[brettheap]

Goal

#1132, the parts that hold on main: the assembly-root template says what to run where GNU make reads a kept GNUmakefile or makefile first; a kept DIRECTORY of either name is named as one GNU make stops at, in the CONTRACTS_DIR line and in execute's NEXT; every sentence that names python3 scripts/bootstrap.py for a person to run names python on Windows too; and two follow-up texts no longer say more than is true (RV1105b's NIT-2 and NIT-3a).

This PR was built on main 961640c6. Lane 4's #1121 (#1047) landed as 0ed3e57f8 at 21:37Z, and this round merged it in (6fa1edd56): NEXT now words a kept file (#1121's) and a kept directory (this PR's) through one lookup and one clause. #1105 landed as 12122952b on 2026-11-08, and this PR merges it in (see "Main merge after #1105"). The items that live only in #1105's code, and the parts of #1121's that this merge does not do, are in their own issue, #1240 (see "Left"). Item 5 is #1206.

Main merge after #1105

Lane 4's #1105 (#1048) landed as 12122952b at 16:45Z. This round merges main into the branch with git merge origin/main, never a rebase, and does it twice:

• main 46fe6ac19 as merge 7b0192bde;
• then NIT-A in its own commit, e5e077940;
• then main 9729315f1 as merge 01e93b6d1, the head. Lane 1's README.md, adopt-project.py and the handbook: a kept directory entry makefile/ is tested, README says make at the root reads a kept GNUmakefile or makefile instead, and a docstring says the shape's Makefile alone holds CONTRACTS_DIR (#1061) #1123 had landed meanwhile, and one of its rows meets this PR's directory line (see "The second merge, after README.md, adopt-project.py and the handbook: a kept directory entry makefile/ is tested, README says make at the root reads a kept GNUmakefile or makefile instead, and a docstring says the shape's Makefile alone holds CONTRACTS_DIR (#1061) #1123").

The first merge conflicted in scripts/shape_materialize.py only: 3 hunks, all in collision_follow_up, as "Landing order" said. adopt-project.py and templates/assembly-root/AGENTS-shape.md merged with no conflict.

Both sides' behaviour stays: #1105's second GNU_MAKE_READS, its _predict_collisions lookup and its collision line, and this PR's _make_reads_first, _make_reads_first_clause and NEXT. None of #1240's unification is done here (one tuple, one lookup).

The three hunks

hunk	this PR (6fa1edd56)	main (46fe6ac19, #1105)	resolution
1. The docstring's NIT-4 sentence	"This said "both callers say it", and execute calls it nowhere"	"Materialized.follow_ups() would say this of them, and nothing calls it"	Main's. Both sides rewrote the same sentence for the same finding. Main's also names Materialized.follow_ups(), which #1240's item G changes. So this PR no longer touches that sentence, and the last scripts/shape_materialize.py bullet under "What changed" is #1105's text now.
2. The docstring's Makefile paragraphs	the BOOTSTRAP_COMMAND sentence and NIT-2's tail	the heading "WHERE make READS THE MERGED FILE", and the "BUT GNU MAKE READS" paragraph	Both. Main's heading and both paragraphs. This PR's sentences on BOOTSTRAP_COMMAND and on NIT-2's tail go in the first paragraph, and one new sentence goes in the second: "Both name that command as BOOTSTRAP_COMMAND does, python on Windows as well (#1132)."
3. The returns	two returns, each one sentence	arrive, joint, meanwhile and until set per branch, then one pair of returns	#1105's structure, with a per-branch then in place of until. With no reads_first, the line is this PR's, byte for byte. With one, it is #1105's, byte for byte, except that the command a person runs is BOOTSTRAP_COMMAND, both in meanwhile and in "and then ... runs it".

• Hunks 1 and 2 carry no code. In the merged file they sit at lines 1022-1023 and 1032-1070, inside collision_follow_up's docstring (lines 1014-1084).
• The rendered lines. I loaded the function from 6fa1edd56, from main, and from the merge, and rendered each reads_first (none, GNUmakefile, makefile) with and without scripts/bootstrap.py aside.
  • With no reads_first, the merge equals 6fa1edd56 (2 of 2).
  • With one, it equals main with " (python on Windows)" after python3 scripts/bootstrap.py in those two places (4 of 4). For a root that keeps Makefile and GNUmakefile, the line now ends "...; until then python3 scripts/bootstrap.py (python on Windows) is what the shape's make bootstrap runs".
• NIT-2's tail where make reads another file stays 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's own: "neither make bootstrap nor python3 scripts/bootstrap.py runs the shape's bootstrap until then, and then python3 scripts/bootstrap.py (python on Windows) runs it". 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's review fixed that branch for the same finding, so it needs nothing more.
• Of adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240's list, this round does item C for 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's collision line, as adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240 asks a resolution to say. So the "with python3" part of "Left"'s second bullet is done. Item B, the kept directory, stays: 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's lookup asks _root_keeps, so a root that keeps Makefile and GNUmakefile/ is told GNU make "reads the source's GNUmakefile". Mutant U3 below shows that no row pins this line for a directory.

#1105's test pins

tests/test_adopt_collision_make_reads.py is #1105's file, and it pins the texts this PR changes, as "Landing order" expected.

• Four pins move. MERGE_IS_ENOUGH and BOTH_MERGES (no reads_first) move to this PR's words. MEANWHILE and SCRIPT_MERGED (with one) move to python on Windows. All four go through a new COMMAND.

• One comment changes. The comment on the Makefile+scripts row now says what NIT-2's tail says.

• Measured: before the move, the six adopt and pin files gave 11 failed and 242 passed, and the 11 are exactly the rows that use those four pins. After the move, 253 passed.

• On main's code: main 46fe6ac19's four production files (adopt-project.py, scripts/shape_materialize.py, templates/assembly-root/AGENTS-shape.md and README.md) were swapped into the mutant worktree at e5e077940. 64 of the three files' 106 rows fail:

  • in this file, exactly those 11 rows, and the other 20 pass;
  • in the new file, 34 of 35, as before (NIT-4's row passes);
  • in tests/test_adopt_next_make_reads_first.py, 19 of 40, as before.

  The four files were then restored, each matched the head by sha256 (4 of 4), and the tree was clean.

NIT-A

tests/test_adopt_next_make_reads_first.py:64 called #480's NOTE "unchanged", but the merge of #1121 moved that pin to python on Windows. The comment now says the NOTE "names python on Windows too since #1132". The change is a comment only, in its own commit.

The second merge, after #1123

Lane 1's #1118 and #1123 landed while the first merge was being gated: 9ec0dc79c at 17:30Z and 9729315f1 at 17:47Z. Both merge with e5e077940 with no conflict in the text. The predicted merge with 9ec0dc79c passed (gate P below). The predicted merge with 9729315f1 failed 1 row of 254 (gate P2):

• The row: README.md, adopt-project.py and the handbook: a kept directory entry makefile/ is tested, README says make at the root reads a kept GNUmakefile or makefile instead, and a docstring says the shape's Makefile alone holds CONTRACTS_DIR (#1061) #1123 adds test_a_name_only_a_disk_that_ignores_case_takes_for_makefile[makefile-directory] to tests/test_adopt_make_reads_first.py. For a kept DIRECTORY makefile/, it expects the CONTRACTS_DIR line a kept makefile FILE gets ("only once makefile does"). README.md, adopt-project.py and the handbook: a kept directory entry makefile/ is tested, README says make at the root reads a kept GNUmakefile or makefile instead, and a docstring says the shape's Makefile alone holds CONTRACTS_DIR (#1061) #1123's own message says that line is not true for a directory, and that the row works as a crash guard for twins in _makefile_follow_up. This PR changes exactly that line: a kept directory is one GNU make stops at.
• The fix, in the merge 01e93b6d1: the row now expects SHAPES + written_as("makefile"), as README.md, adopt-project.py and the handbook: a kept directory entry makefile/ is tested, README says make at the root reads a kept GNUmakefile or makefile instead, and a docstring says the shape's Makefile alone holds CONTRACTS_DIR (#1061) #1123 wrote it, which is the part that guards twins. This PR's directory clause follows, through a new stops_at helper beside reads. The comment above ANSWERED and the row's docstring now say so. Nothing else changes in the merge.
• Mutants on the row. Each ran alone in the mutant worktree at 01e93b6d1 against tests/test_adopt_make_reads_first.py (34 rows). Each was restored with git checkout and matched by sha256 (adopt-project.py 3c2cb5f5bea5), and the tree was clean after each.

id	mutant	rows failing	result
D1	_make_reads_first drops a directory's /, so the line is the one #1123's row pinned	1 of 34, the row	killed
D2	#1123's R10: twins reads the entry with its slash	1 of 34, the row, with IndexError	killed

Mutants on the resolution

H3a-H3g and U1-U3b are the first merge's. All ran in the mutant worktree at e5e077940.

• Method: each mutant was applied alone, then restored with git checkout. Each restore matched by sha256 (shape_materialize.py b18f09d7c396, adopt-project.py 2055a81acdc6), and the tree was clean after each.
• Rows: H3a-H3g, U3 and U3b ran against 93 rows: tests/test_adopt_collision_make_reads.py (31), the new file (35) and tests/test_adopt_e2e.py (27). U1 ran against the new file, tests/test_adopt_make_reads_first.py (33) and tests/test_adopt_collision_make_reads.py, 99 rows. U2 ran against the new file and tests/test_adopt_next_make_reads_first.py (40), 75 rows.
• Unmutated, every row passes (gate A below).

id	mutant	rows failing	result
H3a	hunk 3, no reads_first: meanwhile back to main's plain python3	4 of 93	killed
H3b	hunk 3, no reads_first: the script tail back to main's "before both merges are made"	4 of 93	killed
H3c	hunk 3, with reads_first: meanwhile without BOOTSTRAP_COMMAND (main's)	5 of 93	killed
H3d	hunk 3, with reads_first: "and then ... runs it" without BOOTSTRAP_COMMAND (main's)	3 of 93	killed
H3e	hunk 3, with reads_first: this PR's tail in place of #1105's	3 of 93	killed
H3f	hunk 3: reads_first ignored, so the resolution is this PR's side whole	15 of 93	killed
H3g	hunk 3, with reads_first: #1105's joint ("; and") lost	3 of 93	killed
U1	_makefile_follow_up drops _make_reads_first (first = None)	31 of 99	killed
U2	_next_steps drops _make_reads_first (first = None)	35 of 75	killed
U3	_predict_collisions asks _make_reads_first in place of #1105's lookup	0 of 93	survived
U3b	_predict_collisions passes no reads_first, so #1105's lookup is lost in the merge	15 of 93	killed

• H3c and H3d are killed only by rows that use the moved MEANWHILE and SCRIPT_MERGED. Those pins are what hold BOOTSTRAP_COMMAND in 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's branch.
• U3 survives, and that is adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240's item B. For a kept file, both lookups give the same name. They differ only for a kept directory, where _make_reads_first adds the /, and no row runs _predict_collisions on a root that keeps Makefile and GNUmakefile/. A row now would pin the line adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240 is to correct, so this round adds none.

Runs

Each run went through the lane's runner, with git-filter-repo 2.47.0 (CI's pin) on PATH, on a commit with a clean tree.

run	git	files	passed	failed	skipped
gate A2	2.43.0	01e93b6d1, the head: the 12 files below	1096	0	3: the PowerShell rows (test_shape_doctor.py:10516), which run only on a Windows host
gate B2	2.47.1	01e93b6d1: the same 12 files	1096	0	3: the same
gate A	2.43.0	e5e077940, after the first merge: the 12 files below	1095	0	3: the same
gate B	2.47.1	e5e077940: the same 12 files	1095	0	3: the same
gate P	2.43.0	the predicted merge of e5e077940 with main 9ec0dc79c (#1118; tree ba3f3c1f3, a local commit object, not pushed): the six adopt and pin files, plus #1118's tests/test_adopt_root_shape_paths.py and tests/test_adopt_symlink_target_bytes.py	509	0	0
gate P2	2.43.0	the predicted merge of e5e077940 with main 9729315f1 (#1123; tree 0eb290ae2, a local commit object): the six adopt and pin files	253	1: #1123's makefile-directory row, which the second merge moves	0
gate P4	2.43.0	the predicted merge of 01e93b6d1 with main 801a7d6d0 (#1106; tree 25bc27d39, a local commit object): the six adopt and pin files, plus #1106's tests/test_adopt_one_byte_locale.py and tests/test_adopt_unenterable_source_and_locale.py	277	0	0

The files are the six adopt and pin files (tests/test_adopt_collision_make_reads.py, tests/test_pinned_copy_text.py, the new file, tests/test_adopt_next_make_reads_first.py, tests/test_adopt_e2e.py and tests/test_adopt_make_reads_first.py, 253 rows), plus tests/test_shape_doctor.py and tests/test_scaffold_leg_path_collisions.py, which adopt changes need, plus the hygiene five (test_path_guard_hygiene.py, test_repo_hygiene.py, test_agents_md_codes.py, test_cli_reference.py and test_pinned_copy_text.py). That is 1098 rows in one run at e5e077940, and 1099 at 01e93b6d1, where #1123 adds one row.

