[brettheap]

The two Python 3.9 jobs and the tests job now run on ubuntu-24.04: neither 3.9 job fails in setup-python on an Ubuntu 26.04 runner, and the full suite does not run out of disk there (round 3, below).

The problem. ubuntu-latest has started giving out Ubuntu 26.04 runners, and setup-python has no Python 3.9 build for them: its manifest lists 3.10 to 3.15 there, and 3.8 to 3.15 for 24.04. Two jobs asked for 3.9 on ubuntu-latest:

• the package job of tests.yml, which failed on shape-doctor.py: a leg's gitlink that no .gitmodules section registers, beside one whose sections register other paths or beside none, is named with its own way out, as #914 names one beside a .gitmodules with no section (#1080) #1092's run at 04:35:25Z with The version '3.9' with architecture 'x64' was not found for Ubuntu 26.04.;
• the release's install-check row {os: ubuntu-latest, python: '3.9'}, which would fail the same way on the v1.2.0 release.

What changed.

• .github/workflows/tests.yml, the package job (:289): runs-on: ubuntu-latest is now runs-on: ubuntu-24.04, with a comment above it (:280-287) saying why: 3.9 is requires-python's floor, and the Ubuntu 26.04 that ubuntu-latest now gives out has no 3.9 build for setup-python.
• .github/workflows/tests.yml, the tests job (:74): runs-on: ubuntu-latest is now runs-on: ubuntu-24.04, with a comment above it (:65-72) saying why (round 3, below).
• .github/workflows/release.yml, the install matrix's last row (:908): os: ubuntu-latest is now os: ubuntu-24.04, python: '3.9' as before, with a comment (:903-907) saying why.
• The other jobs keep their runners: tests-git-bash runs on windows-latest, and in release.yml build asks for 3.12 on ubuntu-latest, which 26.04 has. The package comment says so, and that tests is on ubuntu-24.04 too, for its own reason.
• tests/test_release_workflow.py: the install-matrix test (:562-569) pins the new row and that the ubuntu-latest rows ask for 3.x alone; the package test (:584-596) pins the job to ubuntu-24.04 and its one 3.9; a new test (:599-608) pins the tests job to ubuntu-24.04; and a new test (:640-658, with its helper setup_python_pairs, :611-637) holds every setup-python step of both workflows to what 26.04 has.
• tests/test_doctor_legs_members_next.py (:37): a docstring said tests.yml runs the suite on ubuntu-latest alone; it says ubuntu-24.04.

The new test, test_no_setup_python_on_ubuntu_latest_asks_for_a_python_26_04_lacks, is an allow list over tests.yml and release.yml:

• for every actions/setup-python@ step, each ${{ matrix.<key> }} a strategy.matrix.include row has is filled from that row's key (a job with no matrix is one row), and the values are compared as str(), so an unquoted python-version: 3.9, which the repository's YAML reader makes the float 3.9, is still "3.9";
• a job whose matrix has any key but include, the standard form (os: [...], python: [...]), fails the helper's assertion, because the test does not expand it, so it cannot pass unchecked;
• every pair whose runner is ubuntu-latest must ask for 3.x or 3.12;
• every 3.9 pair must run on ubuntu-24.04;
• no ${{ may be left unfilled, so a reference an include row does not fill fails the test rather than slipping past it;
• the two 3.9 jobs (package and the release's install row) must be among the pairs, so the loop cannot pass over nothing.

What reads the row's name. In release.yml, matrix.os is read twice, by the job name (install check (${{ matrix.os }}, Python ${{ matrix.python }}), :880) and by runs-on (:889), and runner.os once, by the Windows-only PowerShell step (:1238). No if:, artifact name, cache key or other step branches on the row. So the new row runs exactly as the old one did, on the 24.04 image ubuntu-latest used to be, and only its job name changes: install check (ubuntu-24.04, Python 3.9). No test, doc or script names that job.

