with match (or search) takes "atlas\n": it is read with fullmatch, or it is in ID_READERS with the reason a line feed cannot reach it, or this fails."
   • The guard is assert id_pattern_calls() == expected (:200), a dict equality over every (module, method) of every call. It is exact in both directions, so a new fullmatch reader fails it too, and so does a listed reader taken away. The assert has no message, so the author of a sibling PR sees only a dict diff.
3. The guard reads spellings, not meaning (rc1166b's NIT 4, optional). _ID = PROJECT_ID_RE followed by _ID.match(value) passes it, and by reading so does from repo_shape import PROJECT_ID_RE as X. id_pattern_calls (:170-192) counts only a call whose receiver is the name PROJECT_ID_RE or an attribute of that name. The body of validate-manifest.py, validate-family.py, family.py and update-shape.py: an id:, a date or a repository: ending in a line feed is refused by name, and a pin's source_repository: that is no path or owner/repo name is named as the pin's (#1139) #1166 claims only the two spellings, and id_pattern_calls's docstring names the same two, so nothing in either is false; an alias is simply outside the guard.

Measured

Lane 5's re-check of #1166 (#1139), rc1166b, NIT 1, NIT 2 and NIT 4 (#1166 issuecomment-6371408915), at 5102b4294:

• NIT 2. adopt-project.py plan --id=northwind<LF>, from a local source through conftest's write_plan, exits 2 and writes no plan: REFUSED adopt-manifest-invalid: --id is 'northwind\n', which the project's own validate-manifest.py refuses: manifest-id: id is 'northwind\n'; it must match ^[a-z0-9][a-z0-9-]*$. Measured for plan only; apply and the other adopt paths were not.
• NIT 1. def _rc_probe(value): return PROJECT_ID_RE.fullmatch(value) appended to update-shape.py fails the guard with Left contains 1 more item: {('update-shape.py', 'fullmatch'): 1}. Making one adopt-project.py reader unreachable (if False: in place of the plan-id: test at :6791) fails it too: {('adopt-project.py', 'match'): 2} != {('adopt-project.py', 'match'): 3}. Neither failure names ID_READERS.
• NIT 4. _ID = PROJECT_ID_RE then _ID.match(value) in update-shape.py: both guards, 2 passed. A PROJECT_ID_RE.match(...) inside a string or a comment is correctly not counted: 2 passed.
• The file on the head, through the review's run9.sh (stub gh, GH_TOKEN=bogus, EMAIL exported, dirty=0): 30 passed on git 2.43.0 (31.59s) and on 2.47.1 (29.43s).

The drafter re-read the three places and the checked_value calls at 5102b4294 (git show of the test file, adopt-project.py, scaffold-project.py, scripts/family.py and tests/test_derived_id_refusals.py).

Suggested fix

Wording and one message, in the test file only; no production file.

1. Correct the reason. In the module docstring bullet (:42-45), say that the first two hold the id to checked_value after the pattern, and that adopt-project.py does not: plan takes it with match and the project's own validate-manifest.py refuses the line feed (adopt-manifest-invalid, in manifest-id's words). In the adopt-project.py row (:162-166), replace "argv --id, as above" with that reason. The scripts/family.py and scaffold-project.py rows can stay.
   • Optional, so the reason is true by test as the other two are by TOOLS: one row that runs adopt-project.py plan --id with a trailing line feed and asserts exit 2 and adopt-manifest-invalid naming --id. The review measured plan only.
2. Say "listed in ID_READERS" for any new reader. In both docstrings (:57-60, :196-198), drop "it asks with fullmatch, or": a new reader, fullmatch included, needs a row, and a reader taken away loses its row, because the table is exact. A sibling that adds any reader must add one. The body of validate-manifest.py, validate-family.py, family.py and update-shape.py: an id:, a date or a repository: ending in a line feed is refused by name, and a pin's source_repository: that is no path or owner/repo name is named as the pin's (#1139) #1166 found none among the open PRs.
3. Give the assert a message that names ID_READERS and what to do, for example that it lists every call of a method of PROJECT_ID_RE outside tests/, fullmatch included, and that the author adds, changes or drops the row of the module and method the diff shows.
4. Optional, NIT 4: make the guard see an alias, or say in the docstring that it reads spellings and stops there. To see one: an ast.Assign whose value is the PROJECT_ID_RE name or an attribute of that name, and an import as that renames it, each a failure that names the alias and the file. A match in a string or a comment must stay uncounted. Leaving it and saying so is a fair answer.

Verify

• git grep -n 'checked_value' tests/test_value_pattern_line_break.py shows the checked_value reason only against scaffold-project.py and scripts/family.py, and the adopt-project.py row and the module docstring name validate-manifest.py and adopt-manifest-invalid for it.
• A scratch clone with one reader added to a tracked module outside tests/, once per method (fullmatch, match, search) on update-shape.py: the guard fails each time, and its message names ID_READERS. Each plant is restored and the file's sha256 checked, as rc1166b did (restored=True each time).
• A listed reader made unreachable (if False: at adopt-project.py:6791): fails, and the message names ID_READERS.
• The file passes: 30 passed on git 2.43 and 2.47.1 as at 5102b4294 (31 if the optional plan row is added), and the hygiene five (test_pinned_copy_text.py, test_path_guard_hygiene.py, test_repo_hygiene.py, test_agents_md_codes.py, test_cli_reference.py: 306 passed on git 2.43.0 at that head).
• The diff touches only the test file, adds no non-ASCII line and no assert that joins conditions with and or or.
• If NIT 4 is done: the alias plants (_ID = PROJECT_ID_RE then _ID.match(value), and from repo_shape import PROJECT_ID_RE as X) fail the guard, and a PROJECT_ID_RE.match(...) in a string or a comment still passes it.

Found by lane 5's re-check of #1166 (#1139), rc1166b, NIT 1, NIT 2 and NIT 4 (#1166 issuecomment-6371408915), at 5102b4294; the three places and the checked_value calls were re-read at that head by the drafter.

Lane: openreposhape-5 (openRepoShape-5)