What held on main, before the change

All of this was measured on main 961640c68 with GNU make 4.3, offline.

item	on main
1, the template	AGENTS-shape.md (the "Work from here" paragraph and the five-commands table) and README.md (Get started, and the parking steps) give make bootstrap with no caveat. With the template's own Makefile beside a plain GNUmakefile or makefile, make -n bootstrap exits 2, No rule to make target 'bootstrap'.
2, a kept directory	With GNUmakefile/ or makefile/ beside Makefile, make bootstrap exits 2 with make: *** GNUmakefile: Is a directory. Stop., or the same with makefile. It does this for every target, and for each directory alone too. Main's CONTRACTS_DIR line says "make there exports CONTRACTS_DIR only once GNUmakefile does". NEXT prints make bootstrap for a root that keeps only GNUmakefile/. Beside a kept Makefile, it prints #480's NOTE that the targets "arrive only once the follow-up merges the two". Rendered by the lane's probe_texts.py on main.
3, python3	The template's scripts/bootstrap.py says python3 scripts/bootstrap.py (python on Windows). #480's NOTE and the Makefile collision follow-up's "until then" name python3 alone.
NIT-2	The no-reads_first script-aside tail says "neither make bootstrap nor python3 scripts/bootstrap.py runs the shape's bootstrap before both merges are made". In a root that keeps its own Makefile and scripts/bootstrap.py, merging the script alone makes python3 scripts/bootstrap.py print the shape's SHAPE-BOOT, while make bootstrap still exits 2. Measured again by a row below.
NIT-3a	_predict_collisions says the wording comes from "the same function execute reports its actual collisions through", and collision_follow_up says "Both callers say it". execute never calls collision_follow_up: it carries the plan's stored follow_ups: into the split message.

What changed

adopt-project.py:

• _make_reads_first(surviving) gives the first of GNUmakefile and makefile the root keeps. It reads GNU_MAKE_READS' order, asks the way _root_keeps asks, and adds a / for a DIRECTORY. _make_reads_first_clause words the clause. For a file, the words are main's, byte for byte. For a directory, it says: "GNU make at the root stops at the source's directory GNUmakefile/, the first of GNUmakefile, makefile and Makefile it finds (GNUmakefile: Is a directory.), so ... only once the root keeps no directory of that name".
• _makefile_follow_up (the CONTRACTS_DIR line) asks both. A kept file's line is unchanged: tests/test_adopt_make_reads_first.py passes as it is.
• _next_steps takes surviving=, the plan's root paths, which cmd_execute passes as paths_for["root"], as adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121 does.
  • Where a kept DIRECTORY stops make, the command is python3 scripts/bootstrap.py. The NOTE says GNU make stops there. Beside a kept Makefile, it also says the targets run there once the follow-up merges the two, and only once the directory is gone. Where the script went aside too, alone or with the Makefile, the NOTE names that command after the merges.
  • Since the merge of adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121, a kept FILE goes through the same lookup and the same clause. For a file the clause gives adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's words byte for byte, and its NOTEs end in BOOTSTRAP_COMMAND too.
• BOOTSTRAP_COMMAND is in every NOTE: adopt-project.py: a source with its own Makefile keeps it, so execute's NEXT make bootstrap (and every shape make target) fails in the adopted root until a human merges shape/Makefile #480's, adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's and the directory's. Each command line a person pastes stays python3 scripts/bootstrap.py.
• _predict_collisions' docstring says what the plan stores and what execute carries (NIT-3a).

scripts/shape_materialize.py:

• BOOTSTRAP_COMMAND = "python3 scripts/bootstrap.py (python on Windows)", as the template's script names it.
• The Makefile collision follow-up says "until then python3 scripts/bootstrap.py (python on Windows) is what make bootstrap runs".
• Where the script went aside too, the follow-up says "so python3 scripts/bootstrap.py (python on Windows) runs the shape's bootstrap once the script is merged, and make bootstrap once both merges are made" (NIT-2).
• The docstring's "Both callers say it" paragraph says what is true (NIT-3a's twin, 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's review NIT-4). 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 made the same fix, and the merge keeps main's text (see "Main merge after 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").

templates/assembly-root/:

• AGENTS-shape.md: the "Work from here" paragraph adds "; where the root keeps a GNUmakefile or a makefile, which GNU make reads before Makefile, run python3 scripts/bootstrap.py (python on Windows), which is what the shape's make bootstrap runs." A sentence after the five-commands table says they are the shape's Makefile's targets, which make at the root reaches only where the file it reads first holds them, or includes a Makefile that does.
• README.md: one sentence after Get started covers make bootstrap "here and below". A second says make park and make resume below work there only where that file holds them, or includes a Makefile that does.
• This is a template byte change, so every project re-pins. No digest of either file is stored in this repository. tests/ was grepped for every changed line, and only this PR's own rows pin them (see "Template pins").

Tests:

• tests/test_adopt_e2e.py: two pins move to the new words, at :312 (the "until then" sentence) and at :447-450 (NIT-2's tail).
• tests/test_adopt_next_make_reads_first.py (adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's): the merge moves three pins, OWN_MAKEFILE_NOTE, RUNS and script_aside, to python on Windows. adopt-project.py: a source with its own Makefile keeps it, so execute's NEXT make bootstrap (and every shape make target) fails in the adopted root until a human merges shape/Makefile #480's NOTE merged with no textual conflict, so without them its rows fail.
• The new file tests/test_adopt_make_bootstrap_followups.py has 35 rows:
  • the template's sentences, held to its Makefile (PYTHON ?= python3, bootstrap: runs $(PYTHON) scripts/bootstrap.py) and to scripts/bootstrap.py's own words;
  • make -n bootstrap against the template's Makefile beside each kept name, each row asserting the AGENTS-shape.md caveat that measurement backs;
  • _make_reads_first (10 roots);
  • the CONTRACTS_DIR line for 4 directory roots;
  • NEXT for a directory, beside a kept Makefile (both names), after both merges, and after the script's merge alone;
  • adopt-project.py: a source with its own Makefile keeps it, so execute's NEXT make bootstrap (and every shape make target) fails in the adopted root until a human merges shape/Makefile #480's NOTE, called both as main's signature allows and with surviving=;
  • the collision follow-up's two sentences;
  • NIT-2 measured with the template's Makefile and stand-in scripts, and tied to the follow-up's new tail;
  • three real adoptions of a source holding GNUmakefile/rules.mk. One is kept by a --path-policy rule, so plan writes the line, check prints it and the split message carries it. One is answered root. In both, NEXT is asserted, make -n bootstrap in the pasted clone exits 2 with GNUmakefile: Is a directory., and the printed command exits 0. In the third, the question is answered code: NEXT names make bootstrap, and that command runs in the pasted clone.

This round (#1207's review, issuecomment-6374516382)

finding	what changed
SHOULD-1	A row pins NEXT where only scripts/bootstrap.py went aside beside a kept GNUmakefile/ (test_next_after_the_script_merge_names_python3_where_make_stops). It fails under the review's mutant X1.
SHOULD-2	The "Left" list is one issue, #1240.
SHOULD-3	"Landing order", below. #1121 landed while this round ran, and is merged and resolved here.
NIT-1	AGENTS-shape.md's sentence after the table loses "(above)": nothing above names a Makefile that includes another.
NIT-2	README.md names make park and make resume.
NIT-3	Beside a kept Makefile, the directory NOTE names the merge as well as the directory. Its row is parametrized over GNUmakefile/ and makefile/.
NIT-4	An adoption sends GNUmakefile/ to the code leg, and NEXT names make bootstrap. It fails under the review's mutant X6.
NIT-5	The notes for the squash message at landing, below.

Rows that fail on main

Measured at the merge head 6fa1edd56, in a scratch worktree:

• main 0ed3e57f8's bytes of the four production files (adopt-project.py, scripts/shape_materialize.py, templates/assembly-root/AGENTS-shape.md, templates/assembly-root/README.md) were checked out over the head;
• the new file (35 rows) and tests/test_adopt_next_make_reads_first.py (40 rows) ran through the lane's runner;
• the four files were restored with git checkout, and each matched the head by sha256 (c94bd19b, 12555f39, 8660bafb, b3f9409b).

file	fail on main	pass on main
the new file	34	1: test_an_adoption_sending_a_gnumakefile_directory_to_a_leg
tests/test_adopt_next_make_reads_first.py	19	21

Main now has #1121, so its _next_steps takes surviving=. The NEXT rows for a kept directory therefore fail on their words, not on a TypeError. Main reads GNUmakefile/ as a file: "then run python3 scripts/bootstrap.py; GNU make at the root reads the source's GNUmakefile, ...". This is the case for the SHOULD-1 row, the after-the-merges row, both directory rows, and NIT-3's makefile/ row.

The new file's other rows fail on main for these reasons:

• the template rows: main's AGENTS-shape.md and README.md lack the sentences;
• _make_reads_first: main has no such helper;
• the CONTRACTS_DIR line: "only once GNUmakefile does";
• the NOTE and collision rows: plain python3, and NIT-2's old tail;
• the two adoptions that keep the directory.

The one new row that passes on main is NIT-4's. It sends GNUmakefile/ to the code leg, and there main's NEXT is already make bootstrap. It guards what main has: under the review's mutant X6, which hands NEXT every leg's paths, it is the row that fails.

The 19 rows of #1121's file that fail on main are exactly the rows that use the three pins this merge moved (OWN_MAKEFILE_NOTE, RUNS, script_aside). The 21 that pass use none of them.

Earlier heads, on main 961640c68 and with the same method: 1c7ca5c6d failed 29 of 35 rows, and 8b499842f failed 34 of 35 (the new file plus the three tests/test_adopt_e2e.py rows the mutants use).

Mutants

This round, at 75c90b6db, in a scratch worktree. Each mutant was applied alone and restored, and each file's sha256 was checked against the commit (4/4 every time). Each ran against 95 rows: the new file (35), tests/test_adopt_e2e.py (27) and tests/test_adopt_make_reads_first.py (33). Unmutated, all 95 pass.

id	mutant	rows failing (of 95)	result
X1	the review's: the script-only NOTE names make bootstrap after the merge (if stops and makefile)	1	killed
X6	the review's: cmd_execute passes every leg's paths	1	killed
N1	NIT-3 undone: beside a kept Makefile, no merge named	2	killed
N2	the merge named where the script went aside too	1	killed
N3	the merge named where no Makefile was kept	4	killed
T1	NIT-1 undone: "(above)" back	1	killed
T2	NIT-2 undone: README's make park / make resume sentence removed	1	killed

The merge with #1121 rewrote the clause code that N1-N3 mutate (stopped became reads, and the merge condition gained stops). The mutants were not run again at 6fa1edd56.

Earlier, 18 mutants at 1c7ca5c6d against 43 rows, each applied alone in a scratch worktree and restored with its sha256 checked:

id	mutant	rows failing (of 43)	result
M1	_make_reads_first drops the / of a directory	14	killed
M2	the clause's directory branch never taken	10	killed
M3	NEXT ignores a kept directory (stops = None)	6	killed
M4	NEXT's command stays make bootstrap for a directory	4	killed
M5	cmd_execute passes surviving=()	2 (the two adoptions)	killed
M6	BOOTSTRAP_COMMAND without (python on Windows)	12	killed
M7	NIT-2's tail back to "neither ... before both merges are made"	2	killed
M8	AGENTS-shape.md's "Work from here" caveat removed	1	killed
M9	the sentence after the five-commands table removed	1	killed
M10	README.md's sentence removed	1	killed
M11	#480's NOTE back to plain python3	1	killed
M12	the collision follow-up's "until then" back to plain python3	2	killed
M13	after the merges, NEXT says make bootstrap where a directory stops make	1	killed
M14	the lookup takes the LAST kept name (reversed)	4	killed
M15	the directory NOTE's lead ignores a kept Makefile	1	killed
M16	the lookup folds case (fold=True)	4	killed
M17	the clause drops GNU make's own words (GNUmakefile: Is a directory.)	10	killed
M18	the file clause says does whatever has is	0	survived

M18 was equivalent before the merge: the file clause's only caller was the CONTRACTS_DIR line. Since the merge, #1121's NOTE calls it with has them, and tests/test_adopt_next_make_reads_first.py pins that wording. M18 was not run again.