The search. tests/, README.md, docs/ and AGENTS.md name no runner of the two 3.9 jobs, and the one docstring that named the runner of tests, tests/test_doctor_legs_members_next.py:37, is changed (round 3); the only ubuntu-latest mentions in docs/ are the v1.0.0 and v1.1.0 release notes, which record what those releases ran and stay as they are.

Out of scope. requires-python stays >=3.9. Dropping 3.9, past its end of life since October 2025, is Brett Heap's call.

Tests, each file alone under a stub gh, with EMAIL set, at head 99b097fe3c1e unless marked:

• tests/test_release_workflow.py: 59 passed
• tests/test_workflow_concurrency.py: 14 passed
• tests/test_install_sh.py: 399 passed
• tests/test_repo_hygiene.py: 100 passed
• tests/test_cli_reference.py: 23 passed
• pyflakes tests/test_release_workflow.py: no output, as at f464631aa89f; tests/test_doctor_legs_members_next.py (a docstring changed): the same 207 warnings as at f464631aa89f, and its 192 tests collect.
• Not run again at this head (earlier results): tests/test_release_install_check_children.py 30 passed and tests/test_no_auto_maintenance.py 60 passed at 3df09c6f5d81; tests/test_shape_advisory.py 47 passed and tests/test_openreposhape_package.py 213 passed at 6031bf1b. Round 3 changed tests.yml's runs-on and comments, one new test and one docstring, and none of these reads them.

The pairs. setup_python_pairs() yields eight, sorted, before and after round 3, and diff shows one line changed, the tests pair: ('tests.yml', 'tests', 'ubuntu-latest', '3.x') is now ('tests.yml', 'tests', 'ubuntu-24.04', '3.x').

Probes and mutants, run at 99b097fe3c1e, each restored and checked with cmp:

• mutant 3 (round 3), the tests job back on ubuntu-latest: test_the_tests_job_runs_the_suite_on_ubuntu_24_04 fails (1 failed, 58 passed); the allow list passes, as tests asks for 3.x;
• probe A, python-version: 3.9 unquoted on the tests job: with the job on ubuntu-24.04 the file passes (59 passed), correctly, since a 3.9 job belongs on 24.04; with the job also put back on ubuntu-latest the allow list fails on ('tests.yml', 'tests', 'ubuntu-latest', '3.9') and so does the pin test, test_the_tests_job_runs_the_suite_on_ubuntu_24_04 (2 failed, 57 passed), and with the pin test deselected the allow list fails alone (1 failed, 57 passed, 1 deselected), so the unquoted float is still caught;
• probe B, release.yml's build set to '3.9': the allow list fails on ('release.yml', 'build', 'ubuntu-latest', '3.9') (1 failed, 58 passed);
• mutant 1, the release row back on ubuntu-latest: the allow list fails on ('release.yml', 'install', 'ubuntu-latest', '3.9'), and test_the_install_check_runs_on_three_systems_and_python_3_9 fails too (2 failed, 57 passed);
• mutant 2, package back on ubuntu-latest: the allow list fails on ('tests.yml', 'package', 'ubuntu-latest', '3.9'), and test_the_package_job_runs_its_python_3_9_on_ubuntu_24_04 fails too (2 failed, 57 passed);
• probe C, the tests job on a matrix of the standard form (runs-on: ${{ matrix.os }}, matrix: {os: [ubuntu-latest], python: ['3.9']}, python-version: ${{ matrix.python }}): the helper's assertion fails with ('tests.yml', 'tests', ['os', 'python']), and so does the pin test, since ${{ matrix.os }} is not ubuntu-24.04; with the pin test deselected the helper's assertion fails alone (1 failed, 57 passed, 1 deselected);
• probe D, the same job on an include row that has os and no python: the ${{ check fails on ('tests.yml', 'tests', 'ubuntu-latest', '${{ matrix.python }}'); probe E, the row {os: ubuntu-latest, python: '3.9'}: the allow list fails on ('tests.yml', 'tests', 'ubuntu-latest', '3.9'), so the row is filled (each 2 failed, 57 passed: the pin test, and the allow list).

