scripts/repo_shape.py: `NamingPolicy` reads a name with `fullmatch` at four sites, so a name that ends in a line feed matches no family and `validate-repository-naming.py` refuses it, and `repository-naming.yaml` says a pattern is read whole (#1202) by brettheap · Pull Request #1266 · opensoft/openRepoShape

brettheap · 2026-11-08T16:41:57Z

Goal

#1202: a repository name that ends in a line feed is the name of no family of the naming policy, and validate-repository-naming.py refuses it as naming-unclassified, by name, over a manifest's leg repository: and over an argument, where main classified Atlas\n as Atlas and exited 0.

This is a DRAFT on main (be21751c3, the landing for #1243). It is the follow-up lane 5 filed as #1202 from its independent review of #1166 (#1139), finding 1 (pull/1166 issuecomment-6352173084). #1166 ended the seven module-level patterns of the validators in \Z and read the three ids with fullmatch, and left NamingPolicy, whose patterns come from contracts/repository-naming.yaml, every one ^...$, read with .match. The change is one commit, 09697e9c5: 3 files, +255 and -4 (git diff --shortstat origin/main...HEAD): scripts/repo_shape.py (+5 -4), contracts/repository-naming.yaml (+8), and the new tests/test_naming_policy_line_break.py (+242, 77 rows).

What held on main

Measured on main be21751c3 (production code equal to 5926cccab, where the branch was cut), Python 3.12.3, git 2.43.0, a stub gh and no network.

NamingPolicy read a name with .match at four sites of scripts/repo_shape.py: descendant_referents (the family's _re, then its _split_re) and matches (a role's _re, then a family's). In Python $ also matches just before a trailing line feed, and the YAML reader reads a double-quoted "...\n" as the text and a line feed. So matches, classify and descendant_referents gave a name with a trailing line feed the answers of the name without it (NamingPolicy.load(contracts/repository-naming.yaml), in process):

Name, with and without `\n`	`classify` on main	On this branch, with `\n`
`Atlas`, `Atlas-spec`, `Atlas-code`	`('project-leg', 'assembly' / 'spec' / 'code')`	`matches == []`, `classify is None`
`openFoo`	`('neutral-product', None)`, also `project-leg/assembly`	`matches == []`, `classify is None`
`Thing-Install`	`('install', None)`	`matches == []`, `classify is None`
`user-wip`	`('workspace', None)`	`matches == []`, `classify is None`
`InkxRouter`, `MedxScribe`	`('project-leg', 'assembly')`, with `descendant_referents` `('openRouter', 'openxRouter')` and `('openScribe', 'openxScribe')`	`matches == []`, `classify is None`, and `descendant_referents == ()`

Atlas-Spec (a capital S) is None with and without the line feed, on both. topic_pattern.match('xf-project-atlas\n') is a match on main; no tool reads topic_pattern (see Limits).

The gate (validate-repository-naming.py, run by the root template at Makefile:48 and validate.yml:310), measured by the new rows run on main's code:

--project project.yaml over a scaffolded root whose leg repository: is written "testorg/Atlas\n" (and likewise -spec and -code): exit 0, the table naming the leg and its family on two lines (' Atlas', then ' project-leg/assembly').
The same name as an argument (Atlas\n, Atlas-spec\n, Atlas-code\n): exit 0, the same split. A name read from stdin is stripped (line.strip()), and is untouched.
The issue also measured that validate-manifest.py has refused such a leg since 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 (manifest-leg-repository, QUALIFIED_RE ending in \Z), so the root's gate as a whole already failed; the naming step passed. I did not re-run that measurement.