Runs

Each run went through the lane's runner: gh stubbed first on PATH with a bogus token, GIT_SSH_COMMAND=false, GIT_CONFIG_GLOBAL=/dev/null, EMAIL set, -p no:cacheprovider, and tmp_path_retention_policy=failed. Each run is on a commit, with a clean tree.

run	commit	git	files	passed	failed
gate M	6fa1edd56, the merge	2.43.0	the new file, test_adopt_next_make_reads_first.py, test_adopt_e2e.py, test_adopt_make_reads_first.py, test_path_guard_hygiene.py, test_repo_hygiene.py, test_agents_md_codes.py, test_cli_reference.py, test_pinned_copy_text.py, test_scaffold_leg_path_collisions.py	618	0
gate N	6fa1edd56	2.47.1	the new file and test_adopt_next_make_reads_first.py (75), then test_adopt_e2e.py (27), run one after the other	102	0
gate H	6fa1edd56	2.47.1	the hygiene five	379	0
gate A2	75c90b6db	2.43.0	the new file, test_adopt_e2e.py, the hygiene five (test_path_guard_hygiene.py, test_repo_hygiene.py, test_agents_md_codes.py, test_cli_reference.py, test_pinned_copy_text.py), test_adopt_make_reads_first.py, test_adopt_contracts_dir.py, test_adopt_printed_sentences.py, and the template's readers: test_scaffold_e2e.py, test_scaffold_free_text_values.py, test_free_plan_hint.py, test_template_quoted_values.py, test_template_values_single_pass.py, test_template_text_placeholders.py, test_bootstrap_placement_texts.py, test_park_resume_targets.py, test_update_shape_hardlink.py	1128	0
gate A	1c7ca5c6d	2.43.0	17 files (the new file, test_adopt_e2e.py, the hygiene five, test_scaffold_leg_path_collisions.py, test_adopt_make_reads_first.py, test_adopt_contracts_dir.py, test_adopt_printed_sentences.py, test_adopt_attributes_eol_block.py, test_adopt_leg_path_collisions.py, test_scaffold_e2e.py, test_bootstrap_placement_texts.py, test_park_resume_targets.py, test_template_text_placeholders.py)	966	0
gate B	1c7ca5c6d	2.47.1	the new file, test_adopt_e2e.py, test_path_guard_hygiene.py, test_repo_hygiene.py	209	0
doctor	1c7ca5c6d	2.43.0	test_shape_doctor.py, test_scaffold_leg_path_collisions.py	549 (3 skipped: the PowerShell rows, which run only on a Windows host)	0

Template pins

tests/ was grepped for every line this PR changes in templates/assembly-root/ ("They are the shape", "reads first holds", "includes a Makefile that", "(above)", "here and below", "Where this root keeps", "it is what the shape", "before Makefile", "python on Windows"). Only tests/test_adopt_make_bootstrap_followups.py pins them. Four other files read the README as a whole, and each passes in gate A2:

• test_scaffold_free_text_values.py counts its lines against the template;
• test_free_plan_hint.py looks for "Free plan" and --repo;
• test_template_quoted_values.py and test_template_values_single_pass.py render it.

Merges

git merge-tree --write-tree of the head 01e93b6d1, on git 2.43.0, at 18:11Z.

• Main 9729315f1, fetched again at 18:11Z, is the main this round merged last. The merge-tree is clean, and its tree is the head's own, f6488542f.
• Main moved to 801a7d6d0 at 18:14Z, when lane 3's adopt-project.py: under a one-byte locale such as Latin-1, check reads a plan written in UTF-8 whose source is there, asking the plan's UTF-8 bytes where the spelling as written names nothing, rather than refusing plan-source-gone (#1078) #1106 landed (adopt-project.py and two locale test files). The merge with it is clean, tree 25bc27d39, and gate P4 passed on it. It was not merged in: it touches none of the code this round resolves, and landing refreshes the branch anyway.
• The open PRs: one gh pr list --state open --limit 100 --json number,headRefName,headRefOid,files read at 18:11Z returned 51 open PRs. 26 of them touch one of this PR's paths. That includes tests/test_adopt_collision_make_reads.py and tests/test_adopt_make_reads_first.py, which this PR now changes. 25 merge cleanly: adopt-project.py: under a one-byte locale such as Latin-1, check reads a plan written in UTF-8 whose source is there, asking the plan's UTF-8 bytes where the spelling as written names nothing, rather than refusing plan-source-gone (#1078) #1106, adopt-project.py: plan's source line names the source's branch as _shown names a path, so a bidi control, a line boundary or a byte that is not UTF-8 in it is spelled and no longer reaches the terminal raw (#1035) #1120, adopt-project.py: an org/repo --source too long for the disk is refused as a path is, naming no GitHub repository and offering no url, so --source <300 a's>/Thing is no longer offered https://github.com/<300 a's>/Thing, which can never clone (#1003) #1122, adopt-project.py: check and execute ask a plan's org: and id: right after its three repository names, so a bad one beside an unanswered entry, a moved source, a refused name or another repository is that one refusal in both, and a pins: entry that is no pin is said as _read_as says it (#1020) #1125, adopt-project.py: a recorded source whose .git this user cannot read is refused source-unreadable, naming that .git, so check and execute no longer call a clone plan made with its .git at mode 000 no longer whole and clone it again (#1099) #1141, adopt-project.py: a --work-dir link into a loop that Python 3.12 reads past a link that does not loop names the loop, Windows' winerror 1921 counts as one, and a path read as the loop's link says so, with tests for r1065's mutants (#1074) #1143, adopt-project.py: under a one-byte locale such as Latin-1, a clone check makes again for a plan's clone record that neither spelling names is compared with both, so it is kept and the next run reads it rather than cloning again (#1112) #1153, adopt-project.py: plan written to and cloning <repository> (read only) into name the caller's --out and --work-dir with _shown, as the other lines of the report name a path, so a bidi control, a line feed or a tab in either no longer reaches the terminal as it was typed (#1115) #1154, adopt-project.py: check names each case twin at the shape's copy, above it and at the manifest, before a link there or after it, and at a pin's file whose exact spelling holds nothing, as _KeptTree keeps every row, not the first (#1138) #1170, 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, adopt-project.py: the --plan path in check and execute, and the --local-remote-dir path in execute's bare, leg and split lines, are named with _shown, as #1154 names --out and --work-dir, so a bidi control, an escape, a line feed or a byte that is not UTF-8 in either no longer reaches the terminal as it was typed (#1152) #1177, 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, adopt-project.py: a --source, or a plan's source.local_path, with a part of its repository git reads at mode 000 (the .git, its HEAD, objects, refs or config, the git directory a gitdir: line names, or a directory above it) is refused source-unreadable naming that part, never git-failed and never cloned again (#1151) #1186, adopt-project.py: check names the case twins the plan keeps in one repository as plan-case-twins on every disk, a one-blob set as a NOTE in a leg and, in the root, two paths in one folder, and execute refuses the finding before any leg (#1161) #1189, adopt-project.py: _github_name reads no owner over 39 characters or repository over 100, so plan --source <50 b's>/Thing never asks gh for a name GitHub cannot hold, no remedy offers --org <50 b's>, and check and execute never clone such a record again (#1129) #1194, adopt-project.py: _pin_row_problem says a pin field whose value is not text as _read_as says what the plan's reader read, "records commit: as the number 7", where it spelled it as Python does, commit: 7 and True, so check and execute name it in the words the plan's other refusals use under the same plan-pin-malformed (#1114) #1198, adopt-project.py: check and execute read one list of a plan's problems against its source, in the order a person mends them (another repository, a refused name, a moved source, an unanswered entry), so check's first finding is execute's refusal for each pair (#1113) #1210, adopt-project.py: a refused command's REFUSED block spells each character a terminal acts on in the command, its directory and what it printed, as _first_said spells gh's, so --local-remote-dir no longer reaches the terminal raw there; a NOTE before NEXT: says when the source's path in its clone command holds a character a terminal does not show as itself; and rows pin six sites #1177's 148 left to chance (#1195) #1225, shape_materialize.py: the docstring of gives_a_branch_another_name says a full name that is no pattern takes the one ref git completes it to, which may be a branch, where it said any other source names no branch (#1238) #1244, repo_shape.py and adopt-project.py: a plain integer of more digits than this Python reads, or a decimal too large to be finite, is yaml-unparsable by its line, where it was a ValueError traceback or "the number inf", and _read_as says a float in digits with no exponent, where it said "the number 1e+300" (#1212) #1249, adopt-project.py: execute unsets the leg clone's remote.origin.url before git filter-repo, so under a Latin-1 locale a source under a directory named with the byte 0xE9 is extracted, where filter-repo died printing that url as U+FFFD, and every branch of the source is still rewritten as before (#1247) #1251, adopt-project.py: a source whose refs this user cannot list is refused source-unreadable naming the directory, never source-default-branch-not-local "(its branches: none)" (#1219) #1252, adopt-project.py: check's compare for the mount's git rm skips the .gitattributes it leaves and reads its exit code, so the root's failing filter is plan-attributes-reencoded alone, and a refused compare is named in git's words (#1192) #1253, adopt-project.py: a stale split branch in a source refs directory this user cannot list is refused source-unreadable by name, never plan ok from check or two legs made by execute (#1257) #1265 and adopt-project.py: each clone execute makes names its remote origin, whatever clone.defaultRemoteName says, so execute no longer ends in REFUSED mounting the legs ("No such remote 'origin'") after both legs are pushed (#1261) #1268. adopt-project.py: under a one-byte locale such as Latin-1, check reads a plan written in UTF-8 whose source is there, asking the plan's UTF-8 bytes where the spelling as written names nothing, rather than refusing plan-source-gone (#1078) #1106 has landed since (above).
• Lane 1's adopt-project.py: check names the case twins the plan keeps in one repository as plan-case-twins on every disk, a one-blob set as a NOTE in a leg and, in the root, two paths in one folder, and execute refuses the finding before any leg (#1161) #1189 (472944b5d, read fresh at 18:11Z) edits three test files that this round changes too: tests/test_adopt_next_make_reads_first.py, tests/test_adopt_collision_make_reads.py and tests/test_adopt_make_reads_first.py. The merge is clean, tree 7fd030b29. adopt-project.py: check names the case twins the plan keeps in one repository as plan-case-twins on every disk, a one-blob set as a NOTE in a leg and, in the root, two paths in one folder, and execute refuses the finding before any leg (#1161) #1189's hunks are in the imports, in the clone rows' parameter lists and docstrings, and in a new refusal row, away from the moved pins, the makefile-directory row and NIT-A's comment. On that predicted merge, a local commit object, the three shared files pass on git 2.43.0: tests/test_adopt_make_reads_first.py and tests/test_adopt_collision_make_reads.py give 62 passed, and tests/test_adopt_next_make_reads_first.py gives 38 passed. There are fewer rows than at this head because adopt-project.py: check names the case twins the plan keeps in one repository as plan-case-twins on every disk, a one-blob set as a NOTE in a leg and, in the root, two paths in one folder, and execute refuses the finding before any leg (#1161) #1189 leaves out the clone rows that keep case twins.
• adopt-project.py: the mount removes a path a case twin is at or under with git rm -f and writes the twins the root keeps back from the index, so on a disk that ignores case execute keeps a set the plan splits across repositories or drops (#1183) #1237 (b557859fd) conflicts in tests/test_adopt_check_throwaway_repos.py, a file this PR does not change. Main conflicts with adopt-project.py: the mount removes a path a case twin is at or under with git rm -f and writes the twins the root keeps back from the index, so on a disk that ignores case execute keeps a set the plan splits across repositories or drops (#1183) #1237 in that file too, while 6fa1edd56 did not, so the conflict is main's, carried in by the merge.
• Main moved again after the round (rv1207e, at 19:25Z): lane 1's adopt-project.py: a --work-dir link into a loop that Python 3.12 reads past a link that does not loop names the loop, Windows' winerror 1921 counts as one, and a path read as the loop's link says so, with tests for r1065's mutants (#1074) #1143 landed as 81eaf757c (adopt-project.py) and lane 4's family.py: the remove that both member-duplicate-id remedies name and the add that bump's member-unknown names carry --family-root <holder>, quoted for the host, so each runs as printed (#1005) #1117 as 9252ef208 (README, the handbook and scripts/family.py). The head merges clean with both (trees e40373967 and 279f8c6ce), and rv1207e's gate P passed on the predicted merge with 9252ef208 (359 passed). At 19:22Z, 48 PRs were open, 24 touch this PR's files, and 23 merge clean; adopt-project.py: the mount removes a path a case twin is at or under with git rm -f and writes the twins the root keeps back from the index, so on a disk that ignores case execute keeps a set the plan splits across repositories or drops (#1183) #1237's conflict is main's, as above.
• No longer on the list: 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, adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121, 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 and README.md, adopt-project.py and the handbook: a kept directory entry makefile/ is tested, README says make at the root reads a kept GNUmakefile or makefile instead, and a docstring says the shape's Makefile alone holds CONTRACTS_DIR (#1061) #1123 are on main. Since 6fa1edd56, 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, 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, adopt-project.py: check's rehearsal and _stored_as_written take a parent's -c settings, attr.tree too, and the source's objects, GIT_ATTR_SOURCE is refused by name before any question, and each throwaway reads its environment first (#1136) #1169, adopt-project.py: check and execute run with GIT_NO_LAZY_FETCH=1 and refuse a partial-clone source that lacks an object its branches, tags or HEAD reach as source-partial-clone, so neither fetches into the source any more (#1145) #1180, adopt-project.py: plan runs with GIT_NO_LAZY_FETCH=1, refuses a partial-clone --source that lacks objects as source-partial-clone, and names a partial --pin-source it could read only by fetching and the fetch that makes it whole (#1162) #1190 and scaffold-project.py: a partial-clone --pin-source short of the pinned commit or its trees is pin-source-unreadable, named a partial clone with git fetch --refetch --no-filter <remote> <commit>, and no longer fetched into (#1182) #1223 landed too.

