[brettheap]

What changed

A follow-up outside v1.2.0 (Brett Heap, 2026-10-21T10:32:04Z). It lands only after the v1.2.0 tag, on a later merge word.

One new file, tests/test_bom_tool_io.py (126 rows), and no tool code. #1050's byte-order-mark rows (tests/test_bom_rewrites.py) hold what each line rewriter returns. They do not hold the tools' own read and write, nor that the cut in front of line 1 takes U+FEFF and nothing else, so rv1050's five mutants each passed every file that reads a pin (#1176).

• The tools' own read and write. update-shape.py's _repin (the shape pin and the manifest, through the Rollback ledger) and bump-leg.py's _leg_pin_text, rewrite_pin and write_lf (in the order its main calls them) run over a file that starts with 0, 1 or 2 marks. The file written starts with exactly the marks it was read with, and every byte after them is the rewrite. The marks are bytes (codecs.BOM_UTF8), never typed.
• Each tool end to end. update-shape.py apply (a local --upstream only, never the default GitHub one) on a root whose pin and manifest are committed with the marks, and bump-leg.py on a leg pin with the marks: both write the marks back, and the project's own gate is green after.
• Only U+FEFF is cut, and only in front of line 1. Another zero-width or format (Cf) character as the second character (ZWSP, ZWNJ, ZWJ, word joiner, soft hyphen, LRM) stays: in front of the key the file is refused unchanged in each tool's words, and in front of a key of its own it is written back where it was. A space or a tab in front of the key (with no mark, behind a mark and in front of one) and a file indented two spaces on every line are refused as before update-shape.py and scripts/bump-leg.py: the three line rewriters find a first-line key after a byte-order mark, as every reader does, and write the mark back as it was (#933) #1050. A mark that is the second character of a comment is text and is written back.
• No fault found. Every row passes at main, so no tool changes. The issue's own measurement holds: the tools keep the mark today, and only a test was missing.

Measured at main (961640c6)

• Before: tests/test_bom_rewrites.py + tests/test_bom_reads.py = 37 passed. rv1050's five mutants (utf-8-sig at _repin's two reads and at _leg_pin_text's read; an lstrip that also cuts spaces and line feeds in _marks_off and in rewrite_pin) each pass the six files of the issue (272 passed under each).
• After: tests/test_bom_tool_io.py = 126 passed in about 13 s. Each of the five fails 5 to 20 of its rows (table below), and every row passes without them.

Mutants

Each is one exact-once edit, run against the 3 or 6 existing files (the BOM files, plus the files that read a pin for M1 to M5 and M19 to M22) and against the new file. The tool file is restored from a saved copy after each and checked with cmp (22 of 22 cmp exit 0; git diff of the tools against HEAD is empty). Runner: W1176's own mutants.py, in the lane's scratch.

#	Mutant	Existing files	New rows
M1	_repin reads the shape pin as utf-8-sig	survives (272 passed)	KILLED (20 failed)
M2	_repin reads the manifest as utf-8-sig	survives (272 passed)	KILLED (20 failed)
M3	_leg_pin_text reads the leg pin as utf-8-sig	survives (272 passed)	KILLED (20 failed)
M4	_marks_off: lstrip also cuts spaces and line feeds	survives (272 passed)	KILLED (10 failed)
M5	rewrite_pin: lstrip also cuts spaces and line feeds	survives (272 passed)	KILLED (5 failed)
M6	_marks_off: cut widened to every Cf character	survives (152 passed)	KILLED (24 failed)
M7	rewrite_pin: cut widened to every Cf character	survives (152 passed)	KILLED (12 failed)
M8	_marks_off: cut left out	killed (4 failed)	KILLED (6 failed)
M9	rewrite_pin: cut left out	killed (2 failed)	KILLED (4 failed)
M10	_marks_off: cut also applied to the second character	survives (152 passed)	KILLED (6 failed)
M11	rewrite_pin: cut also applied to the second character	survives (152 passed)	KILLED (3 failed)
M12	_marks_off: cut takes one mark only	killed (2 failed)	KILLED (3 failed)
M13	rewrite_pin: cut takes one mark only	killed (1 failed)	KILLED (2 failed)
M14	_marks_off: cut takes every mark anywhere	killed (2 failed)	KILLED (6 failed)
M15	rewrite_pin: cut takes every mark anywhere	killed (1 failed)	KILLED (3 failed)
M16	_repin writes the shape pin as utf-8-sig	survives (152 passed)	KILLED (22 failed)
M17	_repin writes the manifest as utf-8-sig	survives (152 passed)	KILLED (22 failed)
M18	write_lf (bump-leg.py's write) as utf-8-sig	killed (1 failed)	KILLED (22 failed)
M19	_marks_off: a space or a tab in front of the marks is cut first (lstrip(" \t"), then the marks)	survives (272 passed)	KILLED (10 failed)
M20	rewrite_pin: the same cut, in front of the marks, first	survives (272 passed)	KILLED (5 failed)
M21	_marks_off: a space or a tab is cut only where a mark follows it	survives (272 passed)	KILLED (4 failed, the rows with the lead in front of a mark)
M22	rewrite_pin: the same, only where a mark follows	survives (272 passed)	KILLED (2 failed, the same rows)