Merges. git merge-tree --write-tree of this head (99b097fe3c1e) with origin/main (de75bf02d3f8) is clean (rc 0), and its tree (034caec526d6) is the head's own. With #1092 (l3-1080-unregistered-gitlinks, 83627b229737) it is clean (rc 0), tree 57cb15c3b3cf; #1092 touches shape-doctor.py, tests/test_doctor_plan_overrides.py and tests/test_shape_doctor.py only. No other open PR touches .github/, tests/test_release_workflow.py or tests/test_doctor_legs_members_next.py.

The proof is this PR's own CI run: its tests and package jobs are the two this change moves onto ubuntu-24.04; nothing local runs Actions.

Review round (RV1095, issuecomment-6308584611, READY with no MUST):

• S1: the 3.9 guard looked for the string "3.9" in tests.yml alone, so an unquoted 3.9, release.yml's build on 3.9 and a runner or Python that a matrix gave all passed; it is now the allow list above over both workflows, which reads a matrix's include rows (a matrix of any other form was still let through until S3, below), and probes A and B and both mutants failed it (round 3 changes probe A, below).
• S2: the package comment (and this body, and the first commit's message) said "Every other job asks for 3.x or 3.12 ... and stays on ubuntu-latest", which was not true; it now says the sentence under "What changed", and the squash message should use that one. release.yml's comment (:903-907) is true as written and is unchanged: 26.04 lists 3.10 to 3.15 and 24.04 lists 3.8 to 3.15 (read from the manifest again at 05:13Z), and the three rows above ask for 3.x.
• N1: runner.os is at release.yml:1238 at this head, not :1234.
• N2: of the release notes, only v1.0.0 and v1.1.0 mention ubuntu-latest, not v1.1.1.

Review round 2 (RV1095d, issuecomment-6308733787, READY with no MUST, one SHOULD and one NIT):

• S3: the helper expanded include rows only and filled a key a row lacked with the text "None", so the ${{ check never fired on a standard matrix (probe C survived); it now asserts a job's matrix has no key but include (tests/test_release_workflow.py:623-624 at this head) and fills only the keys a row has (:633-636), and probe C fails the file.
• N3: the package test's docstring said the next test holds "every other job" to what 26.04 has; it now says every job of both workflows on ubuntu-latest (:589-590), and the allow-list test's docstring says what it holds for both matrix forms (:641-647 at this head).

Round 3: the tests job (#1096, filed by lane 3):

• The failure: shape-doctor.py: a leg's gitlink that no .gitmodules section registers, beside one whose sections register other paths or beside none, is named with its own way out, as #914 names one beside a .gitmodules with no section (#1080) #1092's tests job ran on an Ubuntu 26.04 runner (job 119050702775, image ubuntu-26.04 version 20261012.163.1) and ended 2436 failed, 11382 passed, 380 skipped, 5579 errors in 3048.23s. Its log has 28325 lines of Disk quota exceeded (9287 of them [Errno 122], the rest git's own copy errors), and later No space left on device, all under /tmp/pytest-of-runner/pytest-0, pytest's basetemp. It is the same image move as this PR's, and the job passed on ubuntu-24.04.
• The change: .github/workflows/tests.yml:74, runs-on: ubuntu-latest is now runs-on: ubuntu-24.04, with a comment (:65-72) that gives the job, its counts and the Disk quota exceeded under pytest's basetemp in /tmp, and says it is pinned until the suite runs on 26.04 (tests.yml: on an Ubuntu 26.04 runner the full suite fills /tmp under pytest's basetemp (Disk quota exceeded, 2436 failed and 5579 errors), so tests runs on ubuntu-24.04 until the suite can run there #1096).
• A sentence this makes false: the package comment (:280-287) said "tests here asks for 3.x ... on ubuntu-latest". It now says that tests is on ubuntu-24.04 too, for a reason of its own, that tests-git-bash runs on windows-latest, and that in release.yml build asks for 3.12 on ubuntu-latest, which 26.04 has. The 3.9 reason is as it was. release.yml's comment (:903-907) names no runner of tests, so it is unchanged, and no other workflow comment does.
• The test: test_the_tests_job_runs_the_suite_on_ubuntu_24_04 (tests/test_release_workflow.py:599-608), whose docstring names tests.yml: on an Ubuntu 26.04 runner the full suite fills /tmp under pytest's basetemp (Disk quota exceeded, 2436 failed and 5579 errors), so tests runs on ubuntu-24.04 until the suite can run there #1096 and the disk quota. setup_python_pairs() and the allow-list test are as they were, and pass.
• The squash message should name both reasons, as the title does.

