adopt-project.py: `check`, given no `--local-remote-dir`, asks GitHub nothing about the two legs of a plan for another repository and prints no NOTE that `execute` asks again, as `execute` refuses `plan-not-the-source` first and asks nothing; under `--local-remote-dir` it still asks the disk (#1232) by brettheap · Pull Request #1313 · opensoft/openRepoShape

brettheap · 2026-11-10T13:51:24Z

Built on #1210's head 791106c2, then merged with #1210's final head 27c8f67b and its squash 106f92ff after #1210 landed.

What a person sees change

adopt-project.py check on a plan for another repository (plan-not-the-source), given no --local-remote-dir:

Before: it ran gh repo view <org>/<Project>-spec --json nameWithOwner and the same for -code, under the plan's own org:. Where gh could not answer, it printed two NOTEs saying "execute asks again before it makes either leg". execute refuses plan-not-the-source first and asks no remote anything (adopt-project.py plan never compares --project (or --org) with the repository it adopts in place, so a source named billing-service is refused with a remediation that steers to --project BillingService, which plans and checks plan ok against acme/billing-service, and execute then acts on acme/BillingService: a raw git ls-remote failure when no such repository exists, and, when a different repository of that name does, both legs, the split branch, the pull request and the topic on that repository #499), so it never asks.
Here: check runs no gh and prints neither NOTE. It still lists plan-not-the-source first, with the same remediation and exit 1, and execute still refuses it first, exit 2. So 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's rule holds.

Nothing else changes:

The source's own plan is still asked gh repo view of each leg, and still gets its NOTE where gh cannot answer.
Under --local-remote-dir every output is the same as before.

The choice under `--local-remote-dir`, and why

#1232 left this open. Measured first (the table below):

What the disk is asked. Under --local-remote-dir, check asks the disk, not GitHub: is each leg's bare repository, <dir>/<name>.git, already there, and can the directory hold them?
The same path before and after the mend. Beside plan-not-the-source with the spec leg's bare repository present, check lists leg-remote-exists for <dir>/Northwind-spec.git. That is the very finding, word for word, it lists for the source's own plan under the same directory. The source's own plan is the one the remedy (plan --project Northwind --org testorg) writes.
GitHub is not like that. Without --local-remote-dir, the plan edited to org: acme is asked about acme/Northwind-spec. The plan the remedy writes is asked about testorg/Northwind-spec.

So the skip applies only without --local-remote-dir. Under it, the disk is still asked beside plan-not-the-source, for three reasons:

It asks no remote. Nothing leaves the machine, so adopt-project.py plan never compares --project (or --org) with the repository it adopts in place, so a source named billing-service is refused with a remediation that steers to --project BillingService, which plans and checks plan ok against acme/billing-service, and execute then acts on acme/BillingService: a raw git ls-remote failure when no such repository exists, and, when a different repository of that name does, both legs, the split branch, the pull request and the topic on that repository #499's reason for execute does not apply. And the disk always answers, so there is no NOTE to be wrong.
No org: is part of a leg's path. The plan the remedy writes is asked of the same two paths wherever the plan keeps the source's own leg names: for an org: edit, the one adopt-project.py: without --local-remote-dir, check asks gh repo view about the two legs of another repository's plan, and its NOTE says execute asks again, where execute refuses plan-not-the-source first and asks nothing (follows #1210) #1232, 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's rows and the relative-url rows measure, and for a legs.assembly: or project: edit. A plan written for another project, with its own legs.spec: and legs.code:, is asked of that project's paths: measured by RV1232, with <dir>/Thing-spec.git present, check lists leg-remote-exists for it beside plan-not-the-source on both heads. That is a path the plan plan --project <source> writes does not name, and the one the other remedy, a plan from the repository the plan is about, keeps. It is kept: it is true of the disk, and it asks no remote. adopt-folder-cannot-be-made names the directory, which no mend changes.
One round shows both. This is the same reasoning check already applies to the topic finding (family.py, adopt-project.py and the four tools' --org: init asks GitHub's limits on the topic, the --org and the holder's name before any repository exists, check finds what execute refuses from one list, and an --org GitHub gives no organisation (., a trailing hyphen, --) is refused where GitHub is asked (#863) #871's review, S1). Skipped, the finding comes back only after the mend, in a second round.

The other way, skipping under both, is mutant M2 below. The row test_a_local_remote_dir_is_still_asked_beside_another_repository kills it.

One case the skip is blunter than the question. A plan that names another repository only by its legs.assembly: (or project:) keeps the source's org: and the legs' names. On 791106c, check asked GitHub about testorg/Northwind-spec and -code, the legs the plan the remedy writes names (measured: the another-name probe). It is skipped there too:

execute asks GitHub nothing about such a plan (adopt-project.py plan never compares --project (or --org) with the repository it adopts in place, so a source named billing-service is refused with a remediation that steers to --project BillingService, which plans and checks plan ok against acme/billing-service, and execute then acts on acme/BillingService: a raw git ls-remote failure when no such repository exists, and, when a different repository of that name does, both legs, the split branch, the pull request and the topic on that repository #499);
the NOTE said it does;
the question is asked again once the plan is mended.

Measured by RV1232, with a gh that answers {"nameWithOwner": ...}: on the legs.assembly: Elsewhere plan, main's check listed plan-not-the-source and a leg-remote-exists for each of testorg/Northwind-spec and -code (3 findings), and this head lists plan-not-the-source alone (1 finding). Those are the legs the remedy's plan names, so a person meets them after the mend, one round later. #1232's first "Done when" asks for no gh call on any plan for another repository, so the skip holds there too.

Measured, before and after

The probe is probe1232.py in my scratch. It builds each plan with the test suite's own helpers (planned_with with a held source, whose origin names testorg/Northwind). It runs check and, for the plans for another repository, execute --yes --work-dir, with:

a stub gh first on PATH that logs its arguments and exits 97;
GH_TOKEN=bogus and an empty GH_CONFIG_DIR;
GIT_SSH_COMMAND=false and an empty global git config;
git-filter-repo 2.47.0, and LANES_LANE unset.

"Another (org)" is org: acme. "Another (assembly)" is legs.assembly: Elsewhere. "lrd" is --local-remote-dir: absent (a directory not made yet), "taken" (holding Northwind-spec.git), or "a file".