Landing order

The conflicts are the same in every order, measured offline both ways round and with predicted squashes:

• adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121: 4 hunks, 84 lines, in adopt-project.py's _next_steps and cmd_execute's comment.
• 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: 3 hunks, 65 lines, in scripts/shape_materialize.py's collision_follow_up: its docstring's NIT-4 sentence, its paragraph on the Makefile's merge, and its two returns.
• adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121 and 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 merge with each other cleanly, and so do 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 this PR.

So the order decides only who resolves them. Both have landed, and lane 5 resolved both on this branch:

• adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121 (merge 6fa1edd56). The resolution is one lookup and one clause, BOOTSTRAP_COMMAND in adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's NOTEs, and adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's three test pins moved.
• 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 (merge 7b0192bde, see "Main merge after 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"). The resolution keeps 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's reads_first branch and runs BOOTSTRAP_COMMAND through it. NIT-2's tail is this PR's where make reads the merged Makefile, and 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's where make reads another file. Four of 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's test pins moved: MERGE_IS_ENOUGH, BOTH_MERGES, MEANWHILE and SCRIPT_MERGED.

Main was merged in once more after lane 1's #1123 (merge 01e93b6d1): one of #1123's rows pinned the line this PR changes for a kept makefile/ directory, and it moved (see "The second merge, after #1123").

No other PR has to land first now. 6fa1edd56 was re-checked READY (rv1207b, rv1207c and rv1207d, issuecomment-6722602375). The head, 01e93b6d1, needs a check of its delta since then: the two merges' resolutions, the four moved pins in #1105's file, #1123's moved row, and NIT-A.

Left

#1240 tracks what this PR leaves:

• adopt-project.py and the assembly-root template: follow-ups from RV1105's and RV1121's reviews of #1105 and #1121 (make bootstrap in the template, a kept GNUmakefile/ directory, python3 on Windows, GNU_MAKE_READS twice, stale follow-ups) #1132's item 4 (one GNU_MAKE_READS, one lookup);
• 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's collision line, for a kept directory (its python3 part is done here: see "Main merge after 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");
• adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's file NOTE beside a kept Makefile, which does not name the merge (as this PR's directory NOTE now does);
• 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's second review's NIT-1, NIT-3b, NIT-3c, NIT-4 and NIT-5;
• _make_reads_first_clause's has argument, now live;
• rv1207e's NIT-1: the docstring sentence "With none, the Makefile's follow-up is what it was, byte for byte" (scripts/shape_materialize.py about :1067), kept from 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, is no longer true, since this PR changes the line with no reads_first. It sits beside adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240's items F and G, which reword the same docstring.

The local branch fix/1132-template-bootstrap-stacked (9eb5cce0b, never pushed and never tested) is a model of the end state on #1105 + #1121 + main.

For the squash message at landing

• The first commit, 1c7ca5c6d, says "Left, as adopt-project.py and the assembly-root template: follow-ups from RV1105's and RV1121's reviews of #1105 and #1121 (make bootstrap in the template, a kept GNUmakefile/ directory, python3 on Windows, GNU_MAKE_READS twice, stale follow-ups) #1132's body names". It is this PR's body that names them, and they are now adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240. The squash message should say "Left: adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240".
• Sourcery's summary below says "Centralize GNU make precedence and bootstrap command wording". Only the bootstrap command's wording is in one place (BOOTSTRAP_COMMAND). GNU make's order is still in two tuples once 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 lands, which is item 4, left to adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240. The squash message should not say "centralize" of it.
• It should say that adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121 was merged in, and that tests/test_adopt_next_make_reads_first.py's three pins moved to python on Windows.
• It should say that main was merged in again after 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 and after README.md, adopt-project.py and the handbook: a kept directory entry makefile/ is tested, README says make at the root reads a kept GNUmakefile or makefile instead, and a docstring says the shape's Makefile alone holds CONTRACTS_DIR (#1061) #1123, that collision_follow_up keeps 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's reads_first lines with BOOTSTRAP_COMMAND in them, that four pins in tests/test_adopt_collision_make_reads.py moved with them, and that README.md, adopt-project.py and the handbook: a kept directory entry makefile/ is tested, README says make at the root reads a kept GNUmakefile or makefile instead, and a docstring says the shape's Makefile alone holds CONTRACTS_DIR (#1061) #1123's makefile-directory row now expects the line that says GNU make stops at a kept makefile/.

Limits

• Item 5, follow-ups that go stale after a hand edit of the plan, is its own issue, filed as adopt-project.py: plan writes its follow-ups once, so after a hand edit of the plan check and the split commit name a GNUmakefile the root no longer keeps, or say nothing of one it now keeps #1206. It also holds the three related cases this work found:
  • a kept makefile/ directory gets "until the two are merged";
  • a directory kept only through deeper entries is not seen by _root_keeps;
  • a kept Makefile/ directory is treated as a Makefile.
• This PR's answered-root row meets item 5 too. plan asks about GNUmakefile/, and its stored CONTRACTS_DIR line, written before the answer, names no directory. The row asserts what NEXT and make do there, and not that line.
• Not measured: a disk that ignores case, and GNU make other than 4.3.

Closes #1132

Lane: openreposhape-5 (openRepoShape-5)

🤖 Generated with Claude Code

Summary by Sourcery

Clarify bootstrap and GNU make behavior across assembly-root templates and adoption workflows, including kept GNUmakefile/makefile directories and Windows command guidance.

New Features:

• Document the correct bootstrap command and make-target behavior for assembly roots containing earlier GNU make file names or directories.
• Detect kept GNUmakefile/ and makefile/ directories and provide accurate adoption follow-up and NEXT guidance.

Bug Fixes:

• Correct bootstrap guidance for Windows by identifying python as the interpreter alternative.
• Fix collision follow-ups describing when the shape bootstrap becomes runnable after script and Makefile merges.
• Correct collision-plan documentation to reflect how follow-ups are stored and reported.

Enhancements:

• Share GNU make precedence and bootstrap command wording across adoption planning, execution guidance, and materialization follow-ups.

Documentation:

• Update assembly-root AGENTS and README templates with caveats for GNU make precedence and Windows bootstrap invocation.

Tests:

• Add comprehensive coverage for template guidance, GNU make file and directory precedence, adoption follow-ups, NEXT output, Windows wording, and real directory-collision adoptions.
• Update existing expectations for unified bootstrap command wording and directory handling.

[sourcery-ai]

Reviewer's Guide

Implementation

Define the shared bootstrap command and update collision follow-ups to describe accurately which command works after each merge, including the script-only case.

• scripts/shape_materialize.py L435-440, L1036-1048, L1138-1146 (+2 small edits)

Detect the first GNU make precedence name retained at the root, distinguish directories from files, and use that distinction in the CONTRACTS_DIR follow-up.

• adopt-project.py L318-321, L3806-3852, L3879-3884 (+1 small edits)

Correct the collision-prediction comment to describe how planned follow-ups are stored and carried into the split commit.

• adopt-project.py L3911-3913, L6600-6607, L16285

UI

Make NEXT choose a runnable bootstrap command and explain GNU make’s behavior when a retained precedence name is a file or directory.

• adopt-project.py L16900-16927, L16943-16947, L16957-16964 (+1 small edits)

Documentation

Explain in the assembly-root guidance when to run the bootstrap script instead of make, and when the shape’s make targets are reachable.

• templates/assembly-root/AGENTS-shape.md L19-21, L108-112
• templates/assembly-root/README.md L20-25

Tests

Update existing adoption expectations for the revised follow-up wording and Windows interpreter guidance.

• tests/test_adopt_collision_make_reads.py L85-97 (+3 small edits)
• tests/test_adopt_e2e.py L297-300, L441-445 (+2 small edits)
• tests/test_adopt_make_reads_first.py L103-112, L195-198 (+2 small edits)
• tests/test_adopt_next_make_reads_first.py L64-70, L79-80, L119

Add coverage for template instructions, GNU make file and directory precedence, NEXT and collision follow-ups, Windows command wording, and real adoption scenarios.

• tests/test_adopt_make_bootstrap_followups.py L1-562

---

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]

Independent review of #1207 (#1132) for lane openRepoShape-5

NOT READY

There is no MUST. On every path I measured, the code does what the title and body say. Three SHOULDs need action before landing. SHOULD-1 needs one new test row, so the head moves. SHOULD-2 is a body or issue edit. SHOULD-3 is a landing-order question. With those done, and CI green, I would call it READY.

Findings

SHOULD-1: the script-only-aside branch with a kept directory has no row (mutant X1 survives)

• Where: adopt-project.py:16178-16179, the if stops inside the if script: NOTE.
• What I changed: in my own worktree, mutant X1 turned if stops else " first, and then run \make bootstrap`."intoif stops and makefile else ...`.
• Result: all 92 rows still pass (92 passed in 104.43s).
• What the mutant breaks: a root that keeps only its own scripts/bootstrap.py plus GNUmakefile/ is told "merge shape/scripts/bootstrap.py into scripts/bootstrap.py first, and then run make bootstrap." GNU make 4.3 stops at that directory: make: *** GNUmakefile: Is a directory. Stop., rc 2.
• What the head prints now (correct): "NOTE the root keeps its own scripts/bootstrap.py, and the shape's is shape/scripts/bootstrap.py, so neither make bootstrap nor python3 scripts/bootstrap.py runs the shape's bootstrap yet: merge shape/scripts/bootstrap.py into scripts/bootstrap.py first, and then run python3 scripts/bootstrap.py (python on Windows); GNU make at the root stops at the source's directory GNUmakefile/, ... only once the root keeps no directory of that name."
• Why it matters: the body says "where the script went aside too, it names that command after the merges". Only the case where both files went aside is pinned (test_next_after_the_merges_names_python3_where_a_directory_stops_make, :280-295).
• Fix: add a row, or a parametrize of that row:

  said = next_said(adopter, capsys,
                   [("scripts/bootstrap.py", "shape/scripts/bootstrap.py")],
                   ["scripts/", "GNUmakefile/"])
  assert said == (f"{HEAD}NOTE the root keeps its own scripts/bootstrap.py, and the "
      "shape's is shape/scripts/bootstrap.py, so neither `make bootstrap` nor "
      "`python3 scripts/bootstrap.py` runs the shape's bootstrap yet: merge "
      "shape/scripts/bootstrap.py into scripts/bootstrap.py first, and then run "
      f"{COMMAND}; {stops('GNUmakefile', RUN_THERE)}.\n")

SHOULD-2: Closes #1132 closes the issue while the body's "Left" items have no open issue

• Where: line 171 of the live body, and commit 1c7ca5c6d.
• Measured: gh pr view 1207 --json closingIssuesReferences gives [1132].
• What would be left untracked: item 4, NIT-1, NIT-3b, NIT-3c, NIT-4, NIT-5, and the parents' own instances of items 1-3.
• Where they are recorded now: adopt-project.py: plan writes its follow-ups once, so after a hand edit of the plan check and the split commit name a GNUmakefile the root no longer keeps, or say nothing of one it now keeps #1206 holds item 5 only (its body read in full). The hand-over is recorded only in lane 4's docs/release-notes/v1.0.0.md: the v1.0.0 release notes, a draft for Brett to amend before he tags, saying what 1.0 contains and promises, how to pin it by commit, what changed on the way grouped by the five tracking areas, and every known limit as the input that triggers it and what the person sees (#472) #473 issuecomment-6350568104 ("Your list above is the record of what moves to lane 5"). After the merge, nothing open tracks those items.
• Fix, before landing: do one of these:
  • file the "Left" list as a new issue and name it in the body and in the squash message;
  • or change the line to "Part of adopt-project.py and the assembly-root template: follow-ups from RV1105's and RV1121's reviews of #1105 and #1121 (make bootstrap in the template, a kept GNUmakefile/ directory, python3 on Windows, GNU_MAKE_READS twice, stale follow-ups) #1132" and close adopt-project.py and the assembly-root template: follow-ups from RV1105's and RV1121's reviews of #1105 and #1121 (make bootstrap in the template, a kept GNUmakefile/ directory, python3 on Windows, GNU_MAKE_READS twice, stale follow-ups) #1132 by hand when the left items land.

