closed-issue audit: a close that contradicts the blocked-request record is a finding #1901
No reviewers
Labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
jaylouisw/packetforge!1901
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "fix/1117-2105"
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 #742, #743 — both are reopened by this PR and neither is answered by it.
The two closes were a parser, not a decision
#742 closed at 17:07:57 with the merge of #1873, 83 seconds after its own last comment said the issue stays open. #743 closed at 16:06:14 with the merge of #1867, whose comment above it says the same thing. There is no spec behind either:
research/open-device-requests.jsonstill lists both as blocked on a value the vendor does not publish, and nothing was built.The tracker reads a close keyword followed by an issue number anywhere in a pull request body and ignores the sentence around it. #1867's opening line disclaimed the closure; #1873 carried a heading naming the condition that would end the block. Both were executed as instructions to close. The exact phrasings are quoted with
#Nplaceholders in this body on purpose — reproducing either one beside a live number re-arms it, which is also why the reopen comments were edited after posting.What changes
A
blockedverdict.scripts/check-closed-issues.jsnow readsresearch/open-device-requests.json— this project's own record of the requests it cannot model — and reports any closed issue still in it. Checked before the catalogue: an entry can be matched by a spec and still declared blocked, and that is two contradictions, not an answer.--strictfails on it,--jsoncarries it, and a mirror that will not parse prints a warning rather than quietly turning the newest check off.The window now contains the newest close. The list was fetched in the instance's default order, by issue number, so #742 sat about 1,100 rows below the 120 the CI job reads and no practical
--limitreached it. It is now sorted by most-recently-updated, which puts a close from a minute ago on page 1.Silent findings name the file.
loadCatalog()returns spec objects without the path that produced them, so every live silent finding printed-> null— a report nobody can act on. The walk is repeated locally for that one field.Evidence, live,
--limit 60 --max-age-days 2Before: no blocked count, and every finding ended
-> null.After:
Tests and gates
test/closed-issues-audit.test.js: 31 passed, 0 failed (17 before; the new cases pin the two incident issues against the live mirror file, theblocked-before-catalogue precedence, an unparsable/misshapen mirror, the sort parameter, and the path the loader drops).node scripts/syntax-check.js: 394 files, all OK.npm run check:bytes: clean.test/request-queue.test.js23 passed,test/skip-audit.test.js15 passed — the other suites that touch this script.node scripts/run-tests.jsis running on the branch; this PR touches one standalone script and the docs, not engine code.Docs
docs/RESOLUTIONS.mddescribed three counts and now describes four, with the mechanism and the ordering reason.CONTRIBUTING.mdgains "What a pull request body closes": writeRefs #N, keep the number away from those nine words everywhere else in the body, including headings and quoted comments.bbcf30c77be90b4b0692e90b4b06925289472da65289472da63b4bc0ffea