What changed

scripts/repo_shape.py: the four sites read fullmatch. NamingPolicy is the only code that matches a name against the data's patterns (grep of the tracked *.py; scaffold-project.py:1535 prints the neutral-product pattern's text in a remedy and matches nothing). A one-line note beside topic_pattern says that a reader must use fullmatch, because $ takes a final line feed.
contracts/repository-naming.yaml: a paragraph at the end of the header, "A PATTERN IS READ AGAINST THE WHOLE NAME", for a fork's pattern. A pattern with no $ was read as a prefix under .match and is read as the whole name now, so a fork writes the pattern for the entire name. I took this of the two ways the issue offers (the other is to keep .match and refuse a name that holds a line feed at the entry of matches, classify and descendant_referents) as the shorter. Every shipped pattern begins with ^ and ends with $, and the measured sweep below shows fullmatch changes no answer except a trailing line feed.
tests/test_naming_policy_line_break.py (new, 77 rows):
- the names are read from the data (families[].examples, roles[].examples, the descendant family's claim_without_referent_examples), so a family or a role added to the contract is covered without a line here; a declared_only family is asked under its own id;
- a baseline row per name (the name without the line feed is matched as it was) and, per name with the line feed, matches == [], classify is None and descendant_referents == ();
- rows over a fork's policy (NamingPolicy built from the shipped data with ONE pattern written without $) for each of the four sites;
- the gate rows: validate-repository-naming.py --project over a scaffolded root for each of the three roles, a double-quoted baseline, and an argv name for each, with its baseline.
Nothing else. No refusal code is added, changed or renamed, no sentence of a refusal changes, no command or flag is added, so AGENTS.md, README.md, docs/ and the handbook mirror stay as they are. Before the change I grepped tests/, README.md, docs/ (the handbook), AGENTS.md and contracts/ for text that pins the header's wording or a prefix reading of a policy pattern: nothing reads the contract's comments, and the only patterns a test writes into a policy are tests/test_naming_policy.py:2498's fork form ^~[a-z]+$, which has a $, and tests/test_doctor_validator_verdicts.py:324's corrupted (^open(?:, which does not compile.
What pins scripts/repo_shape.py. This repository holds no copy of it and no digest of it. A project's copy is pinned by the shape pin's tree digest (sorted-ls-tree-r-v1, computed over the shape's tree at the pinned commit), and update-shape.py reaches it as an upstream-changed scripts/repo_shape.py row; contracts/repository-naming.yaml is in the copy lists of scripts/shape_materialize.py (lines 444 and 652) and reaches a project the same way. fullmatch needs no import, so no pinned script meets a name it lacks; tests/test_pinned_copy_text.py passes (hygiene five below).

The rows failing on main

tests/test_naming_policy_line_break.py run on main's production code (n1202-mut, a detached tree at be21751c3 with the new file copied in; sha256 of the file equal to the branch's): 40 failed, 37 passed (run-new-main-be21751c3.log). The 37 that pass are the baselines (21 family names, 6 role names, 6 descendant names, the double-quoted leg, 3 argv names). The 40 that fail, and why:

Row (function)	Failing on main	Because
`test_a_name_that_ends_in_a_line_feed_matches_no_family`	24: `neutral-product-` `openxFactory`, `openXwallet`, `openChart`, `openDox`, `openXdox`; `install-` `xFactory-Hermes-Install`, `Hermes-Install`, `Keycloak-Install`; `domain-descendant-` `MedxChart`, `LedgerxWallet`, `codexFactory`, `MedxGlass`, `codexDox`; `project-leg-` `Atlas`, `Atlas-spec`, `Atlas-code`, `Northwind`, `Northwind-spec`, `Northwind-code`; `family-` `InkRouter`, `Northwind`; `workspace-` `brett-wip`, `scott-wip`, `a1-wip`	`matches(name + "\n")` was the answer for the name
`test_a_descendant_name_that_ends_in_a_line_feed_has_no_referent`	6: `MedxChart`, `LedgerxWallet`, `codexFactory`, `MedxGlass`, `codexDox`, `MedxScribe`	`descendant_referents` gave `openChart` and `openxChart`, and so on
`test_a_forks_descendant_pattern_with_no_dollar_is_read_as_the_whole_name`	2: `family-pattern`, `split-pattern`	a prefix pattern took `MedxScribe`
`test_a_forks_leg_pattern_with_no_dollar_is_read_as_the_whole_name`	2: `role-pattern`, `family-pattern`	a prefix pattern took `Atlas_x`
`test_the_naming_gate_refuses_a_leg_repository_that_ends_in_a_line_feed`	3: `assembly`, `spec`, `code`	exit 0 where 1 is expected
`test_the_naming_gate_refuses_an_argv_name_that_ends_in_a_line_feed`	3: `Atlas`, `Atlas-spec`, `Atlas-code`	exit 0 where 1 is expected

On the branch all 77 pass. The refusal is, for each: exit 1 and the line FINDING naming-unclassified: <name>.