Plan, mode	`check` on `791106c`	`check` here	`execute` (both heads)
another (org), no lrd	exit 1, `plan-not-the-source`; gh: `repo view acme/Northwind-spec`, `repo view acme/Northwind-code`; 2 NOTEs	exit 1, `plan-not-the-source`; gh: none; no NOTE	exit 2, `plan-not-the-source`; gh: none
another (assembly), no lrd	exit 1, `plan-not-the-source`; gh: `repo view testorg/Northwind-spec`, `-code`; 2 NOTEs	exit 1, `plan-not-the-source`; gh: none; no NOTE	exit 2, `plan-not-the-source`; gh: none
another (org), lrd absent	exit 1, `plan-not-the-source`	the same bytes	exit 2, `plan-not-the-source`
another (org), lrd taken	exit 1, `plan-not-the-source`, `leg-remote-exists`	the same bytes	exit 2, `plan-not-the-source`
another (org), lrd a file	exit 1, `plan-not-the-source`, `adopt-folder-cannot-be-made`	the same bytes	exit 2, `plan-not-the-source`
another (assembly), the three lrd modes	as the org rows	the same bytes	exit 2, `plan-not-the-source`
own, no lrd	exit 0; gh: `repo view testorg/Northwind-spec`, `-code`; 2 NOTEs	the same bytes	not run
own, lrd absent	exit 0	the same bytes	not run
own, lrd taken	exit 1, `leg-remote-exists` (the same line as in "another (org), lrd taken")	the same bytes	not run
own, lrd a file	exit 1, `adopt-folder-cannot-be-made`	the same bytes	not run

Byte for byte, with paths and hashes masked: 10 of the 12 probe sources are identical, all 20 runs' exits included. The other 2 are the two "no lrd" plans for another repository. In each, only check changed:
- its gh log, two repo view calls to none;
- its stdout, the two NOTE lines gone.
Its stderr, its exit and execute's whole output are the same. No probe made a leg.
Where the issue's measurement and mine differ: nowhere. RV1210 measured the org: edit on 90eca414, whose adopt-project.py is the same file as 791106c's (sha256 4c8c1c9eb0c5). I also measured the legs.assembly: edit and the three --local-remote-dir modes, which the issue did not.
Why execute was not run on the source's own plan: without --local-remote-dir it goes on to read the remote, which a probe must not.

What changed, in `adopt-project.py`

cmd_check only, the leg question (#798):

another = PLAN_NOT_THE_SOURCE in {problem.refusal.code
                                  for problem in unmended}
legs, fixes, unasked = (
    ([], [], []) if RENAMES_THE_LEGS.intersection(held_codes)
    or (another and local_remote_dir is None)
    else _leg_remote_findings(plan, names, local_remote_dir))

unmended is 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's list of the rule's problems (_against_its_source).
The comment above it says why, beside the one for RENAMES_THE_LEGS.
No refusal, finding, NOTE or remediation changes its words. execute is not touched.
14 lines added, none removed. On 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's head the file goes from sha256 4c8c1c9eb0c5 to 81efae25c83b (my commit e6f2526e). On main 106f92ff and on this head it goes from f42c0e973c33 to 7f0a247ffb65.
The change is unchanged by the restack. The no-context diff of 106f92ff..<this head>, without @@ and index lines, is identical to that of 791106c2..e6f2526e: 152 changed lines in the same 2 files.

Rows

All in tests/test_adopt_plan_against_its_source.py, #1210's module for the rule between check and execute, for three reasons:

adopt-project.py: without --local-remote-dir, check asks gh repo view about the two legs of another repository's plan, and its NOTE says execute asks again, where execute refuses plan-not-the-source first and asks nothing (follows #1210) #1232 names it;
the row also keeps 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's rule (check's first finding is execute's refusal);
its helpers build and read exactly these plans: planned_with's held source, ANOTHER_ORG, named_first, code_of.

A short paragraph is added to the module docstring, because these rows are the exception to its "with the same --local-remote-dir". Each row runs the tool with a gh of its own first on PATH, ahead of the session's guard, that logs each call's arguments and exits 97 (with_a_logging_gh). They skip on Windows, where that #! script cannot run.

Row	What it pins	`791106c`	here
`test_check_asks_github_nothing_about_another_repositorys_legs[org]`	no lrd, `org: acme`: no `gh` call by `check` or `execute` (the row fails if either log holds a `repo view`, or any call), no NOTE, `check` exit 1 and `execute` exit 2, both naming `plan-not-the-source` first	FAILS: `(['repo view acme/Northwind-spec --json nameWithOwner', 'repo view acme/Northwind-code --json nameWithOwner'], []) != ([], [])`	passes
`test_check_asks_github_nothing_about_another_repositorys_legs[assembly]`	the same, `legs.assembly: Elsewhere`	FAILS: the same, `testorg/Northwind-spec`, `-code`	passes
`test_a_local_remote_dir_is_still_asked_beside_another_repository`	the choice: under lrd with the spec leg's bare repository present, `check` lists `plan-not-the-source` then `leg-remote-exists`, the very finding it lists for the source's own plan; no `gh` call	passes	passes
`test_check_still_asks_github_about_the_sources_own_legs`	the control (#1232's last "Done when"): the source's own plan, no lrd, is asked `repo view testorg/Northwind-spec` and `-code`, prints both NOTEs word for word, and passes	passes	passes

On 791106c's adopt-project.py (swapped in and restored with cmp): 2 failed, 2 passed.
Here: 4 passed; the module, 22 passed.

Mutants

Run on my commit e6f2526e, on #1210's head, before the restack. Each was applied alone to adopt-project.py, the module was run (22 rows), and the file was restored, checked by cmp and sha256 (81efae25c83b, OK each time).

Mutant	Killed by
M1, the skip removed (`791106c`'s behaviour)	2: both `asks_github_nothing` rows
M2, the skip under both modes (`or another`)	1: `test_a_local_remote_dir_is_still_asked_beside_another_repository`
M3, the skip for every plan without lrd (`or local_remote_dir is None`)	1: `test_check_still_asks_github_about_the_sources_own_legs`
M4, the skip keyed on `plan-stale` instead of `plan-not-the-source`	2: both `asks_github_nothing` rows

Gates

Run on this head cf658506 (tree c23c2160, the work tree clean), one pytest at a time and never the whole suite, in the environment the probe used:

a stub gh (exit 97) first on PATH;
GH_TOKEN=bogus and an empty GH_CONFIG_DIR;
GIT_SSH_COMMAND=false and an empty global git config;
git-filter-repo 2.47.0;
LANES_LANE unset;
git 2.43.0.

The groups:

the rows' module with the modules it imports from and the ones that share their fixtures;
every module that greps a code whose findings this change gates: plan-not-the-source, leg-remote-exists, adopt-remote-unreadable, adopt-folder-cannot-be-made, and the NOTE's "could not ask GitHub";
every test module 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 changed;
the guards;
test_scaffold_leg_path_collisions.py;
test_shape_doctor.py, once.

Modules that import a fixture from another ran in one pytest with it.

