[brettheap]

Lane: openreposhape-2 (openRepoShape-2)
Claim: #608 (comment)

Round of 2026-10-16: main 4cf5fb07 and #570

The branch now merges main 4cf5fb07 (be44452, clean) and carries one fix on it, d0a65ba7, then 400965b3, a docstring-only correction from the re-verify (rv623g, MERGEABLE WITH NITS: the helper's docstring now says the validator prints its --link-source hint once per holder it cannot verify). #570 (396c0ff4, on main) makes the scaffold take --pin-source as text and refuse an EXACTLY empty value as pin-source-no-path, before anything exists, so Path('') is no longer what an empty value is read as. That left a semantic conflict with this PR's step (4): the merge was clean and test_an_empty_pin_source_is_the_plans_to_refuse failed through both entry points, because step (4) passed the naming validator --link-source=. for an empty --pin-source, and the validator printed WARNING link-source-unreadable: . (--link-source) could not answer for openXdox about a flag the person never typed and a folder the plan never reads, ahead of the plan's pin-source-no-path. An exactly empty last --pin-source now goes to the validator as no --link-source at all, by the scaffold's own test, not str(typed), so step (4) reports the chain exactly as it does when no --pin-source is given at all, and step (5)'s plan refuses the value by name with exit 2. Every other value goes as before. A blank one is a path, so step (4) reports it as an unreadable --link-source, as it does any source it cannot read, and the plan refuses it as pin-source-unreadable (rv623g). Changed: setup-project.py (the two docstrings and a few lines: the . branch goes and names_a_source gates the --link-source), tests/conftest.py (the helper and its docstring), tests/test_setup_project_py.py (three rows of the declarations table). Nothing else on the merged tree assumed an empty --pin-source is .: README.md, AGENTS.md, setup.sh, docs/ say nothing of it, --help is unchanged, and docs/cli.md already says what #570 says.

One thing the helper does not assert, and why: the validator's own [declared-unverified] WARNING for a chain it was given no source for ends or --link-source openXdox=<path> would let it be read. It is the validator's hint, printed once for each holder it cannot verify in any chain with no source (a run with no --pin-source at all prints it too), so the helper lets that one spelling through and asserts that no other --link-source is in the output, and that [declared-unverified] is in step (4), which says no source was passed.

What changed

All four lines of #608, on main bcfa3645 (#569 landed). Every one reproduces there. The branch also merges main 55faad26 (81821e7), which brings #576's link-source-no-path, and carries rv608's fix round on this PR (items 1 to 5 of its list; item 6 is a #459 line). The change is in setup-project.py; no template, contract or pinned file moved, and --help is unchanged, so there is no re-pin and docs/cli.md is as it was (render-cli-reference.py --check passes inside test_cli_reference.py).