No name without a line feed changes its answer (measured, not argued): matches, classify and descendant_referents over 66,688 names (every string of up to 5 characters over A a x X - 1 _, a space and a line feed, and 11 stems with 12 suffixes, with and without open) under 5 declared roles (none, assembly, spec, code, family), on main and on the branch: 333,440 rows, 3,225 differ and every one is a name that ends in a line feed; none that does not; and no name that ends in a line feed matches anything on the branch (probe_inproc.py and sweep.py in lane 5's work/n1202/). tests/test_naming_policy.py passes unchanged.

Mutants

Each applied alone to scripts/repo_shape.py on the branch's copy in n1202-mut, run against the new file, and restored with a sha256 check (restored True, and at the end the file is main's bytes again, blob 00759f1e6):

Mutant	Result	Killed by
M1: `descendant_referents`, the family's `_re.fullmatch` put back to `.match`	1 failed, 76 passed	`test_a_forks_descendant_pattern...[family-pattern]`
M2: `descendant_referents`, `_split_re.fullmatch` put back to `.match`	1 failed, 76 passed	`test_a_forks_descendant_pattern...[split-pattern]`
M3: `matches`, a role's `_re.fullmatch` put back to `.match`	25 failed, 52 passed	18 of the every-name rows, `test_a_forks_leg_pattern...[role-pattern]`, and the 3 leg and 3 argv gate rows
M4: `matches`, a family's `_re.fullmatch` put back to `.match`	31 failed, 46 passed	all 24 every-name rows, `test_a_forks_leg_pattern...[family-pattern]`, and the 3 leg and 3 argv gate rows
M5: M1 and M2 together	14 failed, 63 passed	6 every-name rows, the 6 descendant rows, both fork descendant rows
M6: all four sites	40 failed, 37 passed	the same 40 as on main

No survivor. M1 and M2 are killed only by the fork rows, and that is the shape of the code, not a gap in the rows: the two sites of descendant_referents stand in series, and the shipped split_pattern takes the family pattern's language, so on the shipped data either alone makes the other redundant (M5 is what the shipped-data rows kill). A fork's policy whose one pattern has no $ is what asks each by itself; it is also the case the contract's new paragraph is written for.

Runs

Python 3.12.3 (CI runs 3.14; the rows assert lists, tuples and finding lines, nothing that depends on a repr or an ast.dump), through lane 5's run10.sh (gh stub first on PATH, bogus token, empty GH_CONFIG_DIR, blank git identity, GIT_SSH_COMMAND=false, GIT_CONFIG_GLOBAL=/dev/null), git-filter-repo 2.47.0 (CI's pin; git filter-repo --version prints a40bce548d2c), at head 09697e9c5, dirty=0:

Run	git	Result
the new file, `tests/test_value_pattern_line_break.py` (the sibling it imports two helpers from), `tests/test_naming_policy.py`, and the hygiene five: `test_path_guard_hygiene`, `test_repo_hygiene`, `test_agents_md_codes`, `test_cli_reference`, `test_pinned_copy_text`	2.43.0	751 passed in 131.18s
the same files	2.47.1 (`tools/git247/bin/git`, built from kernel.org's tarball)	751 passed in 120.02s
the neighbours that name `NamingPolicy` or run the naming gate: `test_manifest_validator`, `test_scaffold_referent_chain_links`, `test_chain_pin_sources`, `test_github_name_limits`, `test_adopt_manifest_values`, `test_scaffold_manifest_values`, `test_reader_refusals_named`, `test_pin_source_env_collisions`, `test_family_init_values`	2.43.0	750 passed, 19 skipped in 383.95s (all 19 are template skips in `tests/test_family_init_values.py`: 16 at `:229`, "this template quotes the placeholder", and 3 at `:301`, "this template quotes `id:`", as rv1266 printed them with `-rs`)

Before the rebase onto be21751c3 the same files (without the sibling) passed at the then head f552017f1: 720 on 2.43.0 and 720 on 2.47.1. Not run: the full suite, a real Windows host and macOS, and test_shape_doctor.py (it reads the naming validator through the doctor and no name in it ends in a line feed).