SHOULD-3: the landing order against #1105 and #1121 is not stated, and it conflicts both ways

• Measured with git merge-tree --write-tree:
  • against 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 b28f40de0: CONFLICT (content): Merge conflict in scripts/shape_materialize.py;
  • against adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121 1c5173758: CONFLICT (content): Merge conflict in adopt-project.py.
• What lane 4 plans: docs/release-notes/v1.0.0.md: the v1.0.0 release notes, a draft for Brett to amend before he tags, saying what 1.0 contains and promises, how to pin it by commit, what changed on the way grouped by the five tracking areas, and every known limit as the input that triggers it and what the person sees (#472) #473 issuecomment-6350568104 says they land "as reviewed ... with no fix round (adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121 first)".
• Why it matters:
  • If adopt-project.py, shape_materialize.py and the assembly-root template: a kept GNUmakefile/ is a directory GNU make stops at, python3 scripts/bootstrap.py names python on Windows, and the template says what to run where make reads another file #1207 lands first, 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 and adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121 need the fix round lane 4 says it will not do.
  • If they land first, 8b499842f is superseded, and this review does not carry over to the resolved head.
• Fix: say in the body which lands first, and who resolves the conflict. If it is the parents, plan a new head and a re-review here.

NIT-1: "(above)" points at nothing above

• Where: templates/assembly-root/AGENTS-shape.md:102-105.
• Measured: "...only where the file it reads first holds them, or includes a Makefile that does (above)". Nothing above says "includes a Makefile". "Work from here" (:19-21) names the files but not the include.
• Fix: drop "(above)" or point it at "Work from here". Do it in this PR, which already makes every project re-pin. A later fix makes them re-pin again.

NIT-2: the README caveat covers make bootstrap but not the parking targets

• Where: templates/assembly-root/README.md:20-22.
• Measured: "instead, here and below" covers the two make bootstrap lines. The same parking block has make park (:35) and make resume (:38), which fail the same way at such a root. The README says nothing about them, while AGENTS-shape.md:102-105 does.
• Fix: add one clause saying those targets work only where the file GNU make reads first holds them, or includes Makefile. README.md has no shape-pin row, so this costs no re-pin.

NIT-3: beside a kept Makefile, the directory NOTE no longer mentions the merge

• Where: adopt-project.py:16180-16185.
• Measured by my probe: "NOTE the root keeps its own Makefile and the shape's is shape/Makefile, and GNU make at the root stops at the source's directory makefile/, ... so make bootstrap and the shape's other make targets run there only once the root keeps no directory of that name; ..."
• Why it matters: removing the directory is not enough. The source's Makefile still has no bootstrap rule until the two are merged. The CONTRACTS_DIR line at the same root does say both.
• Fix: where makefile is set, pass a reach like "make bootstrap and the shape's other make targets run there once the follow-up merges the two, and".

NIT-4: no row shows that NEXT reads only the root's paths (mutant X6 survives)

• Where: adopt-project.py:15530.
• Measured: mutant X6, surviving=paths_for["root"] + paths_for["code"] + paths_for["spec"], leaves all 92 rows passing.
• Fix (optional): an adoption row where GNUmakefile/ goes to a leg and NEXT still prints make bootstrap.

NIT-5: two descriptions of what is left are slightly off

• Commit message: 1c7ca5c6d says "Left, as adopt-project.py and the assembly-root template: follow-ups from RV1105's and RV1121's reviews of #1105 and #1121 (make bootstrap in the template, a kept GNUmakefile/ directory, python3 on Windows, GNU_MAKE_READS twice, stale follow-ups) #1132's body names". It is this PR's body that names them, not adopt-project.py and the assembly-root template: follow-ups from RV1105's and RV1121's reviews of #1105 and #1121 (make bootstrap in the template, a kept GNUmakefile/ directory, python3 on Windows, GNU_MAKE_READS twice, stale follow-ups) #1132's. Fix it in the squash message.
• Sourcery's summary: the summary Sourcery appended to the body says "Centralize GNU make precedence and bootstrap command wording". Item 4 is left, so that overstates it.

Checked and true

The PR and its body

• The live body equals $Q/v124/n1132/body-post2.md up to its Lane line (diff gave no difference).
• The head is 8b499842f, a draft on milestone v1.2.1, and it closes [1132].
• I read the whole diff of all 6 files.
• git diff --stat 1c7ca5c6d HEAD touches only tests/test_adopt_make_bootstrap_followups.py.
• At the head, the sha256 of the four production files match the body: 400730ad, 660962cb, 9ff37f19, 4ece14bb.

The code

• The file clause is main's text byte for byte: the same joint, reach and does.
• The directory clause and NEXT match the body for these cases:
  • a kept directory alone;
  • beside a kept Makefile;
  • after both merges;
  • script-only aside (probe).
• A kept FILE still prints main's make bootstrap (probe). That case is adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's.
• NIT-3a's docstrings are true. execute never calls collision_follow_up: git grep finds only _predict_collisions and Materialized.follow_ups, and nothing calls the latter.
• The _next_steps signature and surviving=paths_for["root"] match adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's head 1c517375 exactly.
• The python3 claim: in adopt-project.py, shape_materialize.py and the template, only two places keep plain python3: the pasted command line (:16152) and the "neither ... nor" statement (:16175). Both are as the body says.

GNU make 4.3, measured with the template's Makefile:

• GNUmakefile/ and makefile/ each give rc 2 and "Is a directory. Stop.".
• GNUmakefile/ plus a makefile file gives "GNUmakefile: Is a directory."
• makefile/ plus a GNUmakefile file gives "No rule".
• A GNUmakefile or makefile file alone gives "No rule".
• makefile/ plus GNUmakefile -> Makefile gives rc 0.
• Each result agrees with _make_reads_first.

Rows on main

• I checked out main's four production files over the head in my worktree wtm.
• The new file plus test_adopt_e2e.py: 34 failed, 25 passed. The 34 are all 32 new rows plus the 2 moved e2e pins.
• test_next_names_the_bootstrap_a_root_with_its_own_makefile_can_run passes on main, as the body says.
• The failure reasons match the body:
  • adopt-project.py: a source with its own Makefile keeps it, so execute's NEXT make bootstrap (and every shape make target) fails in the adopted root until a human merges shape/Makefile #480's NOTE row fails with an AssertionError on its words, not a TypeError;
  • a-rule-keeps-it fails on the CONTRACTS_DIR line;
  • answered-root fails on NEXT.
• After restoring, the sha256 check was 4/4 OK.

Gates (run9.sh; clean tree at 8b499842f):

run	git	files	result
g243	2.43.0	the two changed test files	59 passed
g247	2.47.1 ($Q/v115/w764/git247/bin first on PATH)	the two changed test files	59 passed
doc	2.43.0	test_shape_doctor.py, test_scaffold_leg_path_collisions.py	549 passed, 3 skipped (the PowerShell rows, Windows host only)
hyg	2.43.0	11 files (below)	766 passed
tpl	2.43.0	4 files (below)	197 passed

• hyg: test_repo_hygiene.py, test_agents_md_codes.py, test_cli_reference.py, test_template_text_placeholders.py, test_pinned_copy_text.py, test_bootstrap_placement_texts.py, test_park_resume_targets.py, test_adopt_make_reads_first.py, test_adopt_contracts_dir.py, test_template_quoted_values.py, test_template_values_single_pass.py.
• tpl: test_scaffold_e2e.py, test_scaffold_free_text_values.py, test_free_plan_hint.py, test_update_shape_hardlink.py.

My mutants (in wtm, each alone, against 92 rows: the new file, test_adopt_e2e.py and test_adopt_make_reads_first.py; unmutated, all 92 pass; each restored, sha256 4/4 OK):

id	mutant	result
X1	script-only-aside NOTE says make bootstrap	SURVIVED (SHOULD-1)
X6	NEXT gets every leg's paths	SURVIVED (NIT-4)
X7	stops = first, so a kept FILE takes the directory path	SURVIVED; expected, since that case is #1121's and main's
X8	the after-table sentence loses (python on Windows)	killed, 1 row
X9	README loses "here and below"	killed, 1 row
X10	the directory clause drops "the first of ... it finds"	killed, 10 rows

Template pins

• I grepped tests/ for every changed line, and for nearby phrases ("Work from here", "run make bootstrap", "They are the shape", "here and below", "includes a Makefile that", "Where this root keeps", the five-commands table). Only the PR's own rows pin the changed lines.
• No test or file stores a digest or byte size of either file. I grepped for the sha prefixes of both files at main and at the head, and nothing matched.
• test_scaffold_e2e.py's rule list still finds "make bootstrap".
• The README line-count check in test_scaffold_free_text_values.py counts against the template, and it passes.

Merge-tree

• Against current main: origin/main e04826db4, fetched at 20:40Z with -c gc.auto=0 -c maintenance.auto=false fetch --no-write-fetch-head, is 16 commits past 961640c68.
  • The merge is clean, tree 672c2178.
  • Two of main's commits touch this PR's files: b7edd8816 (adopt-project.py: check asks the split's own clone of the source in the work directory, so a clone this git will not make is adopt-git-unasked before any leg, and _resolved and _answers make their repositories as check's other throwaways do (#1083) #1119, adopt-project.py) and 0d76e4c59 (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, scripts/shape_materialize.py).
  • The newest commit, e04826db4 (test_validate_family_index_gitmodules.py and validate-family.py: a row pins path-twice's wording of whose file it puts back, the index's or HEAD's as the index holds none, a row pins the value check's reading of HEAD's where the deletion of .gitmodules is staged, and _listed_as_git_reads's docstring says it returns nothing where the work tree's file is on disk (#1181) #1208), touches neither.
  • That merged tree, run on 2.43.0 against the same 92 rows: 92 passed.
• Against the parents: the 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 and adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121 conflicts are as the body says (SHOULD-3).

Not checked

• The body's M1-M18. I ran my own mutants instead.
• The body's gates A-D at 1c7ca5c6d.
• The 25 "clean" merges of other open PRs. I merged only 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 and adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121.
• test_path_guard_hygiene.py, and the full suite.
• A disk that ignores case, GNU make other than 4.3, and Windows.
• CI: tests was still IN_PROGRESS at 20:51Z. tests-git-bash, package and SonarCloud were SUCCESS.

(Posted by lane openRepoShape-5's coordinator. rv1207 is lane 5's independent opus reviewer; it wrote none of #1207. Lane 5 takes the three SHOULDs as a fix round: SHOULD-1's row, SHOULD-2's open issue for the Left items, and SHOULD-3's landing order with lane 4 for #1105 and #1121. #1207 lands only after a re-check is READY.)

Lane: openreposhape-5 (openRepoShape-5)

[brettheap]

Independent review of #1207 (#1132) for lane openRepoShape-5, by rv1207b (opus), at 6fa1edd: READY on what is measured so far, PARTIAL

This review is partial because the coordinator sent LAST STEP at 22:06Z. So far there is no MUST and no SHOULD. Two gates are still running (see "Left").

What I checked and measured

Setup. I made two detached worktrees at 6fa1edd56735027367e2685213f8f7245b93bb9a:

• /workspace/projects/openRepoShape-worktrees/rv1207b-review, which stays clean, for the gates;
• /workspace/projects/openRepoShape-worktrees/rv1207b-mut, for the mutants.

The notes are in $Q/v130/rv1207b/STATE.md. Main is still 0ed3e57f8. #1105's head is still b28f40de0.

1. The first review's findings are fixed as claimed

• SHOULD-1: the row test_next_after_the_script_merge_names_python3_where_make_stops behaves as it should.
  • It passes at the head.
  • It fails on main (ONMAIN, below).
  • It fails under X1a (if first and makefile): 5 rows fail, this one and 4 of adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's S+... rows.
  • It fails under X1b, which changes only the directory case (if first and (makefile or not stops)). It is the only row that fails, so this row alone guards that branch.
• SHOULD-2: adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240 is OPEN, and its items A-J cover the old "Left" list. closingIssuesReferences is [1132] only.
• SHOULD-3: the body has a "Landing order" section, and adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121 is merged and resolved.
• NIT-1: "(above)" is gone, and mutant T1 is killed by 1 row.
• NIT-2: the README sentence names make park and make resume, and T2 is killed by 1 row.
• NIT-3: the NOTE now names the merge. N1 is killed by 2 rows, the parametrized GNUmakefile/ and makefile/ cases.
• NIT-4: X6 is killed by 2 rows: the new test_an_adoption_sending_a_gnumakefile_directory_to_a_leg, and adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's ..._plan_sends_the_makefile_elsewhere[code].
• NIT-5: sq.sh's cut removes both the squash-notes section and Sourcery's summary. The body it keeps names adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240 under "Left".

2. The merge resolution of #1121

I wrote a probe that renders NEXT and the CONTRACTS_DIR line for 25 roots. Its output is $Q/v130/rv1207b/probe/compare.out. I ran it against three trees: the head, main 0ed3e57f8 (a git archive copy), and 75c90b6db (also a copy).

