ci.yml's all-features Configure step points at a job name that does not exist

[Yaraslaut]

Summary

A comment in ci.yml's linux-all-features Configure step points the reader at
a job that does not exist. It names check-workflow-option-coverage; the job
added by #623 is called option-coverage.

.github/workflows/ci.yml:1838 (on a8511aa6):

          # ODBC/SQLite/yaml-cpp/libzip set the bank example's Lightweight
          # fetch requires. The check-workflow-option-coverage job below is
          # what keeps the next option from repeating this.

The job it means:

$ grep -n '^  option-coverage:' .github/workflows/ci.yml
2424:  option-coverage:

$ grep -rn 'check-workflow-option-coverage' . --exclude-dir=.git
./.github/workflows/ci.yml:1838:          # fetch requires. The check-workflow-option-coverage job below is

Nothing else in the tree uses that name — not a job id, not a job name:, not
a script. The nearest real artifacts are the job option-coverage, its
name: "Every MORPH_BUILD_* option is built by some job", and the script
scripts/check_workflow_option_coverage.py. The comment reads as though it
were formed from the script's filename.

Why it is worth fixing rather than tolerating

This comment is the pointer from morph#605's fix site to the gate that keeps
morph#605 from recurring. A reader following it searches ci.yml for
check-workflow-option-coverage, finds nothing, and is left unable to tell
whether the gate exists at all — which is the one question the sentence was
written to answer.

It is also the same class as #621, one field over: a comment that names the
wrong job. #621 was about placement, this is about the name, and neither is
visible to anything in CI because YAML comments are inert.

Verification status: reproduced

Measured on a8511aa6 by the two greps above, run in a clean worktree. Nothing
inferred: the string occurs once in the tree, and no job or file carries it.

I did not check the other six workflows for comments naming jobs that do
not exist — this is a single site found by grepping for one name, not a sweep.
A sweep is mechanisable (every <name> job mention in a workflow comment,
checked against that file's top-level job keys) and would be the general form
of this issue.

What would change the verdict

Close when the comment names option-coverage, or names the job by its
name: string.

Re-open, or widen into the sweep above, if other workflow comments are found
naming jobs that do not exist.

Found while working #621, which moves banner comments in ci.yml but does not
touch this line. Filed rather than folded in, per AGENTS.md.

[Yaraslaut]

Triage: valid — reproduced, and the fix is one word

Measured on a8511aa6:

$ sed -n '1844p' .github/workflows/ci.yml
          # fetch requires. The check-workflow-option-coverage job below is

$ grep -n '^  option-coverage:' .github/workflows/ci.yml
2431:  option-coverage:

There is no job named check-workflow-option-coverage anywhere in .github/workflows/; #623's job is option-coverage, and scripts/check_workflow_option_coverage.py is the script it runs. The comment names the script where it means the job.

triage: valid, documentation, area: ci.

For whoever takes it: ci.yml is currently held by PR #638, which moves seven banner comment blocks in that file. Take this after #638 lands, or the edit collides with the move for one line's benefit. It is also a natural companion to #638's new check_workflow_job_banners.py, which pairs banners to jobs and would not catch this — a prose reference to a job id inside a comment body is not a banner, so nothing gates it. Worth saying in the fix whether that residual is accepted or closed.

🤖 Generated with Claude Code

https://claude.ai/code/session_01VptDWG2fKr2vBnLSJcgzgW