feat: meet doorlooptijd, ratchetbeweging en of de documentatie klopt #558
No reviewers
Labels
No labels
accepted
bug
declined
docs
duplicate
enhancement
good first issue
in-progress
needs-info
privacy
security
triage
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
LibreKAT/Ocideck!558
Loading…
Reference in a new issue
No description provided.
Delete branch "feat/meetinstrumenten"
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?
Summary
The three measurement issues that came out of the CRA decision (#519). All three ask whether the work is good enough, not whether the code runs.
Closes #537
Closes #538
Closes #539
#537 — how long ordinary issues stay open
tool/check_issue_turnaround.dart(make doorlooptijd). Age per open issue and the oldest, time to first response (median and longest), how many are open with no response at all — the number that matters most — andtriageissues that never moved. Security reports and PRs are excluded; they have their own norms.Advisory is the default here, not a flag — the opposite of
check_service_norms, because no deadline has been agreed for ordinary issues.--strictis ready for the day one is.Against the real tracker: 8 open, 14 closed, exit 0.
#538 — are the ratchets actually moving
tool/check_ratchet_trend.dart(make ratchets). Thirteen baselines against a reference point N days back, read from the source text viagit show, plus the longest-standing baseline entries and coverage per directory. Standing still is not an error; zero does not count as standing still.It pays off immediately: 86.2% on average hides
lib/widgetsat 81.8%.#539 — documentation that is true
test/docs_claims_match_code_test.dart, mechanical half only. Everycurrently **N**in CHECKS.md must name its constant and match it; every percentage next to "floor" must be one of the two real floors; every language count about the interface must equalAppLocalizations.languageNames.length.Two real errors found by it:
ARCHITECTURE.mdclaimed 31 translation files (there are 32), and the "Latest result" table quoted a budget of 21 that was removed on 22-07.The rebase caught a genuine interaction
#518 landed a fourteenth baseline (
classSizeBaseline) while this branch was in flight, and the new measurement's own test — "every baseline intool/is actually followed" — went red. That is the test doing precisely its job on day one.While fixing it: the test typed the baseline count in as a
9, which made it fail on work that had nothing to do with it. That is the same defectdocs_claims_match_code_testexists to prevent, one layer up. Now derived from the ratchet list itself.Verification
Each report tested in both directions with fixed fixtures — quiet tracker versus planted deviation, everything shrinking versus baselines standing still, and four planted documentation lies (all four caught). 34 + 27 tests. No network in any test; the forge token comes from the keychain and appears in no code or fixture.
make checkre-run by me after the rebase: exit 0.Left deliberately
The monthly wiki overview from the #538 discussion (needs pushing), mutation testing per report (
make mutatestays manual), the human half of #539, and no gate over the scale table in CHECKS.md — it is a dated snapshot, and a gate over it would go red on every new test file. All noted where they belong rather than silently dropped.`make doorlooptijd` naast `make servicenormen`, en om dezelfde reden in geen enkel verzameldoel: het heeft een persoonlijke leessleutel nodig, en exit 2 ("kon niet meten") zou bij een medewerker zonder sleutel als een defect in zijn wijziging lezen. Beide doelen ontbraken in .PHONY. Dat is nu geen probleem omdat er geen bestand zo heet, en precies daarom ziet niemand het tot het misgaat. De CRA-positie noemde deze meting als eerste van drie richtingen; daar staat nu bij wat er is uitgevoerd, zodat het besluit en de uitvoering niet uit elkaar lopen. Bewust in assurance/ en niet in docs/: alles in docs/ reist als asset met de app mee, en een cijfer over reactietijd wordt daar een toezegging aan derden. Closes #537 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>