Tests

Under a stub gh (exit 97), GH_TOKEN=bogus, an empty GH_CONFIG_DIR, GIT_SSH_COMMAND=false, a scratch GIT_CONFIG_GLOBAL and EMAIL=tests@openreposhape.invalid, one pytest at a time. The two lines below are the first two heads', with 120 rows; the third head's gates, with 126, are in Fix round 2:

• tests/test_bom_tool_io.py tests/test_bom_rewrites.py tests/test_bom_reads.py tests/test_pin_rewrite_layout.py tests/test_repo_hygiene.py tests/test_path_guard_hygiene.py tests/test_no_auto_maintenance.py tests/test_agents_md_codes.py tests/test_cli_reference.py: 550 passed, 2 skipped in 87.26s.
• tests/test_update_shape.py tests/test_bump_leg.py tests/test_update_shape_apply_lines_and_headers.py tests/test_update_shape_header_comments.py tests/test_bump_leg_dry_run_truth.py (the helpers the new file imports, and the files the issue names): 264 passed, 1 skipped in 200.18s.
• tests/test_shape_doctor.py is not run: no shape-doctor.py or helper changes.

Merges

Re-listed live on 2026-10-23 at about 07:52Z and again at 08:21Z (the same eleven PRs at the same heads), and run at d65340e5b: git merge-tree --write-tree of this head is clean (exit 0) with origin/main (961640c6, frozen until the v1.2.0 tag) and with each of the eleven open PRs that touch update-shape.py, scripts/bump-leg.py, scripts/shape_materialize.py or scripts/repo_shape.py:

