cli: `wheels test --verbose` and `--ci` have no observable effect — docs promise per-spec output and CI tightening

[bpamiri]

Observed (released brew CLI 4.0.3, fresh scaffold with passing + failing specs)

wheels test --verbose > v_sub.out 2>&1
wheels test           > v_none.out 2>&1
wheels -v test        > v_root.out 2>&1
diff v_none.out v_sub.out   # only LuCLI's 'Executing module shortcut' notice differs
diff v_root.out v_sub.out   # byte-identical

No per-spec lines are printed for passing specs in any form — failures print identically with and without the flag. Likewise wheels test --ci vs plain on a failing run: stdout byte-identical, both exit 1.

Docs contract being violated (command-line-tools/wheels-commands/testing.mdx)

• line 61: --verbose, -v — "Print per-spec output instead of the summary line."
• line 62: --ci — "CI mode: tightens exit codes and output for GitHub Actions and similar runners." (exit codes are already non-zero by default since the 4.0.3 exit-code fix, so --ci currently adds nothing)

Prior evidence absorbed

• P1 batch 2 testing audit (IDX-11): ciMode is parsed but never used — confirmed byte-identical output; the testing guide's --ci row was flagged there too (catalog row IDX-11, no issue filed at the time).
• fix(cli): repair broken commands and correct command docs (CLI audit) #2882 CLI audit flagged --ci as inert (deferred); roadmap: high-impact CLI/MCP tooling-honesty gaps — failure exit codes (incl. upgrade check) + MCP tool input schemas #2963 tracks the broader CLI tooling-honesty roadmap — this can fold into either if preferred. The --verbose half is NEW (this audit).

Ask

Either wire both flags (verbose → per-spec TestBox output; ci → whatever differentiated behavior is intended, e.g. no-color/annotations) or correct the testing.mdx flag rows. Module.cfc parses both (parsed.verbose; --ci flag in the test ArgSpec ~line 676) — they just don't change the rendered report.

Dedupe: distinct from #3083 (runner fallback) and #3081 (migrate exit codes); not addressed by in-flight #3106 (runner JSON honesty). Found during the P2 guide-behavioral-audit (p2-2-cliref, claim testing-05).

🤖 Generated with Claude Code

[wheels-bot]

Wheels Bot — Triage

Classified as bug.

The report describes observable wrong behavior with a clear what-happened / what-expected: wheels test --verbose and wheels test --ci produce byte-identical stdout to a plain run (verified via diff), yet testing.mdx documents --verbose as "Print per-spec output instead of the summary line" and --ci as tightening exit codes/output. Both flags are parsed but inert.

Suspected layer

cli — the fix lives in the CLI test runner.

Fix sketch

Both flags are parsed in cli/lucli/Module.cfc (parsed.verbose ~line 692; --ci ArgSpec ~line 676 → opts.ci/ciMode ~line 732) and threaded through to runTests(..., verboseOutput, ..., ciMode, ...) and displayTestResults(result, verboseOutput, ...) (~lines 5091/5188/5351/5413). The verbose branch in displayTestResults (~line 5413, if (arguments.verboseOutput && ...bundleStats...)) only renders a tree when bundleStats is present — so the likely change is in the result-rendering path (ensure per-spec lines actually emit for the default reporter) and the ciMode consumer (currently parsed but never differentiates output). The issue also flags a docs-vs-behavior fork: either wire the flags to do what the guide promises, or correct the testing.mdx flag rows — that choice is a design decision for propose-fix to weigh.

Confidence: `medium`

The report is precise and pins the exact parse sites, but the fix carries a genuine design trade-off (wire the behavior vs. correct the docs), `--ci`'s intended differentiated output is undefined in the issue, and it touches the shared CLI test-runner render path — so at most `medium`.

<!-- wheels-bot:triage:3113 -->
<!-- wheels-bot:triage-class:bug -->
<!-- wheels-bot:triage-confidence:medium -->

[wheels-bot]

Wheels Bot — Fix proposed

Draft PR: #3132

Spec: cli/lucli/tests/specs/commands/TestCommandSpec.cfc — failing → passing
(unit specs for $buildCiAnnotations plus ModuleOutputCapture-based
observable-output specs for --ci and --verbose)
Implementation: cli/lucli/Module.cfc (displayTestResults ciMode wiring +
new $buildCiAnnotations / annotation-encoding helpers)

Scope note — this is a partial-scope fix (hence Refs #3113, not Fixes):

• --ci is the live inert flag and is the focus of this PR. It was parsed
  and threaded into runTests() but never consumed; it now emits one GitHub
  Actions ::error annotation per failed/errored spec.
• --verbose already renders a per-spec tree on develop
  (displayTestResults line ~5413, added 2026-03-13; the JSON reporter returns
  getMemento(includeDebugBuffer=true) which carries bundleStats). The
  released 4.0.3 the report was filed against predates that. I added a
  regression-lock spec for it, but could not confirm the verbose half
  end-to-end against a live server in the bot sandbox — so I left the issue
  open for a human to verify that criterion before closing.

A human review is required before merge. The Reviewer will weigh in shortly.
Doc updates (MDX guides, .ai/wheels/, CLAUDE.md) are handled separately by
bot-update-docs.yml.