Merges

Read again by rv1266 at 2026-11-08T17:10Z: main 46fe6ac19 gives tree 251537e4, clean, and main 12122952b gives c46cbe55, clean. The new file, test_value_pattern_line_break, test_naming_policy and the hygiene five passed on that predicted merge (751, git 2.43.0), since 12122952b changed tests/test_pinned_copy_text.py.

git merge-tree --write-tree <other> 09697e9c5 (git 2.43.0) at 2026-11-08T16:29Z, against main and every open PR that touches one of this PR's files (50 open):

Against	Head	Tree
main	`be21751c3`	`836cb7fec` (clean)
#1249 (`scripts/repo_shape.py`)	`3146c3ca5`	`f41cf1dca` (clean)
#1171 (`scripts/repo_shape.py`)	`8162b874e`	`29eee887f` (clean)
#1236 (`scripts/validate-repository-naming.py`, which the new rows run)	`352dcbc96`	`64990151d` (clean)

The later lander takes the refresh.

Limits

topic_pattern has no row. Nothing outside tests/test_naming_policy.py:422 reads it at main, and the data's topic.pattern and topic.id_pattern are read by no tool, so a row cannot fail on main for it. The note beside it, and the contract's paragraph, say a reader must use fullmatch.
A project's own copy reaches this through update-shape.py. Until a project takes the new scripts/repo_shape.py its validate-repository-naming.py and validate-manifest.py run the old .match. The root's validate-manifest.py already refuses a leg repository: that ends in a line feed since 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.
The refusal prints the name raw. naming-unclassified names the name as typed, so Atlas\n ends the finding's first line and the existing text ("These names match none of the families ...") follows after a blank line; a name holding an ESC prints raw there on main too. Not changed: the issue asks that the name be refused by name, and no wording or code moves.
A fork's data that relied on a prefix reading. A fork's pattern with no $ is read as the whole name now, as the contract's header says. The shipped data has none, and no test writes one.
Not touched, read at their call sites: scaffold-project.py, setup-project.py, family.py and shape_materialize.py take a name from argv, which checked_value refuses first; validate-family.py strips it.

Closes #1202

Lane: openreposhape-5 (openRepoShape-5)

🤖 Generated with Claude Code

Summary by Sourcery

Ensure repository naming policy evaluation uses whole-name matching so names ending in a line feed are rejected consistently.

Bug Fixes:

Reject repository names ending in a line feed as unclassified instead of treating them as their newline-free equivalent.

Enhancements:

Make all repository naming policy patterns match the entire name, including custom patterns without an explicit end anchor.
Document whole-name pattern matching and the corresponding requirement for future readers of topic patterns.

Documentation:

Document that repository naming policy patterns are evaluated against the whole name.

Tests:

Add coverage for trailing-line-feed handling across naming classification, descendant referents, custom patterns, manifest repositories, and command-line names.

…t four sites, so a name that ends in a line feed matches no family and `validate-repository-naming.py` refuses it, and `repository-naming.yaml` says a pattern is read whole (#1202) Every pattern of `contracts/repository-naming.yaml` is `^...$`, and `NamingPolicy` read a name against it with `.match` in `descendant_referents` (the family's pattern, then its `split_pattern`) and in `matches` (a role's pattern, then a family's). In Python `$` also matches just before a trailing line feed, so `Atlas\n` was classified as `Atlas`, and `validate-repository-naming.py --project project.yaml` over a leg `repository: "testorg/Atlas\n"`, or over the argument `$'Atlas\n'`, exited 0. The four sites ask `fullmatch` now. Every shipped pattern begins with `^` and ends with `$`, so no name without a line feed changes its answer. A fork's pattern with no `$` read as a prefix of the name under `.match`; the contract's header now says a pattern is read against the whole name, which is the shorter of the two ways the issue offers. `topic_pattern` has no reader outside `tests/test_naming_policy.py`, so a note beside it says a reader must use `fullmatch`, and no row can fail on main for it. `tests/test_naming_policy_line_break.py` reads its names from the data, so a family or a role added to the contract is covered, and holds a fork's policy whose one pattern has no `$` for each site: the two sites of `descendant_referents` stand in series, and the shipped `split_pattern` takes the family pattern's language, so on the shipped data either alone makes the other redundant. Closes #1202 Lane: openreposhape-5 (openRepoShape-5) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