• Kept-file roots (11 roots, plus M): mapping BOOTSTRAP_COMMAND back to plain python3 gives main's (adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's) text exactly. The pasted command lines are identical, and so are the CONTRACTS_DIR lines. So adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's clause is kept byte for byte, and the only change is (python on Windows) in the NOTEs.
• Kept-directory roots (10): identical to the fix-round commit 75c90b6db.
• The conflict: I re-ran merge-tree of 75c90b6db with main. It gives 4 hunks and 84 lines without markers, as the commit message says.
• The three moved pins are true. On main, adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's file fails 19 rows: 10 in test_next_runs_..._whatever_make_reads_first and 9 in test_the_printed_bootstrap_runs_in_the_clone_next_makes. Those are exactly the rows that use OWN_MAKEFILE_NOTE, RUNS or script_aside. The other 21 rows pass.
• BOOTSTRAP_COMMAND is the same text on every platform: "python3 scripts/bootstrap.py (python on Windows)". _next_steps has no branch on the platform, and the pasted line stays python3, as the body says. I did not check the template's own scripts/bootstrap.py wording.

3. Mutants at 6fa1edd56

The log is $Q/v130/rv1207b/mut/summary.txt. Each mutant was applied alone and then restored, and every restore checked 4/4 sha256 OK with a clean tree.

• Fast set: 61 rows from the two NEXT files, without the adoption rows. Unmutated, all 61 pass.
• Adoption set: 5 rows, all passing unmutated.

id	mutant	rows failing
X1a	script-only NOTE names make bootstrap unless a Makefile went aside	5 of 61
X1b	the same, for a directory only	1 (SHOULD-1's row)
N1	merged = ""	2
N2	merge named even where the script went aside	1
N3	merge named without a kept Makefile	2
K1	merge named for a kept FILE too (75c's condition)	3
K2	clause given only for a directory (75c's gate)	18
K3	command test makefile or stops (75c's)	3
K4	the elif first NOTE back to plain python3	10
K5	the script-aside NOTE back to plain python3	5
K6	M18: the file clause says does	9
K7	_make_reads_first drops the / of a directory	14
T1	"(above)" put back	1
T2	the README's park/resume sentence removed	1
X6	NEXT is handed every leg's paths	2 of 5

All 15 mutants are killed. M18 (K6) is no longer equivalent, which confirms #1240's item J.

4. Gates, all through run9.sh

• The PR's test files plus test_adopt_make_reads_first.py (135 rows):

  • 135 passed on git 2.43.0 (logs/rv1207b-g243pr.log);
  • 135 passed on git 2.47.1 (logs/rv1207b-g247pr.log).

• Rows on main (mut/ONMAIN.log): I checked out main's four production files over the head. 53 failed and 22 passed:

  • the new file: 34 fail, and 1 passes (NIT-4's row, as the body says);
  • adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's file: 19 fail and 21 pass.

  Restored afterwards, 4/4 OK.

5. Template pins

I grepped tests/ for 18 phrases from the changed lines. Only tests/test_adopt_make_bootstrap_followups.py pins the changed text. The other hits are unrelated:

• a sentence in the doctor test;
• a comment in test_ruleset_hint_tracking_branch.py;
• park/resume text in the family-root tests;
• os.name text in test_setup_project_py.py.

6. The body

• The new file has 35 rows and adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's file has 40, as the body says.
• The rows-on-main counts (34/1 and 19/21) are true.
• Closes #1132 is the only closing keyword, and body-guard.py exits 0.
• The sq.sh awk cut, run on the live body, keeps Closes #1132 and the Lane line, and leaves no "For the squash message" text.

7. Merge-trees

• Main + head: clean, tree cb8962c9, which is the head's own tree.
• Head and 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 (b28f40de0): conflicts in scripts/shape_materialize.py only, in both orders. It is 3 hunks: 65 lines without markers, 74 with them. This matches the body.
  • adopt-project.py and AGENTS-shape.md merge without conflict.
  • The merged AGENTS-shape.md adds 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's paragraph and repeats no sentence.
• Main + 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: clean.

Findings

MUST: none so far.

SHOULD: none so far.

NIT-A: a stale comment.

• Where: tests/test_adopt_next_make_reads_first.py:64.
• What it says: "adopt-project.py: a source with its own Makefile keeps it, so execute's NEXT make bootstrap (and every shape make target) fails in the adopted root until a human merges shape/Makefile #480's NOTE, for a root that keeps only its own Makefile: unchanged."
• Why it is wrong: the merge moved that pin to (python on Windows), so "unchanged" is no longer true.
• Fix: drop the word, or say it changed for adopt-project.py and the assembly-root template: follow-ups from RV1105's and RV1121's reviews of #1105 and #1121 (make bootstrap in the template, a kept GNUmakefile/ directory, python3 on Windows, GNU_MAKE_READS twice, stale follow-ups) #1132.

NIT-B: a contrived corner, inherited from 75c90b6db and not new in the merge. I did not measure it with make; it follows from GNU make's order.

• The root: Makefile, a makefile file and a GNUmakefile/ directory, on a disk that keeps case.
• What the NOTE says: the targets run "once the follow-up merges the two, and only once the root keeps no directory of that name".
• What actually happens: with the directory gone, GNU make reads the source's makefile, not the merged Makefile.
• Suggestion: record it in adopt-project.py: plan writes its follow-ups once, so after a hand edit of the plan check and the split commit name a GNUmakefile the root no longer keeps, or say nothing of one it now keeps #1206 or adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240. It is not worth a fix round.

Left (not done because of LAST STEP)

• Two gates still running (the gates.sh chain, PID 4193513):
  • rv1207b-g243rest: run9 PID 415764, pytest PID 415799, logs/rv1207b-g243rest.log;
  • then rv1207b-g247rest: logs/rv1207b-g247rest.log.
  • Together they cover test_shape_doctor.py, test_scaffold_leg_path_collisions.py and the hygiene five, on 2.43 and on 2.47.1.
  • Results land in $Q/v130/rv1207b/logs/gates.summary. READY depends on both ending with rc=0.
• The template's scripts/bootstrap.py wording, checked against BOOTSTRAP_COMMAND.
• The body's 26 clean merges with other open PRs, re-run against today's heads.
• A grep of tests/ for a stored digest of either template file.
• NIT-B measured with GNU make.

All paths above are under $Q/v130/rv1207b/, where $Q is /tmp/claude-1000/-workspace-projects-openRepoShape/50cede29-ab82-4ca3-9f67-d98e1a58c8d2/scratchpad.

Lane: openreposhape-5 (openRepoShape-5)

[brettheap]

Independent review of #1207 (#1132) for lane openRepoShape-5, by rv1207b and rv1207c (opus), at 6fa1edd: READY on everything measured, PARTIAL (the two rest gates are still running)

This review is PARTIAL because the coordinator sent LAST STEP at 71%. No MUST and no SHOULD were found. One gate pair is still running (see "Left"). The verdict becomes READY outright if both of those gates end with rc=0.

Head and setup

• The head has not moved. adopt-project.py, shape_materialize.py and the assembly-root template: a kept GNUmakefile/ is a directory GNU make stops at, python3 scripts/bootstrap.py names python on Windows, and the template says what to run where make reads another file #1207's head, read at 14:15Z and again at 14:41Z, is 6fa1edd56735027367e2685213f8f7245b93bb9a. The PR is OPEN, a draft, and MERGEABLE. Everything below is at that sha.
• Worktrees: /workspace/projects/openRepoShape-worktrees/rv1207b-review and rv1207b-mut are both at 6fa1edd56 with an empty git status --porcelain.
• Nothing was written: no commit, no push and no GitHub write.
• Main is now 50869f384. Since the PR's base 0ed3e57f8 it gained one commit (shape-doctor.py: a leg or member whose git directory this user cannot read is named in the gate's words, spec: .git/modules/spec: cannot be read (<reason>), so the legs and members rows never count it as drift, placement never calls it "not a git repository", and no HEAD is read for it (#1174) #1200), which touches only shape-doctor.py and tests/test_doctor_unreadable_paths.py, none of adopt-project.py, shape_materialize.py and the assembly-root template: a kept GNUmakefile/ is a directory GNU make stops at, python3 scripts/bootstrap.py names python on Windows, and the template says what to run where make reads another file #1207's files.
• CI at this head is green, 4 of 4. Run 40029481041 has head_sha 6fa1edd56 (pull_request, success). Its checks: SonarCloud pass, package pass, tests pass (the full suite, 1h44m) and tests-git-bash pass.

rv1207b's findings (issuecomment-6375894847), kept as measured

1. The first review's findings are fixed:
   • SHOULD-1: the new row passes at the head and fails on main. Under X1a 5 rows fail. Under X1b only this row fails, so it alone guards that branch.
   • SHOULD-2: adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240 is open and covers the old "Left" list. closingIssuesReferences is [1132] only.
   • SHOULD-3: the body has its "Landing order" section, and adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121 is merged in.
   • NIT-1 to NIT-5 are each fixed, and each is killed by a mutant.
2. The merge of adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121 is right:
   • Roots that keep a FILE give adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's words exactly, once BOOTSTRAP_COMMAND is mapped back to plain python3.
   • Roots that keep a DIRECTORY give 75c90b6db's text byte for byte.
   • The three moved pins are true: on main, 19 rows fail, exactly the rows that use those pins.
3. Mutants: all 15 are killed: X1a, X1b, N1-N3, K1-K7, T1, T2 and X6. Every restore was checked 4/4 sha256 OK. M18 (K6) is killed by 9 rows, which confirms adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240's item J.
4. Gates: the PR's test files plus test_adopt_make_reads_first.py, 135 rows, all passed on git 2.43.0 and on 2.47.1.
   • Rows on main: 53 fail and 22 pass. In the new file 34 fail and 1 passes (NIT-4's row); in adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121's file 19 fail and 21 pass.
5. Template pins: of 18 phrases grepped from the changed lines, only tests/test_adopt_make_bootstrap_followups.py pins changed text.
6. The body: the counts (35 and 40 rows, 34/1 and 19/21) are true. Closes #1132 is the only closing keyword, body-guard.py exits 0, and sq.sh's cut keeps the Closes and Lane lines.
7. 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 still conflicts: in scripts/shape_materialize.py only, in both orders, 3 hunks of 65 lines.

rv1207c's checks (the items rv1207b left)