Group	Files	Result
g1, the rows	`test_adopt_plan_against_its_source.py`, `test_adopt_plan_value_order.py`, `test_adopt_root_symlinks.py`, `test_adopt_pins.py`	420 passed, 3 skipped
g2, `plan-not-the-source`	`test_adopt_check_plan_name.py`, `test_adopt_plan_and_remote_paths_spelled.py`, `test_adopt_project_is_the_source.py`, `test_adopt_relative_submodule_url.py`, `test_adopt_split_branch.py`	298 passed (lane 3's run for RV1232's NIT-1; without `test_adopt_plan_and_remote_paths_spelled.py`, which came onto main in `8bcdb2c6` and onto this branch with the restack, the other four gave 150 passed)
g3, the legs	`test_adopt_execute_refusals.py`, `test_adopt_branch_names.py`, `test_adopt_e2e.py`, `test_adopt_filter_repo_failure.py`, `test_adopt_gitmodules_values.py`	435 passed
g4, the legs	`test_adopt_history_gitlinks.py`, `test_adopt_late_failure.py`, `test_adopt_mount_ignored.py`, `test_adopt_root_shape_paths.py`, `test_ruleset_hint_tracking_branch.py`	305 passed
g5, unreadable and folder	`test_adopt_github_exits.py`, `test_adopt_source_clone_record.py`, `test_adopt_temp_work_roots.py`, `test_adopt_unsearchable_source.py`	176 passed
g6, limits and collisions	`test_github_limits_init_check_org.py`, `test_github_name_limits.py`, `test_scaffold_leg_path_collisions.py`	264 passed
g7, the guards	`test_repo_hygiene.py`, `test_cli_reference.py`, `test_agents_md_codes.py`, `test_git_output_decoding.py`, `test_stdout_encoding.py`, `test_path_guard_hygiene.py`, `test_no_auto_maintenance.py`, `test_update_add_codes_documented.py`	470 passed, 3 skipped
m1305, both PRs' rows, on the resolved merge `dd699cc2` with #1305	g1's four and `test_adopt_entry_paths.py`	521 passed, 3 skipped (the same 3 reftable skips as g1). Its module holds both PRs' rows: the 4 here and #1305's at `2edbfcb5`.
g8, the doctor	`test_shape_doctor.py`	446 passed, 3 skipped in 904 s (all 3 at `test_shape_doctor.py:10516`, "the PowerShell line is printed, and powershell.exe exists, only on a Windows host")

The skips are the host's, not this PR's:

g1's 3 are in test_adopt_root_symlinks.py: "the reftable backend needs git 2.45 or later".
g7's are test_stdout_encoding.py:292 ("the default this is about is Windows' own") and 2 at test_no_auto_maintenance.py:222 (the run sets GIT_CONFIG_GLOBAL).

Before the restack, on my commit e6f2526e alone, g1 gave 368 passed and g2 150 passed. That chain was stopped when #1210 landed, and everything was run again here.

Merges

The restack, as new commits, never a rebase:

my commit e6f2526e on 791106c2;
8e15daa9, merging 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's final head 27c8f67b (791106c merged with main 3add78e1): clean;
cf658506, merging 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's squash 106f92ff.
- Its tree is 27c8f67b's (ee6de2a2, and git diff 27c8f67b 106f92ff is empty), so this merge changes no file.
- tests/test_adopt_plan_against_its_source.py, which 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 added, met as an add/add conflict and was taken ours, after measuring that the squash holds 791106c2's bytes of it and ours holds those bytes plus this branch's own change, nothing else. adopt-project.py merged on its own, to the squash's file plus this change.
Main is 106f92ff, the squash itself, so merging main again was "Already up to date".

git merge-tree --write-tree (git 2.43.0) of this head cf658506:

with main 106f92ff: clean, tree c23c2160, which is this head's own tree. git diff --name-only 106f92ff cf658506 lists only this PR's 2 files.
with each open PR that touches either file, beside the same merge of main alone (main now holds 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):

PR     head       with this head                         main alone
#1171  ce1dd661f  clean (tree 3814f2720)                 clean (tree 292ca9412)
#1194  906172b41  clean (tree 96ae61f27)                 clean (tree e24da7a3b)
#1225  cc230f2f1  clean (tree aa789ac39)                 clean (tree f4e9727e1)
#1237  fa6d189ff  clean (tree 3ef7d65c7)                 clean (tree 9e3db4161)
#1249  3e5d1fde1  clean (tree 441d5c687)                 clean (tree f6e21799c)
#1265  39b21c07e  clean (tree 03c9d022a)                 clean (tree 2fcc3c189)
#1280  5cb1abdbc  CONFLICT adopt-project.py              CONFLICT adopt-project.py
#1294  cf07ba0d1  clean (tree 4ba47dd60)                 clean (tree f9a1c9d6f)
#1297  2e020b458  clean (tree a81e10be9)                 clean (tree 1e8b34927)
#1304  df942efee  clean (tree 47edaa910)                 clean (tree 34d58d883)
#1305  2edbfcb54  CONFLICT the test module               CONFLICT adopt-project.py and the test module

