bug: rtk lint does not propagate ESLint non-zero exit code

[pavelbe]

Summary

rtk lint does not propagate ESLint non-zero exit code in a reproducible scenario (--max-warnings=0).

This can produce false-green CI/tooling behavior when lint should fail.

Environment

• rtk: 0.20.0
• bun: 1.3.9
• project: ~/Projects/pricedrop/apps/pricedrop
• checked on: 18 February 2026

Reproduction

cd ~/Projects/pricedrop/apps/pricedrop
npx eslint entrypoints/background.ts --max-warnings=0 >/dev/null 2>&1; echo raw=$?
rtk lint npx eslint entrypoints/background.ts --max-warnings=0 >/dev/null 2>&1; echo rtk=$?

Observed:

raw=1
rtk=0

Expected

rtk lint ... should return the same exit code as underlying ESLint command (1 in this repro).

Actual

rtk lint ... returns 0.

Impact

• CI/automation may continue despite lint failure conditions.
• Exit-code parity is broken for wrapper mode.

[pszymkowiak]

Thanks for the thorough bug reports — #185, #186 and #187 are all really helpful!

This one is straightforward: src/lint_cmd.rs line 176-193 captures the exit code but never propagates it. The function always returns Ok(()) → exit 0.

Fix is to add std::process::exit(exit_code) when the underlying linter fails, same pattern as other commands (e.g. run_compose_passthrough). PRs welcome!

Note: #186 (duplicated diagnostics) has a related root cause — when using rtk lint npx eslint ..., RTK detects npx as the linter name instead of eslint, so it skips the JSON parser and falls back to the generic text filter.