Follow-up: #1096 (run the suite on Ubuntu 26.04)

Closes #1094

Lane: openreposhape-3 (openRepoShape-3)

🤖 Generated with Claude Code

Summary by Sourcery

Keep Python 3.9 workflow jobs on Ubuntu 24.04 while validating that all setup-python jobs use versions supported by their runners.

Bug Fixes:

• Run Python 3.9 packaging and release installation checks on Ubuntu 24.04, avoiding setup-python failures on Ubuntu 26.04 runners.

Enhancements:

• Add workflow validation covering all setup-python runner and Python-version combinations, including matrix-expanded jobs.

Tests:

• Update workflow tests to enforce the Ubuntu 24.04 Python 3.9 placement and prevent unsupported Python versions from running on ubuntu-latest.

[sourcery-ai]

Reviewer's Guide

This PR updates both Python 3.9 workflow paths to run on ubuntu-24.04 because setup-python does not provide a 3.9 build for the Ubuntu 26.04 image now supplied by ubuntu-latest. It preserves ubuntu-latest for jobs requesting 3.x or 3.12 and adds regression tests covering the release matrix and all tests.yml Python 3.9 jobs.

Flow diagram for workflow runner selection

flowchart TD
    Start[Workflow requests Python version]
    VersionCheck{Python 3.9?}
    PinRunner[Use ubuntu-24.04]
    LatestRunner[Use ubuntu-latest]
    SetupPython[setup-python installs Python]

    Start --> VersionCheck
    VersionCheck -->|yes| PinRunner
    VersionCheck -->|no: 3.x or 3.12| LatestRunner
    PinRunner --> SetupPython
    LatestRunner --> SetupPython

Loading

File-Level Changes

Change	Details	Files
Pin Python 3.9 execution to Ubuntu 24.04 wherever workflows use it, avoiding setup-python failures on Ubuntu 26.04 while retaining latest runners for compatible Python versions.	Changed the tests package job runner to ubuntu-24.04 and documented the setup-python compatibility rationale. Changed the release install-check matrix’s Python 3.9 Linux row to ubuntu-24.04 while leaving 3.x rows on ubuntu-latest. Added workflow tests that enforce the runner/version pairing and prevent Python 3.9 from being requested on ubuntu-latest.	.github/workflows/tests.yml .github/workflows/release.yml tests/test_release_workflow.py

Assessment against linked issues

Issue	Objective	Addressed	Explanation
#1094	Run the tests.yml package job, which uses Python 3.9, on Ubuntu 24.04 instead of ubuntu-latest, and document why.	✅
#1094	Run the release workflow’s Python 3.9 install-check row on Ubuntu 24.04 instead of ubuntu-latest, and document why.	✅
#1094	Update workflow tests to enforce the Ubuntu 24.04 runner for these Python 3.9 jobs while leaving unrelated jobs on ubuntu-latest.	✅

Possibly linked issues

• tests.yml and release.yml: the Python 3.9 jobs run on ubuntu-latest, which now gives out Ubuntu 26.04 runners with no 3.9 for setup-python, so package fails there and the release's 3.9 install check would too #1094: The PR implements the issue’s fix: pins the package job and release’s Python 3.9 install row to ubuntu-24.04 and updates tests.

---

Tips and commands

Interacting with Sourcery

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

Customizing Your Experience

Access your dashboard to:

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

Getting Help

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

[brettheap]

(lane openRepoShape-3, 2026-10-21T05:11:37Z) Independent review (RV1095) of head 6031bf1: READY, no MUST. Two SHOULDs for one fix round before this leaves draft; landing waits on tests anyway.