adopt-project.py: a source whose only remote is not named origin, as clone.defaultRemoteName names it, is read from that remote, so plan reads its organisation, default branch and branches where it refused adopt-no-org, took HEAD's branch as the default and asked --tracking-branch of the source's own branches, and one with several remotes and no origin names them and guesses none (#1276) #1280: not this PR's. Its 3 conflict hunks in adopt-project.py (around lines 3620, 3715 and 5818) are the same, byte for byte, with this head and with main alone. This change sits at about line 14951. It is adopt-project.py: a source whose only remote is not named origin, as clone.defaultRemoteName names it, is read from that remote, so plan reads its organisation, default branch and branches where it refused adopt-no-org, took HEAD's branch as the default and asked --tracking-branch of the source's own branches, and one with several remotes and no origin names them and guesses none (#1276) #1280's own conflict with main.
adopt-project.py: check lists an unanswered entry with the plan's problems against its source, before the split branch and the entries, so beside an uncovered path, an unsafe path, a bad leg path, the split branch taken or another entry finding its first finding is execute's refusal, plan-unresolved (#1197) #1305 (lane 3's, for adopt-project.py: check and execute still name different first problems for a plan with an unanswered entry beside an uncovered path, plan-uncovered and plan-unresolved, a pair #1113's four do not cover (follows #1113) #1197), also built on 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's head, at 2edbfcb5:
- With main alone it conflicts in both files, because it has not been restacked yet: its own.
- With this head, one conflict is this PR's, in tests/test_adopt_plan_against_its_source.py, the import block only. adopt-project.py: check lists an unanswered entry with the plan's problems against its source, before the split branch and the entries, so beside an uncovered path, an unsafe path, a bad leg path, the split branch taken or another entry finding its first finding is execute's refusal, plan-unresolved (#1197) #1305 adds re, make_source_repo, planned, ANSWERS and BAD_LEG_PATH; this PR adds os, ADOPT, run_script and PLAN_FILE. The resolution is the union:
```
import os
import re

import pytest

from conftest import ADOPT, git, make_source_repo, planned, run_script
from test_adopt_pins import ANSWERS
from test_adopt_plan_value_order import (
    ANOTHER_ORG, BAD_LEG_PATH, BAD_TRACKING_BRANCH, PLAN_FILE, REFUSED_NAME,
    both_on, edited, findings_of, named_first, needs_filter_repo,
    planned_with)
```
- The docstring paragraphs, the rows and adopt-project.py merge on their own: this PR's hunks in cmd_check come 33 unchanged lines after adopt-project.py: check lists an unanswered entry with the plan's problems against its source, before the split branch and the entries, so beside an uncovered path, an unsafe path, a bad leg path, the split branch taken or another entry finding its first finding is execute's refusal, plan-unresolved (#1197) #1305's nearest, in 791106c2's numbering.
- The resolved merge, dd699cc2 (tree 98625a93, never pushed), adds over this head exactly adopt-project.py: check lists an unanswered entry with the plan's problems against its source, before the split branch and the entries, so beside an uncovered path, an unsafe path, a bad leg path, the split branch taken or another entry finding its first finding is execute's refusal, plan-unresolved (#1197) #1305's change: 3 files, +239/-25. Only the import lines differ, as the union.
- Both PRs' rows in it: see Gates.
- Before the restack, the same merge with adopt-project.py: check lists an unanswered entry with the plan's problems against its source, before the split branch and the entries, so beside an uncovered path, an unsafe path, a bad leg path, the split branch taken or another entry finding its first finding is execute's refusal, plan-unresolved (#1197) #1305's earlier head 142ca405 gave the same one conflict, and on 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's head 791106c2 alone adopt-project.py: check lists an unanswered entry with the plan's problems against its source, before the split branch and the entries, so beside an uncovered path, an unsafe path, a bad leg path, the split branch taken or another entry finding its first finding is execute's refusal, plan-unresolved (#1197) #1305 merges clean.
- adopt-project.py: check lists an unanswered entry with the plan's problems against its source, before the split branch and the entries, so beside an uncovered path, an unsafe path, a bad leg path, the split branch taken or another entry finding its first finding is execute's refusal, plan-unresolved (#1197) #1305 then moved to 28f6f669, its own restack (2edbfcb5 merged with 27c8f67b and 106f92ff). Its change over main is line for line 2edbfcb5's over 791106c2.
  - Main with it is now clean (tree 8f9ded6c).
  - This head with it gives the same conflicted tree (adbf4af2) as with 2edbfcb5, and the same union resolves it to the same tree, 98625a93, the one m1305 tested.

Independent review

RV1232, lane 3's independent opus review (issuecomment-6769621392), reads READY AFTER FIXES at cf6585063. The fixes are to the body only, and it found no MUST. It judged both choices right:

keeping the disk question under --local-remote-dir, which is what adopt-project.py: without --local-remote-dir, check asks gh repo view about the two legs of another repository's plan, and its NOTE says execute asks again, where execute refuses plan-not-the-source first and asks nothing (follows #1210) #1232's "To settle" asks, with a row on it;
skipping a plan edited only in legs.assembly:, which adopt-project.py: without --local-remote-dir, check asks gh repo view about the two legs of another repository's plan, and its NOTE says execute asks again, where execute refuses plan-not-the-source first and asks nothing (follows #1210) #1232's first "Done when" asks for.

Its gates ran from lane 3's main loop at this head with RV1232's own commands, one pytest at a time:

its g1: 420 passed, 3 skipped;
its g2: 298 passed;
its g3, the guards: 470 passed, 3 skipped.

Taken here: SHOULD-1, the bullet on leg paths; SHOULD-2, the paragraph measured by RV1232; and NIT-1, the g2 row.

Not taken:

NIT-3 (the test module names 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's head 791106c2), to keep the reviewed head;
NIT-4 (docs/cli.md:684), which is out of scope. Lane 3 names it to Brett Heap as a follow-up candidate.

Closes #1232

Lane: openreposhape-3 (openRepoShape-3)

🤖 Generated with Claude Code

Summary by Sourcery

Align check’s remote-leg validation with execute by skipping unreachable GitHub checks for plans targeting another repository while retaining local and source-plan validation.

Bug Fixes:

Prevent adopt-project.py check from querying GitHub or emitting misleading notes about remote legs when the plan is for another repository and execute will refuse it before any remote access.

Enhancements:

Continue checking local bare repositories under --local-remote-dir while preserving GitHub checks for the source repository’s own plan.

Tests:

Add coverage for GitHub-query suppression, local remote-directory checks, and unchanged source-plan behavior.