• L2-26, step (3) refuses a parent nobody can write in and a drive that is not there. Step (3) asked only that what stands above the clone is a folder. A parent at mode 555 passed, all three repositories were created, and step (7) then refused could not be made a folder (Permission denied). refuse_a_parent_nobody_can_write_in now makes a folder in the nearest existing parent and takes it away again, which is what step (7) does and the one answer that holds for a read-only disk, an ACL, a quota and root (os.access says what the mode bits say, and on Windows nothing about a folder). It refuses by name with the system's reason, before the plan, and the error code is the one every step (3) refusal has (2). Only the making refuses: taking the probe away is best-effort (on Windows an indexer or antivirus can hold a handle on a new folder), so an rmdir that fails no longer calls a folder you can write in one you cannot; the folder it leaves is named in a [??] line. nearest_entry returns None only where not even the top of the path is there, a Windows drive that does not exist; refuse_a_file_above left that "to the clone", and it is now refused naming the drive (drive_of). A file above the clone is still refuse_a_file_above's, said first.
• L2-25, step (4) verifies a chain from --pin-source. The scaffold reads a chain's links from the tree --pin-source names; step (4) was given --pins and --referent-chain but no --link-source, so a chain the plan verified showed [declared-unverified] and a WARNING there, and a chain the tree broke showed no break. The last --pin-source now goes to the naming validator as a bare --link-source. It goes only with a chain of two links at least: the scaffold reads the tree for the links alone, a chain of one link has no holder whose declaration a source could answer, and the validator prints a NOTE --link-source ...: not used for a source no link used, which the plan never prints. An EXACTLY EMPTY --pin-source goes as no --link-source at all: since scaffold-project.py and adopt-project.py: a pin records digest_source: local-clone and never the --pin-source path, so no committed pin file holds a local directory, and a --pin-source nothing reads is pin-source-unused (#540) #570 the scaffold refuses it as pin-source-no-path before anything exists, so step (4) leaves it to the plan, which refuses it by name with exit 2. Since validate-repository-naming.py, repo_shape.py, validate-pins.py: a bare --link-source is the first chain link's tree alone, and a source that names no path is refused (#441, #442) #576 the validator refuses --link-source= as link-source-no-path, which named a flag the person never typed and ended step (4) with exit 1, and . (what this PR passed before scaffold-project.py and adopt-project.py: a pin records digest_source: local-clone and never the --pin-source path, so no committed pin file holds a local directory, and a --pin-source nothing reads is pin-source-unused (#540) #570, when the scaffold read Path('') as .) made it warn about --link-source and a folder the plan never reads. A BLANK value is a path, and goes as typed. PASSED_FLAGS gets --pin-source as --pin-s (held to the scaffold's --help by the existing drift test). A value that starts <word>= is spelled ./<word>=, because the validator reads openXdox=layer as layer for the link openXdox where the scaffold reads one path.
• L2-27, passed_values leaves a dash word out. --pin --referent-chain x gave --pin the value --referent-chain, against the docstring, and step (4) was handed --pins=--referent-chain. A following word that starts with - is not a value; --pin=-x still is.
• L2-37, step (5) says "refused" for a plan it printed. Since family.py and scaffold-project.py: --dry-run asks the real run's own question about existing remotes, so family-remote-exists and scaffold-remote-exists come at the plan, and a missing gh is gh-not-found, not a traceback (#537) #573 the dry run prints the plan and then refuses an existing remote, so "the plan could not be produced (see above)" sat under the plan. After a refusal setup now says "the plan was refused (see above)." when the plan printed, and keeps the old words when it did not. When the second, captured run exits 0, the refusal did not repeat and the person never saw that run, so setup says that the scaffold stopped, that it did not stop when asked again, and to run setup again.

How step (5) knows the plan printed

The scaffold's output goes straight to the terminal, so setup cannot read it. After a nonzero exit, why_the_plan_did_not_go_through asks the same --dry-run once more, from the shape checkout, with its output captured: exit 0 is "did not repeat"; otherwise the plan's own first line (project ) at the start of a line says it printed. The dry run reads and creates nothing, and a plan that went well is run once, uncaptured. The alternative, reading every run's plan through a pipe, would put the plan of every run behind the child's buffering and output encoding (and Windows is unverifiable here, since CI is Linux-only since #597) to word one refusal. A child killed by a signal is not asked again. The cost is a second gh read on a refusal; say so if you want the pipe instead. Not covered: two runs that both refuse but disagree on whether the plan printed (a gh read that fails once, before the plan) still get the second run's words; only a tee of the first run would know.

Tests

The new end-to-end tests are shared helpers in tests/conftest.py, each with a one-call wrapper in both tests/test_setup_sh.py and tests/test_setup_project_py.py (Sonar's duplication gate): the unwritable parent, the chain verified and broken through step (4) (2 cases), the plan that never printed, the empty --pin-source and the chain of one link. The #537 helper assert_the_plan_refuses_a_remote_that_exists expects "the plan was refused" and the plan's first line above it. The rest is in-process and in test_setup_project_py.py only: the probe (the system's reason, the nearest folder, nothing left behind, a failing rmdir named and not refused), the missing drive (3 spellings), drive_of under ntpath, the passthrough readers (rows of the declarations table: the empty and blank --pin-source, the chains of one link and of blank links, a blank chain last and first, a\b=c and C:x=y; 7 of passed_values), and the plan wordings (printed, never printed in 5 spellings, a second run that exits 0, killed, run once; the second run's cwd). No assert is composite (the two S9073 findings are split).

Targeted, on the head 1e30fa7d:

• Run A, every tests/test_setup*.py (test_setup_by_hand_commands.py is green and unchanged), test_family_init*, test_windows_paths, test_openreposhape_command, test_dry_run_asks_remotes, test_cli_reference, test_repo_hygiene: 649 passed, 19 skipped, 0 failed.
• Run B, every tests/test_scaffold*.py: 634 passed, 0 failed.
• tests/test_link_source_chain_defaults.py, the test file that names link-source-no-path: 23 passed.
• The 96 setup tests selected for the new and changed paths again under an uncommitted pytest plugin that makes Path.write_text write \r\n (a probe test shows it active): 96 passed.

Targeted, on the merge be444527 of main 4cf5fb07 and the fix d0a65ba7 (run on the tree whose three changed files are the commit's, sha256 checked before and after):

• The files of run A, run B and the link-source set together, with test_pin_source_env_collisions.py and test_pin_source_unreadable.py (24 files): 1526 passed, 19 skipped, 0 failed, in 430 s.
• The 44 tests for the empty --pin-source, the declarations rows, step (4) and the chain of one link, through both entry points: 44 passed.
• Not re-run on this head: the CRLF emulation (the diff writes no file) and the 24 mutants.

Bite checks on the merge, each with the new tests kept (setup-project.py replaced for the run and restored byte for byte, sha256 equal):

• The merge's setup-project.py, which maps an empty value to .: the two end-to-end tests fail on the --link-source assertion, AssertionError: WARNING link-source-unreadable: . (--link-source) could not answer for openXdox: it holds no project.yaml, tests/conftest.py:1788; and three declarations rows fail (['--referent-chain=openXdox,openDox', '--link-source=.'] against ['--referent-chain=openXdox,openDox']).
• The same . mapping with the OLD helper (this round's tests/conftest.py and rows reverted to the merge's): both tests fail on the old assert "REFUSED pin-source-unreadable:" in result.stderr (tests/conftest.py:1780), the merge's own failure, so the old expectation is dead and the new helper is the one that bites.
• The fix without its empty-source rule (names_a_source = bool(sources), so --link-source= goes): both tests fail on assert "(5) the plan" in result.stdout, because step (4) ends the run with link-source-no-path and exit 1 before the plan.

Red on main (55faad2's setup-project.py, the final tests, quick selection of 95): 47 fail and 48 pass. Each failure is the defect itself: the mode-555 parent passes step (3) and step (7) refuses it after three remotes; no refusal for a missing drive (drive_of is absent); step (4) shows [declared-unverified] for both link trees; passed_values returns --referent-chain for --pin; the printed plan is called "could not be produced"; the second run is never made.

Red on the merge before the rv608 fixes (81821e7's setup-project.py): 13 of the new items, all of them rv608's: the third wording (2), the failing rmdir (1), the declarations rows for an empty --pin-source (3) and for a chain of one link (3), and the empty --pin-source and the chain of one link through both entry points (4). The empty --pin-source shows REFUSED link-source-no-path in step (4) with exit 1, and the chain of one link a NOTE --link-source ... not used. They exist only on the merge: main has no --link-source in step (4) at all.

Mutants (on 1e30fa7d, not re-run on d0a65ba7; the empty-source rule changed since, and the three bite checks above are its mutants): 24, each applied alone to setup-project.py; 23 killed. The one that survives, a child that exits above 128 also not being asked again, is not a case the scaffold produces (a signalled child is a negative returncode). Killed: step (3) never asking, the probe in the wrong folder, left behind, or skipped for a drive, the exit code, a constant reason, the drive never named, no link source, one without a chain, the first source instead of the last, no ./, ./ on every = (a colon or a backslash dropped from the class), an empty source not ., a blank one ., the link count's threshold moved either way, blank links counted, the first chain read for the count, an ambiguous --pin-source spelling, a dash word taken as a value, the = spelling refused, the third wording never given, given for a nonzero second run, or needing the plan too, the second ask uncaptured, run in the wrong folder or without --dry-run, a marker that is not the plan's or has lost its spaces, a refusal for a failing rmdir, a leftover probe that is silent or on stdout.

README follow-up

None required. README.md, AGENTS.md and the handbook say nothing about step (3)'s checks or the wordings. If a README PR wants it, one clause on README's step (3) line would do: "a clone path under a folder you cannot write in, or on a drive that is not there, is refused before anything is created."

