Skip to content

fix: gate the curate shell test in CI and pre-commit - #35

Merged
RichardHightower merged 1 commit into
mainfrom
fix/gate-the-shell-test
Aug 3, 2026
Merged

fix: gate the curate shell test in CI and pre-commit#35
RichardHightower merged 1 commit into
mainfrom
fix/gate-the-shell-test

Conversation

@RichardHightower

Copy link
Copy Markdown
Contributor

tests/test_okf_curate.sh shipped in v0.3.1 referenced by neither the CI workflow nor the pre-commit hook. The repo's first shell test protected nothing.

What it covers is the point: the post-edit hook — the component that shipped configured but never firing for two releases, which is the headline bug v0.3.0 fixed. A check that exists but never runs is that same defect wearing different clothes, and reintroducing it in the release that fixed it would have been a poor joke.

Two lines, one per gate.

How it was found

The design-doc agent grepped both gate files instead of assuming a committed test was a running one. Confirmed independently — the only match for "curate" in either file was a comment:

.github/workflows/worklog.yml:20:      # (and okf-curate.sh) keep the lenient default.

Verified the gate actually bites

Same check applied to the graph suite in #8: restore the defect and confirm the gate refuses.

# put back the old hard-coded path filter
-  *.md|*.markdown) ;;
+  *knowledge/*|*sample-okf/*) ;;

$ bash tests/test_okf_curate.sh            # exit 1
$ hooks/pre-commit
worklog: curate tests failing — run: bash tests/test_okf_curate.sh

Restored, gate silent, exit 0.

Worth noting the first attempt at this verification was inconclusive and looked like a pass: pre-commit was failing earlier on roadmap staleness (I'd just filed the work item) and never reached the test line. Rendering the roadmap first made the real result visible. A guard you "verified" against an already-failing gate is not verified.

🤖 Generated with Claude Code

https://claude.ai/code/session_01LKmdobV4dykomYAENb3USC

tests/test_okf_curate.sh shipped in v0.3.1 referenced by neither the
workflow nor the pre-commit hook, so the repo's first shell test protected
nothing. It covers the post-edit hook -- the component that shipped
configured-but-never-firing for two releases. A check that exists but
never runs is that same defect wearing different clothes, which is a poor
thing to reintroduce in the release that fixed it.

Found by the design-doc pass, which grepped both gate files rather than
assuming a committed test was a running one.

Verified the gate bites: restoring the old hard-coded path filter in
okf-curate.sh fails the suite and pre-commit refuses the commit.

01KZ2EWBA88ZH2GBSBPCX8B42H

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LKmdobV4dykomYAENb3USC
@RichardHightower
RichardHightower merged commit fd328d2 into main Aug 3, 2026
2 checks passed
@RichardHightower
RichardHightower deleted the fix/gate-the-shell-test branch August 3, 2026 00:04
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant