Repository navigation
Use plain words for internal tags and names in public hub text - #176
Conversation
The apt-guard step of three release workflows carried a ticket id in its name, and the hosted authorization job of two workflows carried a decision tag, a bot-policy tag and a ticket id in its comments. This change removes those tags and changes nothing else in the workflows. The step name is pinned by an exact step dictionary in tests/test_release_workflows.py and the authorization job text is pinned byte for byte, so the test literals and the workflows change together. The step is now named "Stop automatic apt upgrades (until the release image is rebaked)" in publish.yml, release-finalize.yml and release-redeliver.yml. The three authorization comments in facts-validate.yml and python-test.yml now read "Standard GitHub-hosted label; never a larger-runner label.", "Bot logins whose same-repository pull requests count as ours: dependabot[bot]" and "Bind identity to the triggering run and its re-runner.". One comment in python-test.yml loses its ticket id, and six comments and messages in the test lose theirs. Validation, on a clean Linux arm64 host with the previous tree and the changed tree side by side: the 151 tests of tests/test_release_workflows.py pass on both. With only the changed test file on the previous tree, 13 of them fail (the tests that compare the workflows with the pinned step and authorization text); with the whole change they pass again. The five changed workflow files parse to the same documents as before once the step name is mapped, and the changed test file has the same syntax tree as before once comments and the mapped literals are set aside. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Public documentation and comments in the hub named the release owner by an internal role word, the maintainers' own machine by an internal name, and a kind of work by an internal work-track word. This change rewrites those sentences in plain words and changes nothing else. facts/README.md says "the release owner" where it used the internal role word (sixteen lines, two of them headings). README.md, tools/release-dry-run/README.md, docs/release-chain.md and the docstring of tests/test_release_redeliver.py say "a laptop" where they named the maintainers' own machine. docs/release-chain.md, facts/README.md and tools/release-dry-run/README.md use plain words (receipt roles, consumers, a separate output) where they used the internal work-track word; the sentence in tools/release-dry-run/README.md that named an internal project label now says "In this project". .github/dependabot.yml and a comment in tests/test_release_hpo_image.py lose a role name, a ticket id, a level name and a decision tag. One line of program output changes: the comment that scripts/repo_descriptions.py writes at the top of its review output no longer starts with the role word, and no test or fixture reads it. No file changes its line count, so the pinned README range (lines 215 to 244) and every other line anchor hold. Validation, on a clean Linux arm64 host with the previous tree and the changed tree side by side: the docs-runtime tests (6) and the hpo-image tests (35, including the pin on README lines 215 to 244) pass on both. The broader release run (609 tests) gives the same result on both: the same 18 failing ids and the same 9 test modules that cannot be imported on that host because a dependency is missing. The four changed Python files of the two commits have the same syntax tree as before once comments, docstrings and six mapped string pairs (one of them the program's output line) are set aside, and .github/dependabot.yml parses to the same document. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
Graphify reviewed this change.
No issues found by static checks — no coupling regressions or blocking issues in the code graph. That is not the same as safe to merge: see what was not checked.
Not checked: tests were not run; formal verification proved 0 of 1 changed function(s) (1 not verified).
Formal verification. PR-changed functions: 0/1 verified (0 proven, 0 may-equivalent, 0 distinguished) · 1 not verified (1 vacuous).
Not verified on this run: main (vacuous: never exercised).
Graphify review — findings
Scrubs internal tracker IDs and owner-specific references (LB-504, LB-437, CR-135, owner #1050, and similar) from workflow comments, step names, README.md and docs/release-chain.md, replacing "Mac" with "a laptop" and "lane" with "role" or "consumer". Workflow behaviour is unchanged. The edits to the release workflow, HPO image and redeliver tests, the facts README sections and the repo-descriptions fields script come from the truncated part of the diff, so I can't confirm they match the same rewording.
No blocking issues surfaced.
Analysis details — impact, health, verification
Impact & health
Graphify review
Impact — 610 functions depend on the 591 functions this change touches.
Health — grade A; 7 existing hotspot(s) in the area this change touches (pre-existing, not introduced here):
render_document()— 2 callers, 8 callees (medium)_toolchain_argument_problems()— 4 callers, 4 callees (medium)reconcile()— 1 callers, 10 callees (medium)prepare()— 1 callers, 7 callees (medium)prepare()— 0 callers, 12 callees (high)dispatch()— 0 callers, 10 callees (medium)publish— 0 callers, 8 callees (medium)
Verification — 610 functions in the blast radius were not formally verified this run (proofs are advisory here).
Gate & verification
graphify gate
PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.