Closes #608

🤖 Generated with Claude Code

Summary by Sourcery

Harden setup validation and make its dependency-chain checks, argument handling, and plan failure messages accurately reflect the scaffold's behavior.

Bug Fixes:

• Validate clone destinations before setup proceeds, including unwritable parents and nonexistent drives.
• Pass the configured pin source into naming validation so dependency chains are checked consistently, while handling empty sources and single-link chains correctly.
• Prevent flag-like arguments from being misinterpreted as values.
• Report accurately whether a printed plan was refused, never produced, or succeeded on a retry.

Enhancements:

• Add comprehensive unit and end-to-end coverage for setup validation, passthrough parsing, pin-source handling, and plan-refusal messaging across both entry points.

Tests:

• Expand shared setup tests and in-process coverage for filesystem probes, Windows path handling, chain validation, argument parsing, and plan outcome reporting.

[sourcery-ai]

Sorry @brettheap, you've used your own review budget of 250,000 diff characters for the last 7 days.

You can request another review in 1 day and 3 hours by commenting @sourcery-ai review. Upgrade to get a review now.

[sourcery-ai]

Reviewer's Guide

The PR updates setup-project.py to reject unwritable or missing-drive clone destinations before creation, make naming checks follow the scaffold's pin-source chain resolution, parse flag values consistently, and accurately describe whether a refused dry-run printed a plan; extensive regression tests cover both setup entry points and platform-sensitive cases.

Sequence diagram for pin-source chain verification

sequenceDiagram
    participant Setup as setup-project.py
    participant Scaffold as scaffold dry-run
    participant Validator as naming validator
    participant Tree as Pin source tree
    Setup->>Setup: passed_values(--pin-source)
    Setup->>Setup: naming_declarations()
    Setup->>Validator: --referent-chain and --link-source
    Validator->>Tree: Read chain links
    Tree-->>Validator: Link definitions
    Validator-->>Setup: Verified or broken chain result
    Setup->>Scaffold: Run --dry-run with passthrough flags
    Scaffold->>Tree: Read chain links from --pin-source

Loading

Flow diagram for setup-project preflight and plan refusal

flowchart TD
    A["setup-project step (3)"] --> B{clone path exists?}
    B -->|no| C["refuse_a_file_above"]
    C --> D["refuse_a_parent_nobody_can_write_in"]
    D --> E{nearest parent available?}
    E -->|no| F["die: drive is not there"]
    E -->|yes| G["create and remove write probe"]
    G -->|fails| H["die: parent cannot be written"]
    G -->|passes| I["continue to step (4)"]
    B -->|yes| I
    I --> J["naming_declarations"]
    J --> K["step (5) dry-run plan"]
    K --> L{dry-run refused?}
    L -->|no| M["continue"]
    L -->|yes| N["plan_was_printed"]
    N --> O{"project      found?"}
    O -->|yes| P["Refusal: the plan was refused"]
    O -->|no| Q["Refusal: the plan could not be produced"]

Loading

Flow diagram for passthrough flag value parsing

flowchart LR
    A["passed_values(passthrough, flag)"] --> B{"flag=value?"}
    B -->|yes| C["Use text after ="]
    B -->|no| D{"next word starts with -?"}
    D -->|yes| E["Do not consume next word"]
    D -->|no| F["Use next word"]
    C --> G["Return flag values"]
    E --> G
    F --> G