… 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) #901 made `check` and `execute` name the same first problem among a plan's options. Outside them, the two still named different problems first (#1020, found by lane 1's writer on #901). 1. **One rule, the refusal before what `check` lists.** `check` refuses a bad `org:` or `id:` outright, exit 2, and lists none of its findings. `execute` refuses at the first of them, but it asked the two only after it had refused four things that `check` lists as findings: a plan for another repository (`plan-not-the-source`), a name the naming policy refuses (`naming-unclassified`), an unanswered entry (`plan-unresolved`) and a source that has moved on (`plan-stale`). And `check` asked the two only where the policy took the plan's names (`_manifest_values`), so an unsafe `org:` beside a refused name was a finding about the name, or about another repository. Both commands now ask the two first of the plan's values, right after the three repository names (`plan.names()`, then `_checked_org_and_id`): `execute` in `cmd_execute`, ahead of `_refuse_a_plan_for_another_repository`, and `check` at its head, after the triad advisory. `_checked_plan_values` no longer asks them, since `cmd_execute` has. 2. **#716's L4-48, decided.** An `org:` that `checked_value` refuses, held to a repository, was `unsafe-value` to `check` and `plan-not-the-source` to `execute`. Such an `org:` names no repository, on GitHub or in the manifest, so it is not compared with the source's. It is `unsafe-value` in both commands now, held to a repository or not. Once it is mended, a plan for another repository is still `plan-not-the-source` in both. This reverses three tests in `tests/test_adopt_check_plan_name.py`, which pinned the old answers on purpose: - `test_an_unsafe_org_is_unsafe_in_check_and_compared_by_execute_if_held` expected `plan-not-the-source` from `execute` in its six held rows. As `test_an_unsafe_org_is_unsafe_in_check_and_in_execute_held_or_not`, all twelve expect `unsafe-value` from both, in the same words. - `test_check_compares_an_unsafe_org_where_no_manifest_is_built` expected `check` to compare `org: acme labs` beside the refused name `bad_name`: `plan-not-the-source` and `naming-unclassified`, exit 1. As `test_an_unsafe_org_is_asked_before_the_plans_refused_names`, it expects `unsafe-value`, exit 2, from `check` and from `execute`. - `test_check_asks_an_unsafe_org_first_where_a_kept_submodule_url_is_relative` expected `plan-not-the-source` from `execute`. It now expects `unsafe-value`, in `check`'s words. 3. **A `pins:` entry that is no pin** was spelled with `repr` (`` `pins:` entry 1 is 7 ``, `True`, `None`). #901 made `pins:` itself read as `_read_as` reads it, and an entry reads the same way now: "the number 7", "the boolean true", "null", "a list of 2 item(s)". Docstrings and comments now say this: `_checked_org_and_id`, `_checked_plan_values`, `_manifest_values`, `_held_findings`, `_not_the_source`, `_refuse_a_plan_for_another_repository`, `_read_pins` and the #499 block comment. No finding is added or removed: only which problem is named first changes, and the spelling of an entry. Measured on main at `6da757cb` and on this branch, with a stub `gh`, a bogus token, an empty `GH_CONFIG_DIR`, `GIT_SSH_COMMAND=false` and `--local-remote-dir`. No run made a leg. | The plan has | `check` on main | `execute` on main | both, here | |---|---|---|---| | a bad id beside an unanswered entry | `plan-bad-id` | `plan-unresolved` | `plan-bad-id` | | a bad id beside a source that has moved on | `plan-bad-id` | `plan-stale` | `plan-bad-id` | | a bad id beside a name the policy refuses | `plan-bad-id` | `naming-unclassified` | `plan-bad-id` | | a bad id beside another organisation, held to a repository | `plan-bad-id` | `plan-not-the-source` | `plan-bad-id` | | no `id:`, and `legs.assembly: north_wind` (a derived id the policy refuses too) | `plan-bad-id` | `naming-unclassified` | `plan-bad-id` | | an unsafe `org:` held to a repository (L4-48) | `unsafe-value` | `plan-not-the-source` | `unsafe-value` | | an unsafe `org:` beside an unanswered entry | `unsafe-value` | `plan-unresolved` | `unsafe-value` | | an unsafe `org:` beside a source that has moved on | `unsafe-value` | `plan-stale` | `unsafe-value` | | an unsafe `org:` beside a refused name | finding `naming-unclassified`, exit 1 | `naming-unclassified` | `unsafe-value` | | the same, held to a repository | finding `plan-not-the-source`, exit 1 | `plan-not-the-source` | `unsafe-value` | Each "here" row is exit 2 from both, word for word. The last two rows agreed on main. They change because `check` now asks the two where the policy refuses the names, as `execute` now does: without that, moving `execute` alone would have made them disagree. Plans with only one of these problems (an unanswered entry, a moved source, a refused name, another repository) are named as on main by both. **Tests.** - `tests/test_adopt_plan_value_order.py` gains 16 rows: - the 10 above, through both commands, comparing the first line; - four `pins:` entries that are no pin: number, boolean, null and list; - two guard rows, `id` and `org`: a leg name `checked_value` refuses (`legs.spec: "-spec"`) beside a bad `id:` or `org:` is that name's `unsafe-value` in both, so the three names are still asked first. - The 14 non-guard rows all fail against `6da757cb`'s `adopt-project.py` and pass here. The two guard rows pass on both, by design. - Existing rows changed for the new behaviour: - `test_adopt_check_plan_name.py`: the three tests above, and its docstring's L4-48 line. - `test_adopt_pins.py`: `a-number`. - `test_adopt_execute_refusals.py`: the three entry rows of `PIN_ENTRIES`. Closes #1020 Lane: openreposhape-4 (openRepoShape-4) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

… right after the plan's keys and before `_work_root`, so a bad one beside an unwritable `--work-dir` is that one refusal in both (#1020) RV1125's SHOULD-1 on #1125. A bad `id:` or `org:` beside a `--work-dir` this user cannot write in was `plan-bad-id` (or `unsafe-value`) to `check` and `adopt-work-dir-unwritable` to `execute`: `check` lists the directory as a finding (#524), and `execute` refuses it before it opens the source, ahead of the two asks, which #1125 had placed after `_work_root`. The comments in `cmd_execute` and `_checked_org_and_id` said that cannot happen. - `cmd_check` asks `plan.names()` and `_checked_org_and_id(plan)` right after its mistyped-plan question, before `_work_root`, and `cmd_execute` asks them right after `_refuse_a_mistyped_plan`, before `_work_root`. Both read only the plan. `execute` now asks them before `_require_filter_repo()` too. - A side effect, said in the body: `check` printed the Triad advisory, once it had opened the source, before it refused a bad `org:` or `id:`; it prints nothing now, since the refusal comes before the source is opened. - The comments and docstrings that placed the asks "at its head, after the triad advisory" or "right after the names" say where they are now. - tests/test_adopt_plan_value_order.py: `both_on` takes a `--work-dir` for both commands; one row per bad `id:` and bad `org:` beside an unwritable `--work-dir` (skipped where `chmod` cannot keep this user from writing); and, for NIT-2, one row per value that asks `check` alone and holds stdout empty, the position being told by the Triad advisory the unedited plan prints first. - tests/test_adopt_check_plan_name.py: the docstring of the kept-submodule row says it is a regression row since `check` asks `org:` at its head (NIT-1). Lane: openreposhape-4 (openRepoShape-4) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

…blems 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) `check` listed a moved source first, then the plan's entries, then another repository and then the names; `execute` refused another repository, a refused name, an unanswered entry and then a moved source. So for five of the six pairs of the four, `check`'s first finding was not `execute`'s refusal: the four of #1113's table, and a refused name beside a moved source. `_against_its_source` asks the four in one order, and both commands read it: `execute` refuses its first problem, and `check` lists each, in that order and in its own words, before the split branch and the entries. Each of the first three is mended by running `plan` again, which takes back every answer, so an unanswered entry is named last; another repository comes first, before the naming policy or any remote is asked (#499); and a refused name is mended by a `plan` that reads the source as it is now, so it comes before a moved source. `_refuse_an_unrunnable_plan` asks a moved source before an unanswered entry. No refusal's code or wording changes: only which is named first. tests/test_adopt_plan_against_its_source.py runs both commands on one plan per pair, on all four, and on each of the four alone. In tests/test_adopt_relative_submodule_url.py, six rows hand-edit a plan's `org:` to another organisation beside a relative url, and pinned `check`'s old order, the url first. Such a plan is for another repository, which `execute` refuses first (#499), so the rows now expect `plan-not-the-source` first to `check` too. A new row runs `check`, with and without `--local-remote-dir`, and `execute` on one such plan, and asserts that each names it first and that no leg is made. Lane: openreposhape-4 (openRepoShape-4) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…le's problems beside the split branch taken and beside an unsafe `tracking_branch:` pins that `check` lists the rule's problem, and its remediation, first and `execute` refuses it first, so the rule moved after either finding fails a row (#1113) RV1210 measured two placements of the rule that no row pinned, each bringing back a disagreement #1210 fixes: - `check` listing the rule's problems after the split branch: another repository, a refused name and a moved source beside a branch `adopt` in the source, as `test_adopt_split_branch.py` makes it (`another-and-taken`, `name-and-taken`, `moved-and-taken`); - `execute` asking the rule after `_checked_plan_values`, `_repository_urls` and `_refuse_an_unwritable_clone_url`: each of the four beside `tracking_branch: "a b"`, which `BAD_TRACKING_BRANCH` from `test_adopt_plan_value_order.py` edits in (`another-and-tracking`, `name-and-tracking`, `moved-and-tracking`, `unanswered-and-tracking`). Each row asserts that `check` exits 1 and names the rule's problem first, and lists the ordinary finding after it, and that `execute` exits 2 and refuses the rule's problem first. `another-and-taken` also asserts that `check`'s first `Remediation:` line is the rule's (NIT-4), so the remediation lines follow the findings' order. `unanswered-and-taken` stays out: it is #1197's. Test only; adopt-project.py is unchanged. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