Advisory (not blocking):
- verification_scope: 610 function(s) in the blast radius were not formally verified this run
Formal verification
Could not verify: Could not verify main.
The verifier did not have enough to check main, so it is saying so rather than guessing. No false assurance is the whole point.
Guarantee: No guarantee either way, this is an honest abstention, not a pass.
Note: Reason: not verifiable: all 68 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly NameError — names the real obstacle, not a sampling gap)
Workflow comments, two test comments, README.md and the runner policy page cited a dated decision, a chat message id and an internal rule id for the choice of GitHub-hosted runners. This change states the choice without them and changes nothing else. The same comment appeared 15 times in nine workflows (facts-descriptions, facts-fanout, handle-upstream, publish, release-finalize, release-receipts, release-reconcile, release-redeliver and release-rollback) and twice in tests/test_release_workflows.py; it now reads "pineforge-release runs on GitHub-hosted runners; standing exception for this repository". Two other test comments lose their date and message id. README.md line 200 and six lines of docs/release-chain.md say "a standing exception" where they cited the decision. No file changes its line count, so the README range pinned by the tests and every line anchor hold. Validation, on a clean Linux arm64 host with the previous tree (the head of the change before this one) and this tree side by side: the 151 tests of tests/test_release_workflows.py pass on both with the same test ids; the docs-runtime tests (6) and the hpo-image tests (35, including the pin on README lines 215 to 244) pass before and after; the broader release run (609 tests) gives the same result on both. The nine workflow files parse to the same documents as before and the test file has the same syntax tree. Nothing pins the changed lines, so there is no failing run to show first. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Summary
Public text in the release hub named internal ticket, role and decision tags in workflow step names and comments, an internal role word in the setup checklists, and the maintainers' own machine in several places. This change rewrites them in plain words. It changes no behavior of any workflow or program, except one comment line that a program writes at the top of its review output.
Two commits
tests/test_release_workflows.py). The apt-guard step name and the hosted authorization job text are pinned by test literals, so the tests and the workflows change together. This commit can be dropped without the second.README.md,docs/release-chain.md,facts/README.md,tools/release-dry-run/README.md,.github/dependabot.yml,scripts/repo_descriptions.py,tests/test_release_hpo_image.py,tests/test_release_redeliver.py). Sentences are reworded in place; the one program output line is the comment at the top of the review output ofscripts/repo_descriptions.py, which no test or fixture reads.No file changes its line count, so the pinned README range (lines 215 to 244) and every other line anchor hold.
Validation
On a clean Linux arm64 host, with the previous tree and the changed tree side by side:
tests/test_release_workflows.py: the 151 tests pass on both. With only the changed test file on the previous tree, 13 tests fail (the ones that compare the workflows with the pinned step and authorization text); with the whole change they pass again.test_release_*run (609 tests) gives the same result before and after: the same 18 failing ids and the same 9 test modules that cannot be imported on that host because a dependency is missing.Not claimed: the test modules that cannot be imported on that host, and the repository-description tests, which fail the same way before and after, do not evidence this change. No amd64 run. The hosted CI run follows the push.
Follow-up commit: dated decision references
Summary
Public text in the release hub still cited a dated decision, a chat message id and an internal rule id for the choice of GitHub-hosted runners: in workflow comments, two test comments,
README.mdanddocs/release-chain.md. This change states the choice without them. It changes no behavior of any workflow, test or program.Changes
pineforge-release runs on GitHub-hosted runners; standing exception for this repository. It appears 15 times in nine workflows (facts-descriptions,facts-fanout,handle-upstream,publish,release-finalize,release-receipts,release-reconcile,release-redeliver,release-rollback) and twice intests/test_release_workflows.py.README.mdline 200 and six lines ofdocs/release-chain.mdsay "a standing exception" where they cited the decision.Release runner isolationand its anchor are unchanged.tests/test_release_workflows.py, line 3076). It carries no date, id or rule name.Validation
On a clean Linux arm64 host, with the previous tree (the head of the preceding public-text change) and this tree side by side:
tests/test_release_workflows.py: the 151 tests pass on both, with the same test ids.test_release_*run (609 tests) gives the same result before and after: the same 18 failing ids and the same 9 test modules that cannot be imported on that host because a dependency is missing.tests/test_release_workflows.pyhas the same syntax tree.Nothing pins the changed lines (workflow comments are not part of the parsed documents, and no test asserts the README or docs sentences), so there is no failing run to show first.
Not claimed: no amd64 run, and the test modules that cannot be imported on that host do not evidence this change. The hosted CI run follows the push.
🤖 Generated with Claude Code