sourcery-ai · 2026-11-08T16:42:07Z

Reviewer's Guide

Implementation

Switch all four NamingPolicy checks to fullmatch so names must match the entire pattern, and document the same requirement beside topic_pattern.

scripts/repo_shape.py L2231, L2268-2270, L2436-2439

Configuration

Clarify in the naming-policy contract that patterns are matched against the whole name, including patterns without a trailing dollar anchor.

contracts/repository-naming.yaml L119-126

Tests

Add regression coverage for trailing-line-feed names, whole-name matching of fork patterns, and rejection by the naming validator.

tests/test_naming_policy_line_break.py L1-242

Tips and commands

Interacting with Sourcery

Trigger a new review: Comment @sourcery-ai review on the pull request.
Continue discussions: Reply directly to Sourcery's review comments.
Generate a GitHub issue from a review comment: Ask Sourcery to create an
issue from a review comment by replying to it. You can also reply to a
review comment with @sourcery-ai issue to create an issue from it.
Generate a pull request title: Write @sourcery-ai anywhere in the pull
request title to generate a title at any time. You can also comment
@sourcery-ai title on the pull request to (re-)generate the title at any time.
Generate a pull request summary: Write @sourcery-ai summary anywhere in
the pull request body to generate a PR summary at any time exactly where you
want it. You can also comment @sourcery-ai summary on the pull request to
(re-)generate the summary at any time.
Generate reviewer's guide: Comment @sourcery-ai guide on the pull
request to (re-)generate the reviewer's guide at any time.
Resolve all Sourcery comments: Comment @sourcery-ai resolve on the
pull request to resolve all Sourcery comments. Useful if you've already
addressed all the comments and don't want to see them anymore.
Dismiss all Sourcery reviews: Comment @sourcery-ai dismiss on the pull
request to dismiss all existing Sourcery reviews. Especially useful if you
want to start fresh with a new review - don't forget to comment
@sourcery-ai review to trigger a new review!

Customizing Your Experience

Access your dashboard to:

Enable or disable review features such as the Sourcery-generated pull request
summary, the reviewer's guide, and others.
Change the review language.
Add, remove or edit custom review instructions.
Adjust other review settings.

Getting Help

Contact our support team for questions or feedback.
Visit our documentation for detailed guides and information.
Keep in touch with the Sourcery team by following us on X/Twitter, LinkedIn or GitHub.

sonarqubecloud · 2026-11-08T16:43:13Z

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

brettheap · 2026-11-08T17:11:38Z

(lane openRepoShape-5, 2026-11-08T17:11:37Z) Independent review of #1266 (#1202) for lane openRepoShape-5, by rv1266 (opus), at 09697e9: READY

There are no MUST or SHOULD findings. The four sites are the only change in behaviour, and only for a name that ends in a line feed. Every row, mutant and gate the body claims reproduces. There are 7 NITs, none blocking. CI's tests job (the full suite on 3.14) was still pending at 17:10Z, so the landing needs it green.

Scope: one commit, 09697e9c52f35c9aadbafdebb88e47950105f474, on main be21751c3. It is 3 files, +255/-4. The head equals refs/heads/fix/1202-naming-fullmatch and refs/pull/1266/head by ls-remote at 16:45Z and again at 17:10Z. I reviewed in detached worktrees /workspace/projects/openRepoShape-worktrees/rv1266-review and rv1266-mut. Both are clean at the head (porcelain 0).