• update-shape.py: update-shape.py: apply puts every byte back under any OSError after its first write, says what it put back on stderr, and raises the error as it was, so a file gone between its stat and its read, a broken pipe or an EIO no longer leaves the tree half-written behind a traceback (#1173) #1187 (cc8a87ee93, the apply rollback of update-shape.py: apply rolls back an OSError after its first write only where the system refused a read, so a file gone between the stat and the read, or an error naming no path, ends in a traceback with the tree half-written #1173, this lane's sibling), 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 (2593952c95), update-shape.py: update-accept-local-missing-name asks a copy an earlier apply kept, read unchanged, as a module a written copy imports from and as an importer of a written one, so a plain apply that stops either on its import is refused (#1082) #1133 (a7bfee1bdb);
• scripts/bump-leg.py: bump-leg.py: a tag a leg with no origin holds without its commit, or through a tag it lacks, is told so, a tag of a tree or a blob it lacks is refused by name, and rows hold YAML_NOT_TEXT_RE's exponents of two digits (#1103) #1131 (55c30f20f0), shape_materialize.py and bump-leg.py: the fetch for a branch's answer leaves out a pattern like refs/heads* that puts a branch under another name, its clauses move into _left_out, and the docstrings name the clashes left and the way back (#1100) #1157 (baa19972a2, also scripts/shape_materialize.py);
• scripts/shape_materialize.py: scripts/shape_materialize.py: adopt_makefile_block's docstring says the block lands where place writes the shape's Makefile, the root's or shape/Makefile where the source keeps one, not the root's alone (#1135) #1160 (0068f5af8a), adopt-project.py, shape_materialize.py and tests: a named bad link left out of the copy and pin findings takes no other path's finding with it, and the docstrings and the materializer's refusal name the #979 exception and what helps it (#1089) #1118 (2307036abe), shape_materialize.py and the pinned AGENTS-shape.md: where the root keeps a GNUmakefile or a makefile, which GNU make reads first, the Makefile collision follow-up no longer says the shape's make targets arrive with the merge, and the pinned file says make reads the first of the three and the block reaches it only once that file holds it (#1048) #1105 (b28f40de09), shape_materialize.py: env_commit's docstring says the split commit's message spells a path byte that is not UTF-8 as _shown does, an escape, and no longer says it names it with a ? in its place, which #900 replaced (#1018) #1104 (de89aa81b1);
• scripts/repo_shape.py: adopt-project.py and repo_shape.py: a name that holds a default-ignorable character, which a terminal draws as nothing, is spelled by its code point in plan's lines and in the unsafe-value refusal, so a branch ma<U+034F>in no longer prints as main (#1128) #1171 (8162b874ec) and repo_shape.py and bootstrap.py: COMMIT_RE, SHA256_RE and COMMIT_40 end in \Z, not $, so a pin's commit: or tree_sha256: that ends in a line break is refused as not 40 or 64 hex by every validator and adopt-project.py (#1017) #1126 (4c1f80cf4d).

#1178 no longer touches these files. No open PR touches this file or tests/test_update_shape.py, tests/test_bump_leg.py, tests/test_bom_rewrites.py, tests/test_pin_rewrite_layout.py or tests/conftest.py, which it imports. The rows were not run on a merged tree: a clean merge is a clean merge of text only.

Fix round

CI's full suite on the first head (aa0275be, tests.yml run 39861510476) gave 47 setup errors, all in tests/test_bump_leg.py. tests-git-bash, package and SonarCloud passed.