…cute

… nothing about the two legs of a plan for another repository and prints no NOTE that `execute` asks again, as `execute` refuses `plan-not-the-source` first and asks nothing; under `--local-remote-dir` it still asks the disk (#1232) On #1210's head 791106c, `check` on a plan for another repository, given no `--local-remote-dir`, ran `gh repo view <org>/<Project>-spec --json nameWithOwner` and the same for `-code`, under the plan's own `org:`, and where gh could not answer it printed a NOTE that `execute` asks again before it makes either leg. `execute` refuses `plan-not-the-source` first and asks no remote anything (#499). `cmd_check` now skips the leg question (`_leg_remote_findings`) there, as it skips it after a finding in `RENAMES_THE_LEGS`. The source's own plan is asked as before. Under `--local-remote-dir` the disk is still asked beside `plan-not-the-source`: it asks no remote, and a leg's bare repository is `<dir>/<name>.git`, which no `org:` is part of. Measured, the `leg-remote-exists` it lists beside `plan-not-the-source` names the very path the source's own plan, the one the remedy writes, is refused for. Tests: tests/test_adopt_plan_against_its_source.py gains 2 rows that fail on 791106c (a plan for another repository by its `org:` and by its `legs.assembly:`, under a `gh` that logs each call), a row for the choice under `--local-remote-dir`, and a control for the source's own plan. Lane: openreposhape-3 (openRepoShape-3) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The first of the three restack merges after #1210 landed by squash (106f92f): its final head, 791106c merged with main 3add78e. Lane: openreposhape-3 (openRepoShape-3) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The second of the three restack merges: #1210 as it landed on main. Its tree is the final head's (ee6de2a), so this merge changes no file. tests/test_adopt_plan_against_its_source.py, which #1210 added, met as an add/add conflict and was taken ours: the squash holds 791106c's bytes of it, and ours holds those bytes and this branch's own change, nothing else. Lane: openreposhape-3 (openRepoShape-3) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

sourcery-ai · 2026-11-10T13:51:40Z

Reviewer's Guide

Implementation

Skip leg-remote checks for plans refused as plan-not-the-source when no local remote directory is supplied, while retaining disk checks for local directories.

adopt-project.py L14641-14657

Tests

Document the exception and add coverage for skipping GitHub questions, retaining local-disk checks, and preserving checks for the source repository’s own legs.

tests/test_adopt_plan_against_its_source.py L52-63, L68-75, L190-307

Tips and commands

Interacting with Sourcery

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

Customizing Your Experience

Access your dashboard to:

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

Getting Help

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

sonarqubecloud · 2026-11-10T13:52:35Z

Quality Gate passed

Issues
0 New issues
0 Accepted issues

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

See analysis details on SonarQube Cloud

brettheap · 2026-11-10T14:15:04Z

(lane openRepoShape-3, 2026-11-10T14:15:00Z) RV1232: CHECKPOINT at cf65850

LAST STEP reached me during item 4, the gates. Items 1, 2, 3, 5 and 6 are measured. None of item 4 is. Verdict so far: READY AFTER FIXES. The fixes are to the body only, and the verdict holds only if the gates pass. I found no MUST.

PR: #1313, #1313. Branch l3-1232-not-the-source-asks-no-leg. The head reads back from GitHub as cf65850634768426f36af3bd00a0a513433d2da9, and main is 106f92ff9585f00ddbb2151f8a3a4140ae6c4e69 both locally and on GitHub.

The two choices

Choice 1: under --local-remote-dir, the disk is still asked beside plan-not-the-source. RIGHT, and it is what #1232's "To settle" asks. The issue asked the PR to say which way and to put a row on it. The PR does both:

The row is test_a_local_remote_dir_is_still_asked_beside_another_repository. It kills the writer's M2.
Under --local-remote-dir, every output is byte-identical before and after (11 of 11 check runs, and every execute; see item 2).
The disk answer is true and no remote is asked, so adopt-project.py plan never compares --project (or --org) with the repository it adopts in place, so a source named billing-service is refused with a remediation that steers to --project BillingService, which plans and checks plan ok against acme/billing-service, and execute then acts on acme/BillingService: a raw git ls-remote failure when no such repository exists, and, when a different repository of that name does, both legs, the split branch, the pull request and the topic on that repository #499's reason does not apply.

One of the three reasons the body gives goes too far (SHOULD-1). "The plan the remedy writes is asked of the same two paths" is true only when the plan's legs.spec: and legs.code: are the source's own names. I measured the case where they are not: a plan written for another project, with project: Thing, legs: Thing / Thing-spec / Thing-code, and <dir>/Thing-spec.git present. On both heads, check lists:

plan-not-the-source;
FINDING leg-remote-exists: <dir>/remotes/Thing-spec.git, the spec leg's bare repository, already exists;
Remediation: Choose an empty --local-remote-dir, ....

The plan the first remedy writes (--project Northwind --org testorg) never names that path. The second remedy ("plan from a clone of the one that is") does keep it. So keeping the question is still defensible, but the body should say so.

Choice 2: a plan edited only in legs.assembly: is skipped too. ACCEPTABLE. Two things require it:

adopt-project.py: without --local-remote-dir, check asks gh repo view about the two legs of another repository's plan, and its NOTE says execute asks again, where execute refuses plan-not-the-source first and asks nothing (follows #1210) #1232's first "Done when" asks for no gh call on any plan for another repository, with no exception.
adopt-project.py plan never compares --project (or --org) with the repository it adopts in place, so a source named billing-service is refused with a remediation that steers to --project BillingService, which plans and checks plan ok against acme/billing-service, and execute then acts on acme/BillingService: a raw git ls-remote failure when no such repository exists, and, when a different repository of that name does, both legs, the split branch, the pull request and the topic on that repository #499 says execute asks no remote anything about such a plan.

The body says why in three plain bullets. I measured what its "Not measured" paragraph leaves open, with a gh that answers nameWithOwner:

Main: 3 findings: plan-not-the-source, then leg-remote-exists for testorg/Northwind-spec and for testorg/Northwind-code, with ON_GITHUB_FIX.
Head: 1 finding(s), plan-not-the-source only.

These are the repositories the remedy's plan names, so the person now meets them one round later. That is acceptable, and the body should say it as measured (SHOULD-2).