What held (measured)

The four sites, and nothing else.
- git grep finds no other reader of a policy pattern. The only readers of _re, _split_re and topic_pattern are repo_shape.py:2268, :2270, :2436 and :2439, plus the compile lines. Elsewhere, scaffold-project.py:1535 and validate-repository-naming.py:795/919 only print a pattern's text.
- probes/patterns.py checked all 17 compiled patterns (6 families, 3 roles, the split patterns, topic.pattern and id_pattern):
  - every one begins with ^ and ends with $;
  - none is a top-level alternation (the positive control ^a|b$ parses as one BRANCH);
  - no character class admits \n.
- So fullmatch and .match can differ only on a name that ends in a line feed.
My own sweep confirms it. probes/sweep2.py loads main's and the head's repo_shape.py in one process. It compares matches, classify (with and without declared pins) and descendant_referents.
- Inputs: every string of up to 5 characters over B b x X - 0 . \n \r \u2028, plus 14 stems x 19 suffixes, with and without a leading or trailing LF. That is 112,091 names under 6 roles.
- Result: 672,546 rows; 4,200 differ, every one a name that ends in a line feed, and 0 otherwise. At the head, no name that ends in a line feed matches anything.
The writer's sweep reproduces exactly. I re-ran sweep.py against main's bytes and the head's: 333,440 rows, 3,225 differ, all ending in a line feed, 0 otherwise. That is the body's number.
The rows.
- At the head, 77 of 77 pass (inside gates A and B).
- I swapped main's scripts/repo_shape.py (blob 00759f1e6, sha256 623af121...) into rv1266-mut. The new file gave 40 failed, 37 passed (logs/rows-on-main.log). Afterwards I restored the head's copy and checked it: sha256 009c249e..., restored=True.
- The 40 split as 24 every-name + 6 descendant + 2 + 2 fork + 3 leg-gate + 3 argv-gate. The 24 names are exactly the body's list.
- All 6 gate rows fail as assert 0 == 1, printing ' Atlas' and then project-leg/assembly on the next line.
- Table-driven: the rows read families[].examples, roles[].examples and claim_without_referent_examples from the data, and a declared_only family is asked under its own id. test_naming_policy.py's FAMILY_IDS is also read from the data and indexes family["examples"], so a family added without examples already fails there.
- The gate rows: they exit 1 with exactly FINDING naming-unclassified: <name> for assembly, spec and code over --project, and for each argv name.

Mutants. Each was applied alone in rv1266-mut, run against the new file, and restored with a sha256 check (restored True after every one; porcelain 0 at the end). All are killed:

Mutant	Result	Killed by
M1 `descendant_referents` `_re` back to `.match`	1 failed / 76 passed	fork `[family-pattern]`
M2 `_split_re` back to `.match`	1 failed	fork `[split-pattern]`
M3 role `_re` back to `.match`	25 failed	18 every-name rows, leg fork `[role-pattern]`, 3 leg and 3 argv gate rows
M4 family `_re` in `matches` back to `.match`	31 failed	24 every-name rows, leg fork `[family-pattern]`, the 6 gate rows
M5 = M1+M2	14 failed	6 every-name, 6 descendant, 2 fork rows
M6 = all four sites	40 failed	the same 40 as main
X1 (mine) role site `fullmatch(name.rstrip("\n"))`	24 failed	every-name and gate rows
X2 (mine) family site in `matches`, the same strip	30 failed	every-name and gate rows
X3 (mine) both descendant sites strip	12 failed	6 descendant + 6 every-name rows
X4 (mine) `validate-repository-naming.py:1097` strips the leg name	3 failed	the leg gate rows
X5 (mine) `:1465` strips the argv name	3 failed	the argv gate rows
X6 (mine) M6 plus an LF guard at `classify`'s entry	34 failed	every-name rows (`matches`, `descendant_referents`), descendant and fork rows
X7 (mine) `_split_re.search`	1 failed	fork `[split-pattern]`

M1 and M2 are killed only by the fork rows. That is the code's shape (the two sites stand in series), as the body says. X3 shows the shipped-data rows do hold the descendant path.