• Cause. tests/test_bom_tool_io.py imported tests/test_bump_leg.py's module-scoped fixture advanced. A module-scoped fixture runs once for every module that names it, and both runs push HEAD to bump-leg-advance on the session's one shared remote. The file sorts first, so tests/test_bump_leg.py's own push was refused as non-fast-forward. No gate before the push ran the new file in the same pytest as tests/test_bump_leg.py, so none could see it.
• Fix. One new commit, test file only, no tool change. tests/test_bom_tool_io.py has its own module-scoped advanced, which pushes to bom-tool-io-advance, as the other modules that advance the spec leg do (tests/test_bump_leg_unreadable.py and others). Nothing else the file imports writes shared state: upstream_and_project builds in its own tmp paths and tools loads two modules. All 120 rows are kept, and the three end-to-end bump-leg.py rows still run against a real advance of the leg.
• Combined run, the three files in ONE pytest, stub gh, GIT_SSH_COMMAND=false, EMAIL=tests@openreposhape.invalid:
  • tests/test_bom_tool_io.py tests/test_bump_leg.py tests/test_update_shape.py at aa0275be: 218 passed, 47 errors in 64.96s, each ERROR a setup of tests/test_bump_leg.py with ! [rejected] HEAD -> bump-leg-advance (non-fast-forward). The same 47 as CI.
  • The same command with the fix: 264 passed, 1 skipped in 80.64s.
• Mutants, all 18 of mutants.py re-run at 024084b9, each to its end: 18 of 18 new rows: KILLED (the kill counts of the first head), 18 of 18 cmp exit 0, runner exit 0, and git diff of the three tool files against HEAD empty.
• Gates, one pytest at a time under the stub gh:
  • the nine-file gate above: 550 passed, 2 skipped in 50.49s;
  • the five-file update-shape and bump-leg gate above: 264 passed, 1 skipped in 130.80s;
  • tests/test_bom_tool_io.py with those five files in ONE pytest, in CI's order and reversed: 384 passed, 1 skipped in 131.44s and 384 passed, 1 skipped in 134.05s.
• Merges of 024084b9: git merge-tree --write-tree is clean (exit 0) with origin/main (961640c6) and with update-shape.py: apply puts every byte back under any OSError after its first write, says what it put back on stderr, and raises the error as it was, so a file gone between its stat and its read, a broken pipe or an EIO no longer leaves the tree half-written behind a traceback (#1173) #1187, 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, scripts/shape_materialize.py: adopt_makefile_block's docstring says the block lands where place writes the shape's Makefile, the root's or shape/Makefile where the source keeps one, not the root's alone (#1135) #1160, shape_materialize.py and bump-leg.py: the fetch for a branch's answer leaves out a pattern like refs/heads* that puts a branch under another name, its clauses move into _left_out, and the docstrings name the clashes left and the way back (#1100) #1157, update-shape.py: update-accept-local-missing-name asks a copy an earlier apply kept, read unchanged, as a module a written copy imports from and as an importer of a written one, so a plain apply that stops either on its import is refused (#1082) #1133, bump-leg.py: a tag a leg with no origin holds without its commit, or through a tag it lacks, is told so, a tag of a tree or a blob it lacks is refused by name, and rows hold YAML_NOT_TEXT_RE's exponents of two digits (#1103) #1131, adopt-project.py, shape_materialize.py and tests: a named bad link left out of the copy and pin findings takes no other path's finding with it, and the docstrings and the materializer's refusal name the #979 exception and what helps it (#1089) #1118, shape_materialize.py and the pinned AGENTS-shape.md: where the root keeps a GNUmakefile or a makefile, which GNU make reads first, the Makefile collision follow-up no longer says the shape's make targets arrive with the merge, and the pinned file says make reads the first of the three and the block reaches it only once that file holds it (#1048) #1105, shape_materialize.py: env_commit's docstring says the split commit's message spells a path byte that is not UTF-8 as _shown does, an escape, and no longer says it names it with a ? in its place, which #900 replaced (#1018) #1104 and adopt-project.py: plan's table pads each name by the columns a terminal gives it, not the code points it holds, so a wide or combining name's leg column lines up under leg with every other row's (#1034) #1178. No other open PR touches tests/test_bom_tool_io.py.

Fix round 2

RV1188's review of 024084b9 (issuecomment-6360127603): READY, one SHOULD and three NITs. One new commit, tests/test_bom_tool_io.py only; the three tool files are byte-identical to HEAD (cmp against saved copies, exit 0).

• SHOULD-1, SonarCloud python:S3415 ("Unify assertion argument order in this file"). The one == assert with a tuple on its left, assert (pin.read_bytes(), manifest.read_bytes()) == before, is now now = pin.read_bytes(), manifest.read_bytes() and assert now == before, the shape tests/test_naming_policy.py: bind the compared tuple to a name before asserting it, so Sonar's assertion-order rule stops misreading the literal-looking display as the expected side (#130) #131 (d682127d6) fixed. The name is now, not after, which is a function in this file. A scan of every assert in the file by syntax tree found no other == with the expected value on the left: the in asserts keep the text looked for on the left, as in reads. The local analyzer does not reproduce the rule, so the result is read on SonarCloud once CI has run on this head.
• NIT-1. A space or a tab in FRONT of a mark is now a row for each of the three files (LEAD_PLACES: no mark, behind a mark, in front of one). Each tool refuses it unchanged by its own code (update-pin-unrecognised, update-manifest-no-shape, bump-leg-pin-unrecognised), as the reader does. Six rows more: 126 in all. The module docstring, the row's docstring and the third bullet above say what the rows hold.
• NIT-2. The Merges list above is re-listed live, with scripts/repo_shape.py added to the files watched.
• NIT-3. The two end-to-end refusal rows use FINDERS["shape-pin-commit"].refused and FINDERS["leg-pin-commit"].refused in place of the retyped words.
• Combined runs, tests/test_bom_tool_io.py, tests/test_bump_leg.py and tests/test_update_shape.py in ONE pytest, stub gh, EMAIL=tests@openreposhape.invalid: 270 passed, 1 skipped in 129.58s in CI's order and 270 passed, 1 skipped in 137.65s reversed. With the three other files the issue names (tests/test_bump_leg_dry_run_truth.py, tests/test_update_shape_apply_lines_and_headers.py, tests/test_update_shape_header_comments.py), six files in ONE pytest: 390 passed, 1 skipped in 218.42s and 390 passed, 1 skipped in 195.71s reversed.
• Gates, one pytest at a time: the new file with tests/test_bom_rewrites.py, tests/test_bom_reads.py and tests/test_pin_rewrite_layout.py: 278 passed in 21.85s. The hygiene files, each alone: tests/test_repo_hygiene.py 101 passed, tests/test_path_guard_hygiene.py 49 passed, tests/test_no_auto_maintenance.py 58 passed and 2 skipped, tests/test_agents_md_codes.py 47 passed, tests/test_cli_reference.py 23 passed.
• Mutants, all 18 of mutants.py re-run at this head, each to its end, and four more for the new rows (M19 to M22): 22 of 22 new rows: KILLED, 22 of 22 cmp exit 0, runner exit 0. M4 and M5 fail two and one rows more than before. M21 and M22 cut a space or a tab only where a mark follows it, so only the new rows in front of a mark kill them; the six existing files all pass under both.