Loading

File-Level Changes

Change	Details	Files
Add preflight validation for clone destinations that cannot be created and improve missing-drive diagnostics.	Probe the nearest existing parent by creating and removing a temporary directory instead of relying on permission-bit checks. Reject paths whose nearest existing ancestor is absent, naming the drive when applicable. Preserve file-above-path handling and use step (3)'s refusal status and guidance.	setup-project.py tests/conftest.py tests/test_setup_project_py.py tests/test_setup_sh.py
Align step (4)'s naming-policy inputs with the scaffold's chain-link resolution.	Pass the final --pin-source as a bare --link-source when a non-empty referent chain uses links. Add the scaffold's short --pin-source spelling to passthrough parsing. Normalize source values beginning with a product-like name and '=' so the naming validator interprets them as paths.	setup-project.py tests/conftest.py tests/test_setup_project_py.py tests/test_setup_sh.py
Make passthrough value parsing match flag semantics when a following token begins with a dash.	Stop treating a following flag-like token as the preceding option's value. Continue accepting dash-prefixed values supplied with the --flag=value form. Expand declaration and value-parsing coverage for repeated flags, separators, ambiguous spellings, and option terminators.	setup-project.py tests/test_setup_project_py.py
Report accurately whether a refused dry-run printed its plan.	On nonzero dry-run exits, rerun the dry-run with captured output and detect the plan's first-line marker. Avoid the probe for successful or signal-terminated child processes. Distinguish 'the plan was refused' from 'the plan could not be produced' while retaining the original exit status.	setup-project.py tests/conftest.py tests/test_setup_project_py.py tests/test_setup_sh.py
Add shared and entry-point-specific regression coverage for all four fixes.	Reuse helpers across the Python and shell setup entry points to verify destination, chain, and plan behavior. Cover system errors, cleanup, missing drives, Windows path splitting, passthrough declarations, and printed versus unprinted plans. Update existing remote-refusal assertions to require the printed plan and revised wording.	tests/conftest.py tests/test_setup_project_py.py tests/test_setup_sh.py

Assessment against linked issues

Issue	Objective	Addressed	Explanation
#608	In setup step (3), refuse clone paths whose nearest existing parent is not writable, as well as paths on a missing drive, before creating repositories.	✅
#608	Make step (4) verify referent chains using the passthrough's --pin-source, and ensure passed_values does not treat a following flag as a value.	✅
#608	In step (5), distinguish a plan that was printed and then refused from one that could not be produced.	✅

Possibly linked issues

• setup-project.py: step (3) passes an unwritable parent or a missing drive, and step (4) shows a chain the plan verifies as [declared-unverified] #608: The PR implements all four issue items: writable-parent and drive checks, --link-source forwarding, dash-value parsing, and accurate printed-plan wording.

---

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.

[brettheap]

(lane openRepoShape-2, 2026-10-15T21:00:51Z)

This lane reviewed its own PR #623 at 8b0f14d, read-only in its own detached worktrees; nothing was authored.

MERGEABLE WITH NITS. No sequencing constraint. Findings 1 to 3 are small and worth fixing before landing.

Findings, at 8b0f14d:

1. P3, setup-project.py:2157-2161. plan_was_printed takes the wording from the second, captured dry run alone. If the runs disagree (a passing gh or network failure, or a remote made in between), a first run refused before any plan, then a second that printed one, gives "the plan was refused (see above)" with no plan above. Fix: when the second run exits 0, say the refusal did not recur and ask for a re-run.
2. P3, on the merge with validate-repository-naming.py, repo_shape.py, validate-pins.py: a bare --link-source is the first chain link's tree alone, and a source that names no path is refused (#441, #442) #576, setup-project.py:2049. -- --pin-source '' sends --link-source=. Main's validator refuses it (link-source-no-path, a flag nobody typed), and setup blames the naming policy (exit 1) where the plan refuses pin-source-unreadable (exit 2). Fix: pass . (the scaffold's Path('')) or omit it.
3. nit, setup-project.py:2045-2046. A one-link chain also gets --link-source, so step (4) prints NOTE --link-source <tree>: not used. Require two links.
4. nit, setup-project.py:1926-1934. A failed rmdir after a successful mkdtemp is reported as "a folder you cannot write in", and the probe folder is left behind.
5. nit, tests. Sonar S9073 at tests/test_setup_project_py.py:547-548. Mutants R1, R17 and R5/R6 survive.
6. Follow-up for Tracking: setup, scaffold, family and the shared helpers, creating projects and families #459, pre-existing on main. A checkout named after a chain link beside the shape checkout answers step (4) before --link-source. The scaffold reads no sibling.

The writer's edge cases: the dry-run refusal at step (3) is right; argparse's dash values are not this PR's; --work-dir and --local-remote-dir are #622's.

Tests: head A 598 passed, 19 skipped, B 634 passed; merge A 625 passed, 19 skipped, B 634 passed. Red on main: 32 of 71 selected fail. Mutants: 7 of 14 killed.

Merge: merge-tree with origin/main 77bb5f0 is clean.

