Understanding Results

Anatomy of a finding

In the terminal report, each finding shows:

• its severity and message,
• a --> primary location and related: locations, each retaining its full test title and path:line:column,
• a rule-specific action: line naming the remedy.

Findings are grouped by rule, ordered by first-participant file. The footer summarizes analyzed tests and files, displayed severities, inline suppressions, and baseline debt.

Verdicts drive the remedy

The verdict names what the finding claims, because the remedies differ. See Detection Concepts for which rules produce which verdict — the short version:

Verdict	What it claims	Remedy
redundant	One copy can go.	Review removal, or repair intended coverage.
consolidatable	Every case is needed.	One test.each table — do not delete rows.
name-collision	Two tests share a title.	Rename one.
similar-uncertain	The tests look equivalent; that is not established.	Review decides.

Fingerprints

A finding’s fingerprint is its stable identity across runs: the rule plus the normalized participants, not file positions, so it survives reformatting and line moves. Baselines match on it.

States a finding can carry

• Baselined — visible in reports with a [baselined] marker and the original severity, but excluded from the gate.
• Suppressed — a vitest-dedup-disable-next-line directive with a participant on the next line. A suppressed finding never consumes baseline allowance and is never recorded by --update-baseline, because a baseline records accepted debt and the directive claims there is none. Every report states how many findings were suppressed this way.
• Inert participants — machine formats carry a per-participant .skip/.todo flag, so an integration can tell which side of a duplicate is dead code.

Machine-readable fields

Structured fields are provided rather than asking a consumer to parse message. Beyond verdict and fingerprint, findings may carry registrationFiles — an array parallel to locations listing the sorted caller paths a declaration site represents, when the file declaring a test differs from the file registering it.

Determinism

The same admitted input produces byte-identical output. Findings are ordered by their participants’ locations (root-relative path, then position), then by rule, then by message — compared component by component, so order does not depend on the platform’s path separator.