Left out of scope

docs/release-notes/v1.2.0.md (it records v1.2.0's state, known limits included, and is not edited). No milestone.

Closes #1176

Lane: openreposhape-3 (openRepoShape-3)

🤖 Generated with Claude Code

[sourcery-ai]

Reviewer's Guide

Adds 120 regression tests covering the complete BOM lifecycle in update-shape.py and bump-leg.py, including byte-preserving reads and writes, narrowly scoped U+FEFF stripping, refusal behavior, and end-to-end command validation; no tool implementation changes are included.

File-Level Changes

Change	Details	Files
Added comprehensive byte-order-mark I/O coverage for both pin-manipulation tools.	Tests preservation of zero, one, and two UTF-8 BOM byte sequences through each tool’s actual read and write path. Covers update-shape.py repinning of the shape pin and manifest, plus bump-leg.py pin rewriting and writing. Verifies end-to-end CLI behavior preserves BOMs and leaves project validation green.	tests/test_bom_tool_io.py
Added regression coverage defining the exact scope and position of BOM removal.	Confirms only leading U+FEFF bytes on the first line are removed before key matching. Ensures spaces, tabs, indentation, other Unicode format characters, and later BOMs are preserved or rejected without modifying files. Checks refusals retain original bytes and expected tool-specific error codes.	tests/test_bom_tool_io.py
Strengthened mutation-detection coverage for BOM reads, writes, and stripping behavior without changing production code.	Exercises mutants involving utf-8-sig, widened or misplaced stripping, incomplete mark removal, and unconditional BOM insertion. Uses byte-level assertions to distinguish marks from typed U+FEFF text and verify rewritten content exactly.	tests/test_bom_tool_io.py

Assessment against linked issues

Issue	Objective	Addressed	Explanation
#1176	Test each tool’s real read-and-write path to ensure shape pins, manifests, and leg pins retain exactly the one or two UTF-8 BOMs they were read with.	✅
#1176	Add regression coverage that only U+FEFF is removed at the start of line 1, while spaces, tabs, other format characters, and marks elsewhere remain intact or cause the same refusal as before.	✅

Possibly linked issues

• update-shape.py and bump-leg.py: the byte-order-mark rows (#1050) hold what each rewriter returns but not the tools' own read and write, nor that only U+FEFF is cut, so a utf-8-sig read that drops the mark and a cut that also takes spaces both survive #1176: The PR adds tests exercising each tool’s read/write path and rejects mutants that drop BOMs or strip spaces and other characters.

---

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-3, 2026-10-23T07:48:07Z) Independent review (RV1188, opus, ruling 8) of #1188 at 024084b98ed8a77ae7ec6fe5513b3032f86dd461: READY. One SHOULD and three NITs; none blocks landing after the v1.2.0 tag.

SHOULD-1, tests/test_bom_tool_io.py:446.