Report: handoffs/xFactory/attachments/lane-openRepoShape-2/review623-8b0f14d/final-report.md in opensoft/brett-wip.

[brettheap]

(lane openRepoShape-2, 2026-10-15T21:39:20Z)

This lane re-verified the fix round on its own PR #623 at 1e30fa7, read-only in its own detached worktree; nothing was authored.

MERGEABLE WITH NITS. No sequencing constraint.

Items 1 to 5 of the review at 8b0f14d are fixed and tested:

1. Step (5): when the second run exits 0, setup now says the refusal did not happen again and to run setup again, with the first run's exit code. The other wordings are unchanged (probe P4, 8 cases).
2. An empty --pin-source goes to step (4) as ., which is the scaffold's Path('') read in the same folder. Step (4) warns, and the plan refuses pin-source-unreadable with exit 2, as on main; link-source-no-path is gone (probe P2). . is the right choice over leaving the flag out.
3. A one-link chain gets no --link-source, so step (4) prints no NOTE (probe P3). A two-link chain still verifies.
4. Only a failed mkdtemp refuses. A failed rmdir is a [??] line naming the folder.
5. The S9073 asserts are split (Sonar shows 0 issues), and the rows for mutants R1, R5, R6 and R17 are added.

Two residual cases remain: a first run that refused before the plan followed by a second that printed it and refused, and the reverse. Both need two different refusals in a row from a passing cause. The docstring and the PR body name them, so they are acceptable with no #459 line. Item 6 is L2-62 on #459.

Nit, for the landing: line 1 of the lane's squash message lost its (#608).

Tests on 1e30fa7: set A 649 passed, 19 skipped; set B 634 passed. Mutants: 14 of 14 killed.

Merge: merge-tree with origin/main fbbbc8d, which moved during the review by one adopt-only commit, is clean.

Report: handoffs/xFactory/attachments/lane-openRepoShape-2/review623-1e30fa7/final-report.md in opensoft/brett-wip.

[brettheap]

(lane openRepoShape-2, 2026-10-16T02:23:53Z) A readiness round for main 4cf5fb07, re-verified MERGEABLE WITH NITS (rv623g); the nit is now fixed. The head is 400965b3. No merge word has been given.

• Why the round. The PR merged cleanly with main 4cf5fb07, but two tests then failed. Lane 2's own scaffold-project.py and adopt-project.py: a pin records digest_source: local-clone and never the --pin-source path, so no committed pin file holds a local directory, and a --pin-source nothing reads is pin-source-unused (#540) #570 makes the scaffold refuse an exactly empty --pin-source as pin-source-no-path, and step (4) still passed --link-source=. for it, so the naming validator warned about . under a flag the person never typed.
• be444527 is git's own clean merge of 4cf5fb07. Its tree equals merge-tree, and no hunk was resolved by hand.
• d0a65ba7, the fix. An exactly empty LAST --pin-source goes to the validator as no --link-source, by the scaffold's own test, not str(typed).
  • rv623g ran a matrix through setup-project.py and setup.sh: '', =, src then '', '' then a real folder, blank, and none.
  • Step (4) for an empty last value is identical to the run with no source. The plan refuses pin-source-no-path with exit 2, and nothing is created.
  • A blank value is a path. It is reported like any source the validator cannot read, and the plan refuses it as pin-source-unreadable. That was so before this round too.
• The helper.
  • It kills two mutants end to end: restoring the . mapping, and passing --link-source=.
  • The unit rows of the declarations table kill the other two: reading the first value, and treating a blank as empty.
  • Its re.sub strips only the validator's own hint. Every real --link-source report keeps a form it does not match.
• Tests.
  • f623g: 1526 passed and 19 skipped on 24 files.
  • rv623g: 630 passed on 7 files.
  • The old expectation fails on be444527, as it should.
• 400965b3 fixes rv623g's nit: the helper's docstring now says the hint is printed once for each holder the validator cannot verify. The PR body's wording is corrected the same way.

400965b3 is a new H0 and needs Linux and SonarCloud green. Report: handoffs/xFactory/attachments/lane-openRepoShape-2/review623-d0a65ba/final-report.md in opensoft/brett-wip.

[sonarqubecloud]

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