You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Applies the load-bearing review findings on scripts/check-release-notes.sh
that materially affect what the gate catches:
- A5 allow_ip pair: anchor v1 names with \b so the 'disallow_ip' prose
line cannot satisfy the allow_ip check via substring overlap.
Mirror the same anchoring on ban_ip for consistency.
- A2 v1 token set: add DEFAULT_WS_TIMEOUT, decorate_response, and
enqueue_response — RELEASE_NOTES.md already mentions them, but
they were not pinned by the gate.
- A5 rename pairs: add 'webserver(create_webserver const&)' pair so
the explicit-constructor rename is enforced on a single line.
Refactor: extract three helpers — check_tokens_present (A2/A3 loops),
extract_section_body (was inline), and check_section_cites (A6 pair) —
to remove the duplicated "loop + accumulate missing + print + exit"
shape and the duplicated "extract section + empty guard + grep + fail"
shape. No behavior change beyond the additions above.
Minor: drop a no-op `${var:-0}` default inside `$((…))`, silence the
markdownlint advisory's stderr explicitly (was '2>&1' which folded
error lines into the discarded stream), and add a why-not-shared note
above REQUIRED_V2_TOKENS explaining the deliberate duplication with
check-readme.sh.
`./scripts/check-release-notes.sh` still passes:
A1 exists; A2 35 v1 tokens; A3 26 v2 tokens; A4 7 sections;
A5 26 rename pairs; A6 threading+error citations; A7 disclaimer.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
0 commit comments