Checked and true:

• Only two runners move: tests.yml package and release.yml's 3.9 install row, both to ubuntu-24.04. Every other job and row keeps its runner and its Python (parsed on main and head, matrix rows expanded).
• setup-python's manifest (read 04:39Z): Linux x64 26.04 has 3.10-3.15, 24.04 has 3.8-3.15. 3.9 on 24.04 x64 is 3.9.25. The jobs left on ubuntu-latest ask for 3.x (resolves to 3.15.0) or 3.12 (3.12.15), or for no Python at all (testpypi, pypi).
• The release row is read only by its job name, runs-on, setup-python and the Windows-only runner.os step (release.yml:1238). Only the check name changes, to install check (ubuntu-24.04, Python 3.9). Nothing names it: no test, doc, script or ruleset. The release gate requires tests alone, and package keeps its name.
• The error text in the comment matches failed job 119040615170 (image ubuntu-26.04) exactly.
• Tests at the head, one file at a time under a stub gh: test_release_workflow 57 passed, test_release_install_check_children 30, test_workflow_concurrency 14, test_repo_hygiene 100.
• Both mutants KILLED (release row back on ubuntu-latest; package back on ubuntu-latest), files restored and checked with cmp.
• CI: package passed on ubuntu-24.04 (24.04.5, image 20261012.334.1) with CPython 3.9.25. tests-git-bash and SonarCloud passed. tests was still running at 05:09Z, and its actionlint steps had not started.
• merge-tree with origin/main de75bf0 is clean (tree 659a288b8747). No other open PR touches .github/ or tests/test_release_workflow.py.
• Body: closingIssuesReferences [1094]; it ends Closes, Lane, robot before Sourcery's usual summary; the only non-ASCII character is the robot.

SHOULD 1, tests/test_release_workflow.py:596-600. The guard looks for the string "3.9" in tests.yml only, so it passes with a 3.9 job on ubuntu-latest:

• an unquoted python-version: 3.9 on the tests job: the repo's YAML reader makes it the float 3.9; 57 passed;
• release.yml's build set to '3.9': 57 passed;
• a job whose runner comes from a matrix is never checked.
  Fix: over both workflows, expand ${{ matrix.os }} and ${{ matrix.python }} from each matrix row, and for every setup-python on ubuntu-latest assert str(python-version) is in {"3.x", "3.12"}. That catches both probes and both mutants.

SHOULD 2, .github/workflows/tests.yml:277-278. "Every other job asks for 3.x or 3.12 ... and stays on ubuntu-latest" is not true as written: tests-git-bash runs on windows-latest, no tests.yml job asks for 3.12 (release.yml's build does), and testpypi and pypi ask for nothing. Suggested: "The other jobs keep their runners: tests here asks for 3.x and release.yml's build for 3.12 on ubuntu-latest, both of which 26.04 has, and tests-git-bash runs on windows-latest." The body and the squash message repeat the sentence; fix those too.

NIT (body only): runner.os is at release.yml:1238 at the head, not :1234. Of the release notes, only v1.0.0 and v1.1.0 mention ubuntu-latest, not v1.1.1.

Lane 3's plan: a short fix round, F1095, takes both SHOULDs and the two body NITs on this branch. Then a short delta review, and the pilot. #1092's pilot is chained behind this one, so package on main's push run stays off the Ubuntu 26.04 runners. If main goes red on package first, this lands as the fix for red main (condition 5 of the fast-refresh ruling).

Lane: openreposhape-3 (openRepoShape-3)

[brettheap]

(lane openRepoShape-3, 2026-10-21T05:23:20Z) Delta review (RV1095d) of head 3df09c6: READY, no MUST. One new SHOULD.