Findings

SHOULD-1 (body). In "The choice under --local-remote-dir", replace the bullet "No org: is part of a leg's path. ..." with:

No org: is part of a leg's path. The plan the remedy writes is asked of the same two paths wherever the plan keeps the source's own leg names: for an org: edit, the one #1232, #1210's rows and the relative-url rows measure, and for a legs.assembly: or project: edit. A plan written for another project, with its own legs.spec: and legs.code:, is asked of that project's paths: measured by RV1232, with <dir>/Thing-spec.git present, check lists leg-remote-exists for it beside plan-not-the-source on both heads. That is a path the plan plan --project <source> writes does not name, and the one the other remedy, a plan from the repository the plan is about, keeps. It is kept: it is true of the disk, and it asks no remote. adopt-folder-cannot-be-made names the directory, which no mend changes.

SHOULD-2 (body). Replace the paragraph "Not measured. By the code, not by a run: ..." with:

Measured by RV1232, with a gh that answers {"nameWithOwner": ...}: on the legs.assembly: Elsewhere plan, main's check listed plan-not-the-source and a leg-remote-exists for each of testorg/Northwind-spec and -code (3 findings), and this head lists plan-not-the-source alone (1 finding). Those are the legs the remedy's plan names, so a person meets them after the mend, one round later. #1232's first "Done when" asks for no gh call on any plan for another repository, so the skip holds there too.

NIT-1 (body, Gates). g2 leaves out tests/test_adopt_plan_and_remote_paths_spelled.py. It names plan-not-the-source and has an another-repository row in its "github" mode (:350). It came onto main in 8bcdb2c6 (2026-11-09) and onto this branch through the restack; it is not in 791106c2. So the body's "every module that greps a code whose findings this change gates" does not hold yet. Lane 3 should run it (command below) and add it to g2 with its result line.

NIT-2 (body). After the robot line, the Sourcery bot added a "## Summary by Sourcery" block. Its lines 212 and 218 each hold U+2019. Everything else in the body is ASCII except the robot line, U+1F916 at line 208.

NIT-3 (code, optional; the head would change). The test module names 791106c2 4 times (:57, :233, :257, :286). That is #1210's PR head, and git merge-base --is-ancestor 791106c2 106f92ff says it is not on main, so a plain clone cannot show it. 106f92ff, #1210's squash, behaves the same in my probe, so "Before #1232", or naming 106f92ff, would point at something a reader can find.

NIT-4 (out of scope). The docs/cli.md:684 help still says that without --local-remote-dir, check asks GitHub. It also did not mention the existing RENAMES_THE_LEGS skip, so the new skip is consistent with it, but the sentence now has two exceptions it does not state.

Item 2: before and after (measured)

The probe ran check and, where marked, execute with main's adopt-project.py (sha256 f42c0e973c33) and with the head's (7f0a247ffb65) on the same plan. Each case had a fresh base, a logging gh stub first on PATH, and a proxy that goes nowhere. The stub's modes were exit 97, "exists" (answers nameWithOwner) and "notfound". The result is 16 cases, 28 runs on each head:

Changed: 5 check runs, all of them with no --local-remote-dir on a plan for another repository: org 97, org exists, assembly 97, assembly exists, whole-plan 97.
- The 97 runs only lose their two repo view calls and the two NOTE lines on stdout. Stderr and exit are unchanged.
- The "exists" runs also drop their two leg-remote-exists findings and the ON_GITHUB_FIX line, and the count goes from 3 to 1.
Identical: the other 11 check runs and all 12 execute runs (SAME True). That includes:
- the source's own plan without --local-remote-dir: on 97, gh was asked repo view testorg/Northwind-spec and -code, 2 NOTEs, exit 0; on exists, two leg-remote-exists, exit 1; on notfound, exit 0;
- its execute: exit 2 adopt-remote-unreadable without the flag, and exit 2 leg-remote-exists with the flag and a spec leg already there;
- every run under --local-remote-dir.
No probe made a leg. remotes after held only what the probe put there.

This matches the body's table and its "the same bytes" claims.

Item 3: the rows and mutants (measured)

Run	Result
the 4 new rows at the head	`4 passed, 18 deselected in 7.78s`
the 4 rows with main's file swapped in, then restored (cmp OK, `7f0a247ffb65`)	`2 failed, 2 passed, 18 deselected in 7.20s`; the two failures are the `asks_github_nothing` diffs (`acme/...` and `testorg/...`)

My three mutants were each applied alone. The whole module ran (22 rows), and the file was restored and checked with cmp (RESTORED 7f0a247ffb65 each time).

Mutant	Result	Killed by
M5: `gh` still asked, only the NOTE dropped beside another repository	`2 failed, 20 passed`	both `asks_github_nothing` rows
M6: skip only where `org:` is not the source's own	`1 failed, 21 passed`	`asks_github_nothing[assembly]` only, so the parameter earns its place
M7: `cmd_execute` runs `gh repo view <org>/<spec>` before refusing	`2 failed, 20 passed`	both rows, through `execute`'s log (`At index 1 diff: ['repo view testorg/Northwind-spec --json nameWithOwner'] != []`)

Item 5: merges (git 2.43.0, measured)

Main: clean, tree c23c21609037, which is the head's own tree.
adopt-project.py: check lists an unanswered entry with the plan's problems against its source, before the split branch and the entries, so beside an uncovered path, an unsafe path, a bad leg path, the split branch taken or another entry finding its first finding is execute's refusal, plan-unresolved (#1197) #1305 at 28f6f669: one conflict, in tests/test_adopt_plan_against_its_source.py, the import block only. The union of imports resolves it to tree 98625a93f328, the one the writer tested. Over the head it adds 3 files, +239/-25, as the body says. My conflicted tree is a046dcf8, not the body's adbf4af2. That difference is expected: merge-tree writes the commit names given on the command line into the conflict markers.
adopt-project.py: a source whose only remote is not named origin, as clone.defaultRemoteName names it, is read from that remote, so plan reads its organisation, default branch and branches where it refused adopt-no-org, took HEAD's branch as the default and asked --tracking-branch of the source's own branches, and one with several remotes and no origin names them and guesses none (#1276) #1280 at 5cb1abdb: it conflicts in adopt-project.py both with the head and with main alone. The 3 hunks, at :3620, :3715 and :5818, are identical once the marker labels are stripped. The only other difference is this PR's own 14 lines. The conflict is adopt-project.py: a source whose only remote is not named origin, as clone.defaultRemoteName names it, is read from that remote, so plan reads its organisation, default branch and branches where it refused adopt-no-org, took HEAD's branch as the default and asked --tracking-branch of the source's own branches, and one with several remotes and no origin names them and guesses none (#1276) #1280's.
Restack: confirmed.
- The -U0 diff of 106f92ff..cf658506 (without @@ and index lines) is byte-identical to that of 791106c2..e6f2526e: 152 lines (14 + 135 + 3).
- git diff 27c8f67b 106f92ff is empty, and so is the test module's diff between 791106c2 and 106f92ff.
- The sha256 values are as the body says: 4c8c1c9eb0c5 to 81efae25c83b, and f42c0e973c33 to 7f0a247ffb65.
- All 3 commits are authored as Brett Heap <service@opensoft.dev> and carry the Opus 5.5 trailer.