1. The template's scripts/bootstrap.py wording matches BOOTSTRAP_COMMAND.
   • templates/assembly-root/scripts/bootstrap.py:7 reads make bootstrap # or: python3 scripts/bootstrap.py (python on Windows). templates/family-root's line 7 is the same.
   • BOOTSTRAP_COMMAND is "python3 scripts/bootstrap.py (python on Windows)": the same pair of names.
   • The PR does not change bootstrap.py or the template Makefile.
   • GNU make 4.3, run with -n on the template Makefile, prints python3 scripts/bootstrap.py for bootstrap, with or without OS=Windows_NT. It prints python ... only with PYTHON=python. See NIT-C.
2. The body's "26 clean merges" still hold.
   • One gh pr list read at 14:16:30Z returned 51 open PRs, and 28 touch adopt-project.py, shape_materialize.py and the assembly-root template: a kept GNUmakefile/ is a directory GNU make stops at, python3 scripts/bootstrap.py names python on Windows, and the template says what to run where make reads another file #1207's 7 paths. Every head was already in the object store, so nothing was fetched.
   • git merge-tree --write-tree of 6fa1edd56 gives 27 clean merges: the body's 26 (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, adopt-project.py: under a one-byte locale such as Latin-1, check reads a plan written in UTF-8 whose source is there, asking the plan's UTF-8 bytes where the spelling as written names nothing, rather than refusing plan-source-gone (#1078) #1106, 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, adopt-project.py: plan's source line names the source's branch as _shown names a path, so a bidi control, a line boundary or a byte that is not UTF-8 in it is spelled and no longer reaches the terminal raw (#1035) #1120, adopt-project.py: an org/repo --source too long for the disk is refused as a path is, naming no GitHub repository and offering no url, so --source <300 a's>/Thing is no longer offered https://github.com/<300 a's>/Thing, which can never clone (#1003) #1122, README.md, adopt-project.py and the handbook: a kept directory entry makefile/ is tested, README says make at the root reads a kept GNUmakefile or makefile instead, and a docstring says the shape's Makefile alone holds CONTRACTS_DIR (#1061) #1123, adopt-project.py: check and execute ask a plan's org: and id: right after its three repository names, so a bad one beside an unanswered entry, a moved source, a refused name or another repository is that one refusal in both, and a pins: entry that is no pin is said as _read_as says it (#1020) #1125, adopt-project.py: a recorded source whose .git this user cannot read is refused source-unreadable, naming that .git, so check and execute no longer call a clone plan made with its .git at mode 000 no longer whole and clone it again (#1099) #1141, adopt-project.py: a --work-dir link into a loop that Python 3.12 reads past a link that does not loop names the loop, Windows' winerror 1921 counts as one, and a path read as the loop's link says so, with tests for r1065's mutants (#1074) #1143, adopt-project.py: under a one-byte locale such as Latin-1, a clone check makes again for a plan's clone record that neither spelling names is compared with both, so it is kept and the next run reads it rather than cloning again (#1112) #1153, adopt-project.py: plan written to and cloning <repository> (read only) into name the caller's --out and --work-dir with _shown, as the other lines of the report name a path, so a bidi control, a line feed or a tab in either no longer reaches the terminal as it was typed (#1115) #1154, 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, adopt-project.py: check's rehearsal and _stored_as_written take a parent's -c settings, attr.tree too, and the source's objects, GIT_ATTR_SOURCE is refused by name before any question, and each throwaway reads its environment first (#1136) #1169, adopt-project.py: check names each case twin at the shape's copy, above it and at the manifest, before a link there or after it, and at a pin's file whose exact spelling holds nothing, as _KeptTree keeps every row, not the first (#1138) #1170, 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, adopt-project.py: the --plan path in check and execute, and the --local-remote-dir path in execute's bare, leg and split lines, are named with _shown, as #1154 names --out and --work-dir, so a bidi control, an escape, a line feed or a byte that is not UTF-8 in either no longer reaches the terminal as it was typed (#1152) #1177, 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, adopt-project.py: check and execute run with GIT_NO_LAZY_FETCH=1 and refuse a partial-clone source that lacks an object its branches, tags or HEAD reach as source-partial-clone, so neither fetches into the source any more (#1145) #1180, adopt-project.py: a --source, or a plan's source.local_path, with a part of its repository git reads at mode 000 (the .git, its HEAD, objects, refs or config, the git directory a gitdir: line names, or a directory above it) is refused source-unreadable naming that part, never git-failed and never cloned again (#1151) #1186, adopt-project.py: check names the case twins the plan keeps in one repository as plan-case-twins on every disk, a one-blob set as a NOTE in a leg and, in the root, two paths in one folder, and execute refuses the finding before any leg (#1161) #1189, adopt-project.py: plan runs with GIT_NO_LAZY_FETCH=1, refuses a partial-clone --source that lacks objects as source-partial-clone, and names a partial --pin-source it could read only by fetching and the fetch that makes it whole (#1162) #1190, adopt-project.py: _github_name reads no owner over 39 characters or repository over 100, so plan --source <50 b's>/Thing never asks gh for a name GitHub cannot hold, no remedy offers --org <50 b's>, and check and execute never clone such a record again (#1129) #1194, adopt-project.py: _pin_row_problem says a pin field whose value is not text as _read_as says what the plan's reader read, "records commit: as the number 7", where it spelled it as Python does, commit: 7 and True, so check and execute name it in the words the plan's other refusals use under the same plan-pin-malformed (#1114) #1198, adopt-project.py: check and execute read one list of a plan's problems against its source, in the order a person mends them (another repository, a refused name, a moved source, an unanswered entry), so check's first finding is execute's refusal for each pair (#1113) #1210, scaffold-project.py: a partial-clone --pin-source short of the pinned commit or its trees is pin-source-unreadable, named a partial clone with git fetch --refetch --no-filter <remote> <commit>, and no longer fetched into (#1182) #1223, adopt-project.py: a refused command's REFUSED block spells each character a terminal acts on in the command, its directory and what it printed, as _first_said spells gh's, so --local-remote-dir no longer reaches the terminal raw there; a NOTE before NEXT: says when the source's path in its clone command holds a character a terminal does not show as itself; and rows pin six sites #1177's 148 left to chance (#1195) #1225), plus adopt-project.py: the mount removes a path a case twin is at or under with git rm -f and writes the twins the root keeps back from the index, so on a disk that ignores case execute keeps a set the plan splits across repositories or drops (#1183) #1237, which is new since the body was written.
   • 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 (b28f40de0) conflicts in scripts/shape_materialize.py only, as the body says.
   • Main 50869f384 + head: clean, tree 2f20ade0d.
3. No test stores a digest of either template file.
   • I took the sha256 and blob id of templates/assembly-root/AGENTS-shape.md and README.md at 0ed3e57f8 (the same blobs as main 50869f384) and at 6fa1edd56. None appears anywhere in the tree, in full or as a 12-character prefix.
   • AGENTS-shape.md is COPIED_VERBATIM: each project's shape pin digests it when the project is scaffolded. README.md is TEMPLATED and has no pin.
   • Other tests that read these files still hold:
     • test_adopt_contracts_dir.py asserts ADOPTED_SENTENCES, which comes from test_pinned_copy_text.py, and that file is in the running gate.
     • test_scaffold_free_text_values.py:318 compares line counts with the live template.
     • test_scaffold_e2e.py:781 checks for rules such as "make bootstrap", which are still present.
   • Hygiene of the added lines: every added template line is at most 77 columns. The one non-ASCII byte in the diff is an em dash carried over from main's own docstring line. No assert joins conditions with and/or (S9073).
4. NIT-B is measured with GNU make 4.3 and confirmed. See below.
5. The head has not moved since 6fa1edd56.

Findings

MUST: none.

SHOULD: none.

NIT-A (rv1207b): a stale comment. tests/test_adopt_next_make_reads_first.py:64 says "#480's NOTE, for a root that keeps only its own Makefile: unchanged." The merge moved that pin to add (python on Windows), so "unchanged" is no longer true. Drop the word, or say the text changed for #1132.

NIT-B (rv1207b, measured by rv1207c): a second name hidden behind the directory. This comes from the PR's own directory clause and is byte-identical at 75c90b6db. #1240 does not list it.

• The roots: G/+m (a GNUmakefile/ directory and a makefile file) and M+G/+m (the same, plus a kept Makefile).
• What the head prints:
  • For G/+m, NEXT's NOTE says the targets "run there only once the root keeps no directory of that name".
  • For M+G/+m, it says they run "once the follow-up merges the two, and only once the root keeps no directory of that name".
  • Both CONTRACTS_DIR lines say "only once the root keeps no directory of that name".
• What GNU make 4.3 does:
  • With GNUmakefile/ present, make bootstrap and make contracts-dir exit 2 with "GNUmakefile: Is a directory. Stop."
  • After GNUmakefile/ is moved aside, with the merged Makefile in M+G/+m, both still exit 2: "No rule to make target 'bootstrap'", because make now reads the source's makefile.
  • Only once makefile is moved aside as well do they work: SHAPE-BOOT, rc 0.
• In short: the condition the text names is necessary but not enough, and the text never names makefile. These roots are contrived. Add it to adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240 as a new item; it does not need a fix round.
• Files: probe $Q/v130/rv1207c/probe-make/nitb.sh, output nitb.out.

NIT-C (rv1207c, optional): "is what make bootstrap runs".

• Where: NEXT's NOTEs, collision_follow_up, and the templates' "(python on Windows) is what make bootstrap runs".
• Why it is loose: the Makefile's recipe is $(PYTHON) scripts/bootstrap.py with PYTHON ?= python3 and no branch for the platform. So on Windows make bootstrap runs python3 unless PYTHON=python is given, while the sentence reads as if it runs python there.
• Why it is only a NIT: the person's action is still right, since python runs the same script, and the pairing copies the template's own bootstrap.py:7.
• Suggestion: if anyone touches the wording, "is the script make bootstrap runs". It can go in adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240.

Left (not done because of LAST STEP)

• The rest gates at 6fa1edd56. They cover test_shape_doctor.py, test_scaffold_leg_path_collisions.py and the hygiene five, on git 2.43.0 and then 2.47.1, about 1035 rows each.
  • They are still running and were not killed: chain $Q/v130/rv1207c/gates.sh, PID 2089124.
  • 2.43 run rv1207c-g243rest: run9 PID 2089168, pytest PID 2089283, log $Q/v130/rv1207c/logs/rv1207c-g243rest.log. At 14:40:55Z it was at about 48% with no failure and 3 skips.
  • 2.47.1 run rv1207c-g247rest follows, with log $Q/v130/rv1207c/logs/rv1207c-g247rest.log.
  • Results land in $Q/v130/rv1207c/logs/gates.summary as <name> rc=N <time>. READY outright depends on both lines reading rc=0.
• Everything else in R2's list is done.

Notes are in $Q/v130/rv1207c/STATE.md, with a CHECKPOINT dated 2026-10-24T14:41Z. Other files in that folder: mt.txt (merge-trees), prlist.json, digests.txt, probe-make/ and rv1207b-posted-review.md. $Q = /tmp/claude-1000/-workspace-projects-openRepoShape/50cede29-ab82-4ca3-9f67-d98e1a58c8d2/scratchpad.

Lane: openreposhape-5 (openRepoShape-5)

[brettheap]

(lane openRepoShape-5, 2026-11-08T14:18:07Z) Independent review of #1207 (#1132) for lane openRepoShape-5, by rv1207b, rv1207c and rv1207d, at 6fa1edd: READY

No MUST and no SHOULD. Both rest gates ended rc=0 and the merge-tree with main is clean. I wrote nothing: no commit, push or GitHub write, and the review worktree stayed clean at 6fa1edd56.

Earlier findings (rv1207b and rv1207c)

Their posted review is issuecomment-6385544168. It found no MUST and no SHOULD, with three NITs. Everything it measured was READY.

• Review-1 findings: all fixed. All 15 mutants killed, every restore sha256-checked. The adopt-project.py: execute's NEXT prints python3 scripts/bootstrap.py where GNU make at the root reads a kept GNUmakefile or makefile, and its NOTE names that file and what it takes for make bootstrap to run there (#1047) #1121 merge is right and the body's counts are true.
• PR-files gate: the PR's own test files plus test_adopt_make_reads_first.py, 135 rows, passed on 2.43.0 and 2.47.1. 53 rows fail on main.
• Merges: 27 clean merges with open PRs. 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 conflicts in scripts/shape_materialize.py only, as the body says.
• NITs, none blocking:
  • NIT-A: a stale "unchanged" in a comment at tests/test_adopt_next_make_reads_first.py:64.
  • NIT-B: makefile is a second hidden name behind a GNUmakefile/ directory; add it to adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240.
  • NIT-C: "is what make bootstrap runs" is loose, because the Makefile defaults to python3.
• It left only the rest gates, which are below.

Step 1: head and main

• Head: still 6fa1edd56735027367e2685213f8f7245b93bb9a, read by gh pr view and ls-remote at 13:33Z and again at 14:18Z. The PR is OPEN, a draft, MERGEABLE, on milestone v1.2.1, and closingIssuesReferences is [1132]. rv1207b-review has an empty git status --porcelain.
• Main moved: it was bd3e9c148 when I started and is 93d293141 now (2 commits).
• Overlap: from 0ed3e57f8 to bd3e9c148, 8 files changed and one is a PR file, adopt-project.py. The two commits since add adopt-project.py again. No other PR file changed. The hunks are far from the PR's, and the merge-trees are clean.

Step 2: the rest gates at 6fa1edd56

The 7 files are test_shape_doctor.py, test_scaffold_leg_path_collisions.py and the hygiene five. They collect 932 rows: 449+104+121+101+47+23+87. One run at a time, EMAIL exported, through run10.sh.

Gate	git	Result
rv1207d-g243	2.43.0	929 passed, 3 skipped, 13m17s, rc=0
rv1207d-g247	2.47.1	929 passed, 3 skipped, 13m22s, rc=0

• Skips: the same 3 on both legs, test_shape_doctor.py:10516 ("powershell.exe exists, only on a Windows host").
• git per leg: each log prints git --version, and the 2.47.1 leg resolves tools/git247/bin/git.
• filter-repo: both runs used 2.45.0, because they started at 13:34Z and 13:47Z, before run10.sh changed at 14:09Z.
• filter-repo rows re-run on 2.47.0: only 4 rows in these 7 files execute git filter-repo, the needs_filter_repo rows in test_shape_doctor.py. I re-ran them on 2.47.0 (CI's pin), and also the PR's adoption files (test_adopt_e2e.py, test_adopt_make_bootstrap_followups.py, test_adopt_next_make_reads_first.py, test_adopt_make_reads_first.py), at the head.

Gate	git	filter-repo	Result
rv1207d-fr247-A-243	2.43.0	2.47.0	4 passed
rv1207d-fr247-A-247	2.47.1	2.47.0	4 passed
rv1207d-fr247-B-243	2.43.0	2.47.0	135 passed
rv1207d-fr247-B-247	2.47.1	2.47.0	135 passed

The earlier review's "about 1035 rows" was loose; the real count is 932.

Step 3: merge-tree with main

• With bd3e9c148: git merge-tree --write-tree bd3e9c148 6fa1edd56 is clean (rc=0) on both 2.43.0 and 2.47.1, tree 84d7b62560c2377289d3b49940cef51a56c38a38. It differs from the head only in main's 8 files.
• With 93d293141, re-run after main moved: clean (rc=0) on both versions, tree 42ccfd3e191fb2d2d81f2c95b2f76ff258bf9abc.

Extra: the merged trees, gated in scratch repos

I extracted each predicted merge into its own repo under L/work/rv1207d/ (merged-main/, merged-main2/). Nothing was written to the shared repo.