Checked:

• The diff 6031bf1..3df09c6 touches only the tests.yml comment (:277-279) and tests/test_release_workflow.py. release.yml and every runner are unchanged. The new comment is true.
• tests/test_release_workflow.py at the head: 58 passed.
• KILLED: probe A (unquoted 3.9 on tests), probe B (build on '3.9'), mutant 1 (release row back on ubuntu-latest) and mutant 2 (package back on ubuntu-latest). Each file was restored and checked with cmp.
• The 3.9-on-24.04 rule and the ${{ rule forbid nothing plausible: the manifest's 3.9.25 has no Windows build and no arm64 macOS build, and an inputs-given value should fail closed.
• Body: closingIssuesReferences [1094]; it ends Closes, Lane, robot before Sourcery's summary; N1 and N2 are fixed; the only non-ASCII character is the robot.
• merge-tree is clean with origin/main de75bf0 (tree 6065d23300c5, the head's own) and with shape-doctor.py: a leg's gitlink that no .gitmodules section registers, beside one whose sections register other paths or beside none, is named with its own way out, as #914 names one beside a .gitmodules with no section (#1080) #1092 83627b2 (tree 2a97a9537825).
• CI: package passed on ubuntu-24.04 with CPython 3.9.25 (job 119052545007). tests-git-bash and SonarCloud passed. tests was in progress at 05:22Z, its actionlint steps pending.

SHOULD, tests/test_release_workflow.py:607 and :613-614. The helper expands matrix include rows only, and fills a key a row lacks with the text "None", so the ${{ rule never fires on a matrix without include. Probe C survived (58 passed): the tests job with runs-on ${{ matrix.os }}, matrix {os: [ubuntu-latest], python: ['3.9']}, and python-version ${{ matrix.python }}. Fix:

• assert set(matrix) <= {"include"} for every job;
• replace only the ${{ matrix.<key> }} the row has (for key, value in row.items()).
  I checked this in memory: the same 8 pairs at the head, and probe C is caught.

NIT, tests/test_release_workflow.py:589: "the next test holds every other job of both workflows to what 26.04 has" should say "every job of both workflows on ubuntu-latest", since that is what the next test checks.

Lane 3's plan: one more short round, F1095b, takes S3 and the NIT before the landing. The test's docstring says it covers a runner or a Python that a matrix gives, so the guard must also fail closed on a matrix without include. Probe C must then fail. The same reviewer checks it, then the pilot (key 1095b) runs. Each tests run here takes about an hour, which is the cost of the extra round. #1092's pilot stays chained behind this landing.

Lane: openreposhape-3 (openRepoShape-3)

[brettheap]

(lane openRepoShape-3, 2026-10-21T05:29:01Z) Delta review 2 (RV1095e) of head f464631: READY, no MUST, no SHOULD.

Checked:

• The diff 3df09c6..f464631 touches tests/test_release_workflow.py only (+20/-14).
• S3 is fixed: the helper asserts set(matrix) <= {"include"} (:611-612) and fills only the keys a row has (:621-624).
• N3 is fixed (:589-590). The docstrings at :600-606 and :629-635 are true.
• tests/test_release_workflow.py at the head: 58 passed.
• Probe C (the tests job on a standard matrix, os [ubuntu-latest] x python ['3.9']) is now KILLED: 1 failed, 57 passed, at :611 with ('tests.yml', 'tests', ['os', 'python']). The file was restored and checked with cmp.
• Body: closingIssuesReferences [1094]; it ends Closes, Lane, robot before Sourcery's summary; the only non-ASCII character is the robot; the Review round 2 lines are true.
• merge-tree with origin/main is clean (tree 1b274e7b630c, the head's own). Main has not moved from de75bf0.

Optional NIT, no round needed: the matrix assertion at :611 also fires for a job with a standard matrix and no setup-python step. None exists today. Moving it under if steps: would narrow it, in a follow-up if at all.

Lane 3's plan: this head lands as it is, and the optional NIT is left. The pilot (key 1095b, issue arg 1094) has its go now. It waits for tests on f464631, gates the predicted squash over main, then takes the HOLD in lane 3's rule that the first PR green and gated goes first. #1092's pilot starts once this one has LANDED.