Item 6: the body (measured)

closingIssuesReferences reads [1232], and Closes #1232 is the only closing keyword.
It is ASCII except as NIT-2 says.
It says what a person sees change, both choices, and the restack.
Every number I re-measured matches, except the conflicted-tree id noted in item 5.

Not measured

g1 was stopped at LAST STEP (SIGTERM, exit 143) after 67 rows passed with no F or E printed. No pytest is running.
g2 and g3 were never started.
The doctor was not run (optional; the writer ran it).

Run these from the clone, one at a time:

X=/workspace/projects/openRepoShape-worktrees/_lane3-scratch/RV1232-scratch
cd /workspace/projects/openRepoShape-worktrees/_lane3-scratch/RV1232 || exit 1
$X/envrun.sh /dev/null /usr/bin/python3 -m pytest -q -p no:cacheprovider --basetemp=/tmp/claude-1000/l3bt/RV1232-g1 -rs tests/test_adopt_plan_against_its_source.py tests/test_adopt_plan_value_order.py tests/test_adopt_pins.py tests/test_adopt_root_symlinks.py > $X/g1.log 2>&1
$X/envrun.sh /dev/null /usr/bin/python3 -m pytest -q -p no:cacheprovider --basetemp=/tmp/claude-1000/l3bt/RV1232-g2 -rs tests/test_adopt_check_plan_name.py tests/test_adopt_plan_and_remote_paths_spelled.py tests/test_adopt_project_is_the_source.py tests/test_adopt_relative_submodule_url.py tests/test_adopt_split_branch.py > $X/g2.log 2>&1
$X/envrun.sh /dev/null /usr/bin/python3 -m pytest -q -p no:cacheprovider --basetemp=/tmp/claude-1000/l3bt/RV1232-g3 -rs tests/test_repo_hygiene.py tests/test_cli_reference.py tests/test_agents_md_codes.py tests/test_git_output_decoding.py tests/test_stdout_encoding.py tests/test_path_guard_hygiene.py tests/test_no_auto_maintenance.py tests/test_update_add_codes_documented.py > $X/g3.log 2>&1

Paths

Clone: /workspace/projects/openRepoShape-worktrees/_lane3-scratch/RV1232, detached at the head, clean. adopt-project.py matches the saved head copy by cmp.
Scratch: /workspace/projects/openRepoShape-worktrees/_lane3-scratch/RV1232-scratch:
- checkpoint: /workspace/projects/openRepoShape-worktrees/_lane3-scratch/RV1232-scratch/CHECKPOINT.md
- probe: /workspace/projects/openRepoShape-worktrees/_lane3-scratch/RV1232-scratch/probe/rvprobe.py
- probe results: /workspace/projects/openRepoShape-worktrees/_lane3-scratch/RV1232-scratch/probe/out/SUMMARY.txt
- mutants: /workspace/projects/openRepoShape-worktrees/_lane3-scratch/RV1232-scratch/mut/
- merges: /workspace/projects/openRepoShape-worktrees/_lane3-scratch/RV1232-scratch/merge/
- main worktree: /workspace/projects/openRepoShape-worktrees/_lane3-scratch/RV1232-scratch/wt-main, at 106f92ff
Basetemps: /tmp/claude-1000/l3bt/RV1232-*

Lane 3: g1, g2 and g3 run now from lane 3's main loop with RV1232's own commands, one at a time at cf6585063. Their lines will follow on this thread, and the body fixes wait for them.

brettheap · 2026-11-10T14:34:09Z

(lane openRepoShape-3, 2026-11-10T14:34:09Z) READY at cf6585063. Lane 3's main loop ran RV1232's three gates at this head with RV1232's own commands, one pytest at a time, each exit 0:

g1: 420 passed, 3 skipped;
g2, with tests/test_adopt_plan_and_remote_paths_spelled.py (NIT-1): 298 passed;
g3, the guards: 470 passed, 3 skipped.

The logs are in RV1232-scratch. So RV1232's verdict holds: READY AFTER FIXES, with the fixes to the body only.

Lane 3 took SHOULD-1, SHOULD-2 and NIT-1 into the body word for word, and added an "Independent review" section. It did not take two:

NIT-3, to keep the reviewed head;
NIT-4, which is out of scope; it is a follow-up candidate for Brett Heap.

closingIssuesReferences reads [1232], and the body is ASCII above the robot line. With the fixes in, #1313 is READY: v1.2.1 is set, and it is marked ready. Lane 3 places it 31st by its READY time, after #1305 (#473).

sourcery-ai · 2026-11-10T14:35:01Z

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

You can request another review in 4 days and 6 hours by commenting @sourcery-ai review. Upgrade to get a review now.

Brett Heap and others added 8 commits October 21, 2026 13:59

Merge branch 'main' into fix/1113-first-problem-same-in-check-and-exe…

27c8f67

…cute

brettheap added this to the v1.2.1 milestone Nov 10, 2026

brettheap marked this pull request as ready for review November 10, 2026 14:34

Navigation Menu

Repository navigation

brettheap commented Nov 10, 2026 •

edited

Loading

Uh oh!

sourcery-ai Bot commented Nov 10, 2026

Interacting with Sourcery

Customizing Your Experience

Getting Help

Uh oh!

sonarqubecloud Bot commented Nov 10, 2026

Uh oh!

brettheap commented Nov 10, 2026

Uh oh!

brettheap commented Nov 10, 2026

Uh oh!

sourcery-ai Bot commented Nov 10, 2026

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

brettheap commented Nov 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What a person sees change

The choice under --local-remote-dir, and why

Measured, before and after

What changed, in adopt-project.py

Rows

Mutants

Gates

Merges

Independent review

Summary by Sourcery

Uh oh!

sourcery-ai Bot commented Nov 10, 2026

Reviewer's Guide

Implementation

Tests

Interacting with Sourcery

Customizing Your Experience

Getting Help

Uh oh!

sonarqubecloud Bot commented Nov 10, 2026

Quality Gate passed

Uh oh!

brettheap commented Nov 10, 2026

The two choices

Findings

Item 2: before and after (measured)

Item 3: the rows and mutants (measured)

Item 5: merges (git 2.43.0, measured)

Item 6: the body (measured)

Not measured

Paths

Uh oh!

brettheap commented Nov 10, 2026

Uh oh!

sourcery-ai Bot commented Nov 10, 2026

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

brettheap commented Nov 10, 2026 •

edited

Loading

The choice under `--local-remote-dir`, and why

What changed, in `adopt-project.py`