What happened?

In the ecosystem CI run for oven-sh/bun, oxlint -c oxlint.json src/js reports 20 eslint(no-unreachable) errors. The output is much noisier than expected because the rule appears to report every nested statement/expression inside an unreachable region instead of one diagnostic for the unreachable region.

Run: https://github.com/oxc-project/oxc-ecosystem-ci/actions/runs/24795395393/job/72564872698

Example from the log

The command reports multiple diagnostics for the same unreachable block in src/js/internal/sql/postgres.ts:

× eslint(no-unreachable): Unreachable code.
   ╭─[src/js/internal/sql/postgres.ts:279:5]
279 │     $assert(result instanceof SQLResultArray, "Invalid result array");

× eslint(no-unreachable): Unreachable code.
   ╭─[src/js/internal/sql/postgres.ts:280:5]
280 │     if (typeof commandTag === "string") {

× eslint(no-unreachable): Unreachable code.
   ╭─[src/js/internal/sql/postgres.ts:280:41]
280 │     if (typeof commandTag === "string") {

× eslint(no-unreachable): Unreachable code.
   ╭─[src/js/internal/sql/postgres.ts:281:7]
281 │       if (commandTag.length > 0) {

× eslint(no-unreachable): Unreachable code.
   ╭─[src/js/internal/sql/postgres.ts:284:12]
284 │     } else {

× eslint(no-unreachable): Unreachable code.
   ╭─[src/js/internal/sql/postgres.ts:285:7]
285 │       result.command = cmds[commandTag];

A similar pattern occurs in src/js/node/_http_agent.ts: the rule reports the if, its condition, nested assignments, and following statements as separate no-unreachable errors. The job ends with:

Found 0 warnings and 20 errors.

Expected behavior

Report one no-unreachable diagnostic for each contiguous unreachable region, or at least suppress descendant diagnostics once a parent unreachable statement/range has already been reported.

For the Bun log above, the useful output would be closer to 1 diagnostic for the unreachable region in postgres.ts and 1 diagnostic for the unreachable region in _http_agent.ts, rather than 20 diagnostics.

Why this matters

The current output makes ecosystem CI failures harder to read and makes a single underlying control-flow issue look like many unrelated lint errors.

Notes

This is about diagnostic quality/noise reduction, not necessarily whether the code is considered unreachable. The main issue is duplicate/nested reporting for one unreachable region.