• SonarCloud has one OPEN issue on this PR: python:S3415 (MAJOR), "Unify assertion argument order in this file", filed at :188. The gate is still OK.
• The only assert in the file with a bracketed tuple on the left is :446, assert (pin.read_bytes(), manifest.read_bytes()) == before. That is the same shape tests/test_naming_policy.py: bind the compared tuple to a name before asserting it, so Sonar's assertion-order rule stops misreading the literal-looking display as the expected side (#130) #131 (d682127d6) fixed.
• Fix: now = pin.read_bytes(), manifest.read_bytes(), then assert now == before. Not after, which is a function here. Then check on SonarCloud that the issue has closed.
• The local analyzer does not reproduce the rule, so :446 is inferred from tests/test_naming_policy.py: bind the compared tuple to a name before asserting it, so Sonar's assertion-order rule stops misreading the literal-looking display as the expected side (#130) #131, not measured.

NIT-1, tests/test_bom_tool_io.py:38 and :254-256, and the body's third bullet: "before a mark or after one".

• The rows hold "no mark" and "after one mark"; no row puts the space or tab in front of a mark.
• Measured: that case is refused unchanged today by all three tools (update-pin-unrecognised, update-manifest-no-shape, bump-leg-pin-unrecognised), and the reader refuses it too.
• Add a before-the-mark row, or reword.

NIT-2, the body's Merges.

• The file also imports scripts/repo_shape.py, which adopt-project.py and repo_shape.py: a name that holds a default-ignorable character, which a terminal draws as nothing, is spelled by its code point in plan's lines and in the unsafe-value refusal, so a branch ma<U+034F>in no longer prints as main (#1128) #1171 (8162b874ec) and repo_shape.py and bootstrap.py: COMMIT_RE, SHA256_RE and COMMIT_40 end in \Z, not $, so a pin's commit: or tree_sha256: that ends in a line break is refused as not 40 or 64 hex by every validator and adopt-project.py (#1017) #1126 (4c1f80cf4d) touch.
• update-shape.py: apply puts every byte back under any OSError after its first write, says what it put back on stderr, and raises the error as it was, so a file gone between its stat and its read, a broken pipe or an EIO no longer leaves the tree half-written behind a traceback (#1173) #1187's head is now cc8a87ee93.
• adopt-project.py: plan's table pads each name by the columns a terminal gives it, not the code points it holds, so a wide or combining name's leg column lines up under leg with every other row's (#1034) #1178 no longer appears among the open PRs touching these files.
• Merge-tree is clean with all eleven open PRs. Refresh the list at the next body edit.

NIT-3, :445 and :462 retype "has 0 commit line(s)". FINDERS["shape-pin-commit"].refused and FINDERS["leg-pin-commit"].refused hold the same words. Optional.

Observation, nothing for this PR:

• update-shape.py's _is_block_header (:1695-1698) uses str.strip(), so a no-break space (U+00A0) or an ideographic space (U+3000) in front of shape: is taken as the header. A direct call to _repin wrote such a manifest, while parse_yaml reads that line as another key.
• This is already on main, is not the byte-order-mark cut, and no row here pins it. It was not run through apply end to end.
• A follow-up issue is the lane's call.

Checked:

• Tools: byte-identical at main and the head; only the test file changes.
• Own probe: the three files with 0 to 3 marks are written back with exactly those marks. A mark followed by a space, tab, zero-width space, word joiner, right-to-left mark or U+FFFE is refused unchanged. A mark inside a comment is written back.
• The fixture fix: at aa0275be a narrow selection gives 2 passed and 3 errors (rejected HEAD -> bump-leg-advance); at the head it gives 5 passed. bom-tool-io-advance is unique among the 14 ADVANCE_BRANCH names.
• Five modules together (bom_rewrites, bom_tool_io, bump_leg, pin_rewrite_layout, update_shape): 391 passed, 1 skipped, in CI order and reversed.
• The new file with the three hygiene files: 328 passed, 2 skipped.
• On a merge of this PR with all eleven open PRs touching the tools or repo_shape.py: the BOM files give 157 passed.
• Own mutants, each restored with cmp 0, all KILLED, and all surviving the three existing files:
  • a utf-8-sig read in _repin (20 failed);
  • an lstrip that cuts spaces (8);
  • the cut widened to every Cf (24);
  • a second-character mark cut (6);
  • the shape pin written with a mark (22);
  • a new one, a utf-8-sig write at _apply_rewrite (2, the end-to-end bump-leg rows).
• CI: tests, tests-git-bash, package and SonarCloud all SUCCESS. The Sonar gate is OK, with 0.0% duplication.
• Body and commits: as required.
• Not re-made: M2, M3, M5, M7-M9, M11-M15, M17 and M18. The writer's log shows all 18 killed.