The contract header.
- The contract reaches a project verbatim through shape_materialize.py's COPIED_FROM_SHAPE (line 444), and a family root through FAMILY_COPIED_FROM_SHAPE (line 652).
- Main's and the head's contracts load to equal data (data equal: True).
- Nothing pins its bytes, so the memory note's risk (a byte change breaking another file's regex) does not apply here. The rewrite targets that tests use still count 1 each at the head:
  - kind: repository-naming-policy\n (test_shape_doctor.py:1651, test_doctor_next_is_a_command.py:535);
  - pattern: "\^open\(\?: and - id: assembly(?=\r?\n) (test_doctor_validator_verdicts.py POLICY_EDITS, which asserts count 1).
- The header adds one textual split_pattern: (1 to 2), but no test or tool reads that text; YAML reads the key.
- No doc, handbook mirror or digest quotes the header, and no test holds a 64-hex digest of it. The new lines are ASCII.
No legitimate caller passes such a name.
- git_out strips stdout, and the advisory's other name is a directory name.
- adopt-project.py, scaffold, setup and family.py take names from argv or a plan, which checked_value refuses first.
- validate-family.py:615 strips.
- validate-manifest.py returns at manifest-leg-repository (:1139-1143) before it classifies, so none of its output changes. test_value_pattern_line_break.py[leg-repository], which asserts exactly one finding, passes.
- No other test feeds a line-feed name to the policy, the gate, or a block scalar (grep).
The body's claims and counts hold.
- +255/-4; 77 rows; 40/37; M1-M6 counts; the sweep numbers; 751 passed on each git; 750 passed and 19 skipped for the neighbours; the merge trees.
- Its citations hold: test_naming_policy.py:422 and :2498, test_doctor_validator_verdicts.py:324, scaffold-project.py:1535 (CHAIN_LINK_FORM = "neutral-product"), and shape_materialize.py 444/652.
- Commit: identity Brett Heap <service@opensoft.dev>; subject 249 characters; one Closes #1202; the Lane line and Co-Authored-By: Claude Sonnet 5.5; ASCII; no assert joins conditions with and or or.

NITs

The body's Runs row guesses the 19 skips wrong. It says they "need another OS or a host" and that "the reasons were not printed". My runs print them (-rs), and they are all template skips in tests/test_family_init_values.py:
- :229, 16 rows: "this template quotes the placeholder: the run is fine, and test_a_quoted_value_is_held_as_typed runs it";
- :301, 3 rows: "this template quotes id:: the derived id reads back".
The name is printed raw, and Limits covers only half of it. Probe at the head: validate-repository-naming.py $'Atlas\n' exits 1.
- stdout is ' Atlas', then ' NO FAMILY' on the next line. So the table split the issue names as its second wrong answer is still there, now reading NO FAMILY.
- stderr is FINDING naming-unclassified: Atlas and then a blank line, which looks exactly like a valid Atlas.
- It predates this PR: the validator is untouched, and $'At\nlas' prints the same way.
- repo_shape.spelled() (templates/assembly-root/scripts/bootstrap.py and templates/family-root/scripts/bootstrap.py: main prints the manifest's name with a plain print before any step runs, so name: "a\ud83db" (a lone surrogate) ends make bootstrap in a UnicodeEncodeError traceback, exit 1 (make: 2), and the validator that names the problem is never reached; a valid name the stdout cannot encode ends the same way #391) exists for exactly this, and validate-repository-naming.py:1357 and the table don't use it.
- Suggest a one-line follow-up issue. Changing it would change a refusal's text, which is out of this issue's scope.
Line length. The new topic_pattern comment at repo_shape.py:2231 is 108 characters, the only line in the file over 100 (main's longest is 93). No linter enforces a limit. Optional wrap.
The Merges section is stale. main is now 46fe6ac19, and it is clean (table below).
No milestone. The PR has none, and repo_shape.py: NamingPolicy reads a name with .match against patterns that end in $, so a name that ends in a line feed is classified as the name without it, and validate-repository-naming.py exits 0 over such a leg repository (#1166 follow-up) #1202 is on v1.2.1. The coordinator should set it before landing.
One unclear cell. In "What held on main", the "the same" cells for openFoo, Thing-Install and user-wip read as "the same as main". They mean the first row's matches == [], classify is None. Wording only.
CI tests was pending at 17:10Z (SonarCloud, package and tests-git-bash pass). Watch it before landing.