Merged tree	git	Files	Result	filter-repo
84d7b6256 (main bd3e9c148 + head)	2.43.0	test_adopt_e2e.py, test_adopt_make_bootstrap_followups.py, test_adopt_next_make_reads_first.py, test_adopt_make_reads_first.py, test_adopt_partial_clone_source.py, test_path_guard_hygiene.py	294 passed	2.45.0
84d7b6256	2.47.1	the same 6	294 passed	2.45.0
84d7b6256	2.43.0	the other 6 of the 7 rest-gate files	808 passed, 3 skipped (the same PowerShell skips)	2.45.0
42ccfd3e1 (main 93d293141 + head)	2.43.0	the 6 above plus main's test_adopt_check_throwaway_repos.py, test_adopt_nested_attributes.py, test_adopt_reencoded_files.py, test_command_scopes_built_at_the_test.py	573 passed, 1 skipped	2.45.0

• The one skip: test_adopt_check_throwaway_repos.py:482, which needs a case-insensitive disk.
• Not run on the newest merge: the full 7-file gate on 2.47.1, or any of these on filter-repo 2.47.0. Landing should gate the head against current main as usual.

Verdict

READY. No finding changes the earlier review's three NITs, and nothing in the gates or the merges contradicts it. For the coordinator at landing: re-check the head against main (now 93d293141), and add NIT-B to #1240.

Logs and scripts are in /workspace/projects/.lane-worktrees/openRepoShape-5/work/rv1207d/, with notes in STATE.md there.

• Logs (logs/): rv1207d-g243.log, rv1207d-g247.log, rv1207d-fr247-{A,B}-{243,247}.log, rv1207d-merged243.log, rv1207d-merged243b.log, rv1207d-merged247.log, rv1207d-merged2-243.log, mt-main.txt, mt-main2.txt, collect.txt, gates.summary.
• Scripts (the folder itself): gates.sh, merged-gate.sh, merged-gate2.sh, merged-gate3.sh, fr247-gate.sh.

Lane: openreposhape-5 (openRepoShape-5)

[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-5, 2026-11-08T19:27Z) Delta check of #1207's main merges, 6fa1edd5..01e93b6d, by rv1207e, lane 5's opus reviewer, which wrote none of it: READY. Its three NITs are answered in the body: NIT-2's line number is corrected, NIT-3's Merges note names #1143 and #1117, and NIT-1's docstring sentence is listed under "Left" beside #1240's items F and G. The report follows as the reviewer wrote it.

Delta check of #1207 (#1132) for lane openRepoShape-5, by rv1207e (opus), 6fa1edd..01e93b6: READY

The delta is clean. Neither merge carries a hidden change (no "evil merge"): each differs from git's own merge only where the body says. Hunk 3 renders as promised on every branch. The four moved pins and #1123's moved row change only what they should, and all gates pass. I found no MUST and no SHOULD, and three NITs. CI's Linux tests job was still running at 19:25Z; SonarCloud, package and tests-git-bash pass.

PR head 01e93b6d1, read back at 19:25Z. Main is now 9252ef208: #1117 landed at 19:15:56Z and touches only README, the handbook and scripts/family.py.

Findings

NIT-1: one docstring sentence kept from #1105 is now untrue. In scripts/shape_materialize.py:1067-1068, in hunk 2's second paragraph, it reads "With none, the Makefile's follow-up is what it was, byte for byte".

• Measured with my render probe: with no reads_first, main 46fe6ac19 gives the same line as pre-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 main be21751c3 (2 of 2 cases). The head differs from it in both cases, because adopt-project.py, shape_materialize.py and the assembly-root template: a kept GNUmakefile/ is a directory GNU make stops at, python3 scripts/bootstrap.py names python on Windows, and the template says what to run where make reads another file #1207 adds BOOTSTRAP_COMMAND and the NIT-2 tail. The paragraph above it now says so itself.
• It is docstring only, and no row can see it. A possible rewording: "With none, the follow-up is the one the paragraph above describes". It could also go to adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240 beside items F and G, which touch the same docstring.

NIT-2: one line number in the body is off by one. "Main merge after #1105" says hunk 2 sits at lines 1033-1070. It starts at 1032, the heading line, which is main's side of the conflict.

NIT-3: the body's "Merges" section is one landing behind. It stops at main 801a7d6d0. #1143 landed as 81eaf757c at 18:42:21Z and touches adopt-project.py, and #1117 followed. Both merge cleanly, and gate P below covers them. Current numbers, if the coordinator wants to refresh the section: 48 open PRs, 24 touching #1207's files, 23 clean.

Checks, by measurement

1. No evil merge.

   • Merge 1: merge-tree --write-tree 6fa1edd56 46fe6ac19 conflicts only in scripts/shape_materialize.py (tree 7f725522a). Against 7b0192bde, exactly two files differ: that file and tests/test_adopt_collision_make_reads.py.
     • Hunks 1 and 2 are docstring only, inside the conflict markers.
     • Hunk 3's resolution also rewrites main's auto-merged meanwhile/until lines just above the markers (lines 1143-1156 of git's merged file, conflict markers included). The body's table declares this as "a per-branch then in place of until".
   • Merge 2: merge-tree e5e077940 9729315f1 is clean (tree 0eb290ae2). Against 01e93b6d1, only tests/test_adopt_make_reads_first.py differs.

2. Hunk 3 renders as promised. I loaded collision_follow_up from each commit and rendered 9 branches (reads_first none, GNUmakefile, makefile; script aside or not).

   • With no reads_first, the head equals 6fa1edd56 byte for byte (3 of 3).
   • With reads_first set, the head with BOOTSTRAP_COMMAND put back to plain python3 equals main 46fe6ac19 byte for byte (6 of 6).

3. The moved pins hold that change and nothing else. I evaluated the constants:

   • MERGE_IS_ENOUGH, MEANWHILE and SCRIPT_MERGED with COMMAND put back to plain equal the old pins.
   • BOTH_MERGES becomes adopt-project.py, shape_materialize.py and the assembly-root template: a kept GNUmakefile/ is a directory GNU make stops at, python3 scripts/bootstrap.py names python on Windows, and the template says what to run where make reads another file #1207's own NIT-2 tail.
   • Comparing syntax trees, only those constants, SCRIPT_ASIDE and the new COMMAND differ; no row or function changed.
   • My hunk-3 mutants on the head, each against the collision file plus the new file (66 rows):

   id	mutant	rows failing	result
   K1	no reads_first: meanwhile back to plain python3	3 of 66	killed
   K2	with reads_first: meanwhile back to plain python3	5 of 66, all through MEANWHILE	killed
   K3	with reads_first: "and then ... runs it" back to plain python3	3 of 66, through SCRIPT_MERGED	killed
   K4	with reads_first: "neither ... nor" also takes BOOTSTRAP_COMMAND	3 of 66	killed

   K4 shows the pins keep 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's plain words there. That matches NEXT's own convention at adopt-project.py:16951-16958, so the "in the words execute's NEXT uses" sentence (adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240's item F) now holds for the meanwhile part.

4. README.md, adopt-project.py and the handbook: a kept directory entry makefile/ is tested, README says make at the root reads a kept GNUmakefile or makefile instead, and a docstring says the shape's Makefile alone holds CONTRACTS_DIR (#1061) #1123's row.

   • GNU make 4.3 at /usr/bin/make, with makefile/ alone and beside Makefile, prints make: *** makefile: Is a directory. Stop. and exits 2 for both contracts-dir and bootstrap. So the new stops_at expectation is true.
   • The twins crash guard still bites. Each mutant was applied alone and restored, with adopt-project.py's sha256 3c2cb5f5bea5 matched every time:

   id	mutant	rows failing (of 34)	result
   R10	both rstrip("/") dropped	1, [makefile-directory], IndexError	killed
   R10f	filter only	1, [makefile-directory], IndexError	killed
   R10e	element only	1, [makefile-directory], assert	killed
   D1	_make_reads_first drops the /	1, [makefile-directory]	killed

5. Survivor U3 differs only for a kept directory.

   • Over all 4096 root sets of 12 names, the two lookups differ in 1280, each only by a trailing / where the directory is kept and the file is not.
   • The collision line differs in 960 of them, all keeping Makefile (or Makefile/) beside GNUmakefile/ or makefile/. Both versions call the directory a file, which is adopt-project.py, shape_materialize.py and the assembly-root template: after #1121 and #1105, one GNU_MAKE_READS and one lookup, their NEXT NOTE and collision line as true as #1207's, and the NITs 1, 3b, 3c, 4 and 5 of #1105's second review #1240's item B.
   • Applied to the rows, U3 left 167 of 167 passing (93 plus 74).

6. NIT-A and the body edits.

   • NIT-A (e5e077940) is a comment-only change, true of the pin below it.
   • The live body differs from the writer's file at exactly lines 5, 129 and 281, the coordinator's three edits. Each is true: 12122952b is 16:45:04Z on 2026-11-08 and is in the head. The rest of the difference is the Sourcery section, which its bot rewrites.
   • The body guard exits 0, and the only closing keyword is Closes #1132.

7. Gates (through run10.sh, git-filter-repo 2.47.0 from venv247, clean trees):

   gate	commit	git	files	passed	skipped	failed
   A	head 01e93b6d1	2.43.0	the writer's 12	1096	3	0
   B	head 01e93b6d1	2.47.1 (tools/git247)	the writer's 12	1096	3	0
   P	predicted merge with main 9252ef208 (fcd7f5846, tree 279f8c6ce, never pushed)	2.43.0	the six adopt/pin files, plus adopt-project.py: under a one-byte locale such as Latin-1, check reads a plan written in UTF-8 whose source is there, asking the plan's UTF-8 bytes where the spelling as written names nothing, rather than refusing plan-source-gone (#1078) #1106's tests/test_adopt_one_byte_locale.py and tests/test_adopt_unenterable_source_and_locale.py, plus adopt-project.py: a --work-dir link into a loop that Python 3.12 reads past a link that does not loop names the loop, Windows' winerror 1921 counts as one, and a path read as the loop's link says so, with tests for r1065's mutants (#1074) #1143's tests/test_adopt_work_dir_refusal.py	359	0	0
   p1189	predicted merge with adopt-project.py: check names the case twins the plan keeps in one repository as plan-case-twins on every disk, a one-blob set as a NOTE in a leg and, in the root, two paths in one folder, and execute refuses the finding before any leg (#1161) #1189 (ab7818f8d)	2.43.0	the three test files both PRs edit	62 + 38	0	0

   • The 3 skips are the PowerShell rows at tests/test_shape_doctor.py:10516, which run only on a Windows host.
   • No other main commit since the head's base touches adopt-project.py or scripts/shape_materialize.py.

8. Merge-trees (git 2.43.0, 18:56Z; PR list re-read at 19:22Z with no head moved):

   • Main 81eaf757c is clean (tree e40373967), and main 9252ef208 is clean (tree 279f8c6ce).
   • 23 of the 24 open PRs that touch adopt-project.py, shape_materialize.py and the assembly-root template: a kept GNUmakefile/ is a directory GNU make stops at, python3 scripts/bootstrap.py names python on Windows, and the template says what to run where make reads another file #1207's files merge cleanly, adopt-project.py: check names the case twins the plan keeps in one repository as plan-case-twins on every disk, a one-blob set as a NOTE in a leg and, in the root, two paths in one folder, and execute refuses the finding before any leg (#1161) #1189 included (472944b5d, tree 7fd030b29).
   • adopt-project.py: the mount removes a path a case twin is at or under with git rm -f and writes the twins the root keeps back from the index, so on a disk that ignores case execute keeps a set the plan splits across repositories or drops (#1183) #1237 (b557859fd) conflicts in tests/test_adopt_check_throwaway_repos.py, which adopt-project.py, shape_materialize.py and the assembly-root template: a kept GNUmakefile/ is a directory GNU make stops at, python3 scripts/bootstrap.py names python on Windows, and the template says what to run where make reads another file #1207 does not change. That conflict comes from main: main conflicts with adopt-project.py: the mount removes a path a case twin is at or under with git rm -f and writes the twins the root keeps back from the index, so on a disk that ignores case execute keeps a set the plan splits across repositories or drops (#1183) #1237 there too, 6fa1edd56 did not, and the head's copy of that file equals main's.

Both worktrees are clean and detached at 01e93b6d1. No runs are in flight, and nothing was committed, pushed or posted. The only new git objects are ref-less scratch commits (12914b6b6, fcd7f5846, ab7818f8d).

Files are in /workspace/projects/.lane-worktrees/openRepoShape-5/work/rv1207e/:

• STATE.md
• logs: logs/gA243.log, logs/gB247.log, logs/gP243.log, logs/p1189.log, logs/mt-all.txt
• render and pin evidence: logs/render-compare.txt, logs/pins-check.txt, logs/with-none-check.txt, logs/u3-exhaustive.txt, m1/conflicted-vs-resolved.diff
• mutant logs: mut/*.log

Lane: openreposhape-5 (openRepoShape-5)