Lane: openreposhape-3 (openRepoShape-3)

[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

[brettheap]

(lane openRepoShape-3, 2026-10-23T08:30:48Z) Delta review (RV1188, opus) of #1188's second fix round: head d65340e5, on 024084b9.

Verdict: READY. All four findings from the review of 024084b9 are fixed, and nothing new is wrong. Land only once CI's tests on d65340e5 is green; it was still running at 08:29:55Z (started 08:20:58Z).

• Tools: update-shape.py, scripts/bump-leg.py and scripts/shape_materialize.py each cmp exit 0 against main 961640c6. Both commits since main change tests/test_bom_tool_io.py only.
• SHOULD-1 is done. :457-458 bind the pair to now and assert now == before. On SonarCloud, S3415 (AaFloFMrgxXxprHhSJPb) reads CLOSED and was the PR's only issue; the gate is OK and new duplication is 0.0%. Every == assert in the file has the actual value on the left.
• NIT-1 is done. The six LEAD_PLACES rows (a space or a tab in front of a mark, in each of the three files; 126 rows in all) check each tool's own code (update-pin-unrecognised, update-manifest-no-shape, bump-leg-pin-unrecognised), the finder's words and an unchanged file.
  • Own mutant, a leading space or tab cut only where a mark follows: in _marks_off KILLED (4 failed), in rewrite_pin KILLED (2 failed). Every failure is a before-a-mark row, and both mutants survive the three existing files, so only the new rows hold this.
  • M4 re-run (lstrip also cuts spaces): KILLED, 10 failed.
  • Each mutant was restored, cmp exit 0.
  • The module docstring, the row docstring and the body's third bullet now match the rows.
• NIT-3 is done. :456 and :474 use FINDERS["shape-pin-commit"].refused and FINDERS["leg-pin-commit"].refused.
• NIT-2 is done. The Merges list covers scripts/repo_shape.py, adopt-project.py and repo_shape.py: a name that holds a default-ignorable character, which a terminal draws as nothing, is spelled by its code point in plan's lines and in the unsafe-value refusal, so a branch ma<U+034F>in no longer prints as main (#1128) #1171, repo_shape.py and bootstrap.py: COMMIT_RE, SHA256_RE and COMMIT_40 end in \Z, not $, so a pin's commit: or tree_sha256: that ends in a line break is refused as not 40 or 64 hex by every validator and adopt-project.py (#1017) #1126, and update-shape.py: apply puts every byte back under any OSError after its first write, says what it put back on stderr, and raises the error as it was, so a file gone between its stat and its read, a broken pipe or an EIO no longer leaves the tree half-written behind a traceback (#1173) #1187 at cc8a87ee93. adopt-project.py: plan's table pads each name by the columns a terminal gives it, not the code points it holds, so a wide or combining name's leg column lines up under leg with every other row's (#1034) #1178 touches only adopt-project.py and one test file. merge-tree of d65340e5 is clean with main and all eleven PRs.
• Body: the table, the Fix round 2 section and the Tests note match what was measured. It still ends Closes #1176, the Lane line, then the robot line, with no bot block, and it is still a draft with no milestone.
• Combined run, CI order: tests/test_bom_tool_io.py, tests/test_bump_leg.py and tests/test_update_shape.py in one pytest gave 270 passed, 1 skipped.
• CI on d65340e5: tests-git-bash, package and SonarCloud passed; tests was still running.
• Skipped, to stay lean: the reversed order, the hygiene files and the other mutants. Only the test file changed, and the writer ran all of them.

Lane: openreposhape-3 (openRepoShape-3)

[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 10 hours and 26 minutes by commenting @sourcery-ai review. Upgrade to get a review now.