Gates

All ran through run10.sh (gh stub, bogus token, blank identity, GIT_SSH_COMMAND=false, GIT_CONFIG_GLOBAL=/dev/null), on Python 3.12.3 with git-filter-repo 2.47.0 (a40bce548d2c), each with rc=0. Gates A-D ran at the head with dirty=0.

Gate	Git	Files	Result	Log
A	2.43.0	new file, `test_value_pattern_line_break`, `test_naming_policy`, hygiene five	751 passed, 141.06s, 0 skips	`logs/gateA-core-243.log`
B	2.47.1 (`tools/git247/bin/git`)	same	751 passed, 120.31s, 0 skips	`logs/gateB-core-2471.log`
C	2.43.0	the writer's 9 neighbours	750 passed, 19 skipped, 417.13s	`logs/gateC-neigh-243.log`
D	2.47.1	the 9 neighbours	750 passed, 19 skipped, 419.77s	`logs/gateD-neigh-2471.log`
E	2.43.0	gate A's set, on the predicted merge `c46cbe55` (main `12122952b` + this PR)	751 passed, 125.56s	`logs/gateE-merge-core-243.log`

Every skip in C and D is one of the two template skips in NIT 1. I ran gate E because 12122952b changed tests/test_pinned_copy_text.py. I did not run the full suite, test_shape_doctor.py, Windows or macOS.

Merge-trees

git merge-tree --write-tree <other> 09697e9c5 (git 2.43.0), with heads fetched fresh at 16:51Z and 17:10Z. Of 51 open PRs, only #1249 and #1171 touch repo_shape.py, and #1236 touches the validator the new rows run.

Against	Head	Tree
main	`46fe6ac19` (17:10Z; it touches `shape_advisory.py`, `shape-doctor.py`, `test_doctor_unsearchable_links.py`)	`251537e4` clean
main	`12122952b` (16:51Z)	`c46cbe55` clean
#1249	`3146c3ca5`	`f41cf1dc` clean
#1171	`8162b874e`	`29eee887` clean
#1236	`352dcbc96`	`64990151` clean

The trees for #1249, #1171 and #1236 equal the writer's.

Everything is in /workspace/projects/.lane-worktrees/openRepoShape-5/work/rv1266/: STATE.md, logs/, merge-trees.txt, rows-on-main-failed.txt, and in probes/ the scripts patterns.py, sweep2.py, mutate.py, run-mutants.sh and contract_counts.py. I made no commits, pushes or GitHub writes.

Lane: openreposhape-5 (openRepoShape-5)

brettheap added this to the v1.2.1 milestone Nov 8, 2026

Navigation Menu

Repository navigation

scripts/repo_shape.py: `NamingPolicy` reads a name with `fullmatch` at four sites, so a name that ends in a line feed matches no family and `validate-repository-naming.py` refuses it, and `repository-naming.yaml` says a pattern is read whole (#1202) - #1266

brettheap commented Nov 8, 2026 •

edited

Loading

Uh oh!

sourcery-ai Bot commented Nov 8, 2026

Interacting with Sourcery

Customizing Your Experience

Getting Help

Uh oh!

sonarqubecloud Bot commented Nov 8, 2026

Uh oh!

brettheap commented Nov 8, 2026

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

brettheap commented Nov 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Goal

What held on main

What changed

The rows failing on main

Mutants

Runs

Merges

Limits

Summary by Sourcery

Uh oh!

sourcery-ai Bot commented Nov 8, 2026

Reviewer's Guide

Implementation

Configuration

Tests

Interacting with Sourcery

Customizing Your Experience

Getting Help

Uh oh!

sonarqubecloud Bot commented Nov 8, 2026

Quality Gate passed

Uh oh!

brettheap commented Nov 8, 2026

What held (measured)

NITs

Gates

Merge-trees

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

brettheap commented Nov 8, 2026 •

edited

Loading