Lane: openreposhape-3 (openRepoShape-3)

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

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

See analysis details on SonarQube Cloud

[brettheap]

(lane openRepoShape-3, 2026-10-21T06:20:01Z) Delta review 3 (RV1095f) of head 99b097f: READY, no MUST, no SHOULD.

Checked:

• The diff f464631..99b097f changes only tests.yml (comment :65-72, tests on ubuntu-24.04 at :74, package comment :285-287), the new test at tests/test_release_workflow.py:599-608, and the docstring at tests/test_doctor_legs_members_next.py:37. The pin is right: main's passing tests job 118985229247 ran on image ubuntu-24.04 20261012.334.1.
• Job 119050702775 matches the comment: image ubuntu-26.04 20261012.163.1; "2436 failed, 11382 passed, 380 skipped, 5579 errors in 3048.23s"; the first Errno 122 at 06:01:28Z and the first real ENOSPC at 06:01:42Z; every write the errors name is under /tmp/pytest-of-runner/pytest-0.
• The rewritten package comment and the doctor docstring are true.
• Probe A: allowing 3.9 on 24.04 is right. The guard is about whether setup-python has a build for the runner. Pinning tests to 3.x would be a different guard. Not worth a round.
• tests/test_release_workflow.py at the head: 59 passed. Mutant 3 (tests back on ubuntu-latest) is KILLED: 1 failed, 58 passed. The file was restored and checked with cmp.
• Title: 221 characters, ASCII. Body: closingIssuesReferences [1094]; it ends Closes, Lane, robot; tests.yml: on an Ubuntu 26.04 runner the full suite fills /tmp under pytest's basetemp (Disk quota exceeded, 2436 failed and 5579 errors), so tests runs on ubuntu-24.04 until the suite can run there #1096 appears only as the tracking issue and the follow-up; the only non-ASCII character is the robot.
• CI run 39553859661: package success on ubuntu-24.04; tests carries the ubuntu-24.04 label and was in progress at 06:18Z; tests-git-bash and SonarCloud passed.
• merge-tree is clean with origin/main de75bf0 (tree 034caec526d6, the head's own) and with shape-doctor.py: a leg's gitlink that no .gitmodules section registers, beside one whose sections register other paths or beside none, is named with its own way out, as #914 names one beside a .gitmodules with no section (#1080) #1092 83627b2 (tree 57cb15c3b3cf).

NIT, .github/workflows/tests.yml:69 and the body's "The failure" line: 28325 is the number of lines saying "Disk quota exceeded". Only 9287 of them carry "[Errno 122]"; the rest are git's "fatal: cannot copy ...: Disk quota exceeded" and "error: copy-fd: write returned: Disk quota exceeded". Say "28325 lines of Disk quota exceeded" if there is another round, and in the squash message either way.

Lane 3's plan:

• The NIT is fixed in the body ("28325 lines of Disk quota exceeded, 9287 of them [Errno 122]") and in lane 3's squash message. The tests.yml comment keeps its words until tests.yml: on an Ubuntu 26.04 runner the full suite fills /tmp under pytest's basetemp (Disk quota exceeded, 2436 failed and 5579 errors), so tests runs on ubuntu-24.04 until the suite can run there #1096 drops this pin; another round would cost another hour of tests.
• The pilot (key 1095c, issue arg 1094) has its go now. It waits for tests on 99b097f, which runs on ubuntu-24.04, then gates the predicted squash over main and takes the HOLD.
• After it lands, shape-doctor.py: a leg's gitlink that no .gitmodules section registers, beside one whose sections register other paths or beside none, is named with its own way out, as #914 names one beside a .gitmodules with no section (#1080) #1092 gets gh pr update-branch, so its tests runs on the pinned workflow, and then its pilot runs.

Lane: openreposhape-3 (openRepoShape-3)

[sourcery-ai]

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

You can request another review in 10 hours and 43 minutes by commenting @sourcery-ai review. Upgrade to get a review now.