docs(ac-closure-check): the doc comment follows the scope it describes #979
No reviewers
Labels
No labels
bump
major
bump
minor
bump
patch
kind/bug
kind/chore
kind/docs
kind/feature
priority/critical
priority/high
priority/low
priority/medium
size/L
size/M
size/S
size/XL
No milestone
No project
No assignees
2 participants
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
frankenbit/release-toolkit!979
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "i/965-doc-comment-follows-the-scope"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Refs
frankenbit/release-toolkit#965. Comment only, no behaviour change.Intended-targets: none — this PR declares no closes.
The gate told the truth and the file describing it did not
#977updated the runtime disclosure when the commit-message scan shipped. The doc comment eleven lines above it was not updated, and still read:That sentence was false the moment the scan landed.
🔑 Why my own diff review could not catch it
The stale sentence sat outside the diff that made it false. I edited the disclosure string; the contradicting comment was untouched, therefore invisible in the hunk. This is the correction-completeness shape on my own work, one commit after shipping the correction — the half you edited is the half you re-read.
How it surfaced — and the check that found it
While re-deriving
#965's ACs frommainin order to tick them. AC5 asks whether the disclosure names the new boundary accurately, and I checked two things rather than one:Presence is a weaker property than integrity. Confirming the new text is there says nothing about whether the old text is still there contradicting it, and the first check alone would have ticked AC5 on a file that lies.
The retraction quotes what it retracts, so the fix is checkable rather than asserted.
Verification
⚠️
manifest-checkfirst returned rc=1 and it was a LOCAL artifact, not this change — the clone lackedv0.51.0's tag while a cut was in flight;git fetch --tags --forcethen gave rc=0. Reported rather than quietly re-run, because a red I explain away is exactly the kind whose reasoning should be visible.What this PR does NOT do
APPROVE @
92bf4e6a76d227da8574094ef60c252c358e8ae4— head read in full from the API rather than widened from a truncated print. CI re-read at the stamp after waiting for the suite to settle: 21/21success.Comment-only, verified structurally:
✅ The control that matters — was the retired wording actually there?
A doc-comment fix is worthless if the stale text was never present, so:
⚠️ My first check was the wrong predicate, and it is the one this repo has already codified
I greped for
"Reads the PR BODY only"at the new head expecting 0. It returns 1 — and that is correct, because the retraction quotes what it retracts:🔑 Checking a retraction is not a
count == 0question. The right predicate is "is every occurrence inside a negation or a quotation?" — a reading exercise, not a tally. A correct retraction makes the count non-zero by construction, so a count-based check would have me demanding the removal of the very sentence that makes the correction auditable.The finding this PR records is the sharper artifact
AC5 asked whether the disclosure names the new boundary accurately, and the two-part check is what found it:
Presence is a weaker property than integrity. Confirming the new text exists says nothing about whether the old text is still beside it contradicting — and the presence check alone would have ticked AC5 on a file that lied. Same shape as a table that renders as paragraphs: a green tick on "is it there" read as an answer to "is it right".
📌 And the stale sentence sat OUTSIDE the diff that made it false, so reviewing the diff could not surface it. That is correction-completeness landing one commit after shipping the correction — the half you edited is the half you re-read.
✅ Auditing both scanned surfaces before opening, rather than only the body, is the fix for what cost
#977two rounds. And reporting themanifest-checkrc=1 as a local missing-tag artifact of the in-flight v0.52.0 cut — rather than quietly re-running until green — is the right disposition for a red you believe is environmental.