chore: refresh assets/badge-tests.svg from a measured count (post-#294) - #296
chore: refresh assets/badge-tests.svg from a measured count (post-#294)#296cdeust wants to merge 1 commit into
Conversation
#294 (issue #293) made the tests badge the sole artifact carrying an absolute test count, checked as a monotone floor (committed <= live) so two branches adding tests never conflict on it. The version that landed in 05c7169 kept the badge at 6673 — main's pre-rebase figure carried through its own conflict resolution — which still satisfies the floor (6673 <= 6692) but is not itself a fresh measurement, only a number that happens not to overclaim. A floor that is merely non-overclaiming, left unrefreshed indefinitely, drifts further from the true count every time another PR adds tests without also regenerating this file — the badge would still say 6673 after ten more test-adding PRs, none of them individually breaking the gate. Refreshing it whenever it is touched keeps it a real answer to "how many tests," not just a number nothing yet overtook. Regenerated via `python scripts/generate_repo_badges.py --test-count 6692` (not hand-edited); 6692 is `pytest --collect-only -q` on this branch's tip (3f0522b, current main). No other file changed — the five prose/JSON docs already point at this badge rather than restating a number. Verified: `check_doc_claims.py --test-count 6692` -> "doc claims OK"; `generate_repo_badges.py --check --test-count 6692` -> "badges OK (5 checked)"; tests_py/scripts/test_generate_repo_badges.py -> 49 passed, 20 subtests passed; ruff check/format clean.
|
Superseded — no longer needed. Main moved twice more while this was in flight: PR #284 landed right after #294 (main is extremely active right now), and whoever rebased #284 onto main also regenerated This PR's own regeneration (to 6692) was correct against the main tip it was branched from (3f0522b) but would now be a regression (6692 < 6706, reintroducing staleness) against the current tip. Closing rather than rebasing a now-pointless diff. |
1 participant
Summary
Tiny, single-file follow-up to #294 (issue #293). #294's own squash-merge
(05c7169) landed
assets/badge-tests.svgstill stating 6673 — main'scount from before that PR's own rebase — rather than a fresh measurement.
6673 <= 6692 (the true live count on this tree), so the monotone-floor
gate #294 introduced correctly stayed green; the number just isn't
current.
Not a bug in #294's design (a floor is explicitly allowed to lag), but a
floor that is only ever checked, never refreshed, gets stale forever —
worth a one-line regeneration whenever it's touched rather than carried
forward PR after PR.
Change
python scripts/generate_repo_badges.py --test-count 6692(nothand-edited) — 6692 is
pytest --collect-only -qon this branch's tip.No other file changed.
Completion Ledger (§13.2)
assets/badge-tests.svgtitle/label/alt refreshed to the measured countgit diffscoped to this commit touches exactly one fileTest plan
python scripts/check_doc_claims.py --test-count 6692->doc claims OK (1 declared not-a-claim exemption(s))python scripts/generate_repo_badges.py --check --test-count 6692->badges OK (5 checked)pytest tests_py/scripts/test_generate_repo_badges.py -q->49 passed, 20 subtests passedruff check scripts/generate_repo_badges.py-> cleancheck_doc_claims.pyandgenerate_repo_badges.py --checkfail closed (exit 1) — the floor gate fix: eliminate the six-file test-count conflict class with a monotone floor (issue #293) #294 introduced is still blocking, not just decorativeBoy-scout check (§14)
Co-Authored-By: Claude noreply@anthropic.com