Change --fix-only exit semantics to mirror --fix

[charliermarsh]

Summary

After this change, running ruff --fix-only will exit with status code 0 unless the program hits an internal error, while ruff --fix-only --exit-non-zero-on-fix will exit with status code 1 if at least one diagnostic is fixed. This effectively mirrors ruff --fix, taking into account that non-fixable diagnostics are ignored. Previously, the --exit-non-zero-on-fix flag had no effect when combined with --fix-only.

Closes #4092.

[github-actions]

PR Check Results

Ecosystem

✅ ecosystem check detected no changes.

Benchmark

Linux

group                                      main                                   pr
-----                                      ----                                   --
linter/all-rules/large/dataset.py          1.00     14.8±0.53ms     2.7 MB/sec    1.01     15.0±0.51ms     2.7 MB/sec
linter/all-rules/numpy/ctypeslib.py        1.00      3.5±0.18ms     4.7 MB/sec    1.02      3.6±0.09ms     4.6 MB/sec
linter/all-rules/numpy/globals.py          1.06   459.8±14.85µs     6.4 MB/sec    1.00   432.0±21.92µs     6.8 MB/sec
linter/all-rules/pydantic/types.py         1.00      5.9±0.37ms     4.3 MB/sec    1.02      6.1±0.25ms     4.2 MB/sec
linter/default-rules/large/dataset.py      1.00      7.2±0.49ms     5.7 MB/sec    1.04      7.4±0.20ms     5.5 MB/sec
linter/default-rules/numpy/ctypeslib.py    1.01  1614.2±50.54µs    10.3 MB/sec    1.00  1593.5±33.90µs    10.4 MB/sec
linter/default-rules/numpy/globals.py      1.00    177.2±4.62µs    16.7 MB/sec    1.00    178.1±5.58µs    16.6 MB/sec
linter/default-rules/pydantic/types.py     1.00      3.3±0.07ms     7.8 MB/sec    1.05      3.5±0.09ms     7.4 MB/sec
parser/large/dataset.py                    1.00      5.6±0.31ms     7.3 MB/sec    1.05      5.8±0.16ms     7.0 MB/sec
parser/numpy/ctypeslib.py                  1.00  1122.1±28.38µs    14.8 MB/sec    1.03  1160.6±40.73µs    14.3 MB/sec
parser/numpy/globals.py                    1.01    117.4±4.59µs    25.1 MB/sec    1.00    115.8±4.28µs    25.5 MB/sec
parser/pydantic/types.py                   1.00      2.4±0.05ms    10.4 MB/sec    1.02      2.5±0.06ms    10.2 MB/sec

Windows

group                                      main                                   pr
-----                                      ----                                   --
linter/all-rules/large/dataset.py          1.02     12.4±0.17ms     3.3 MB/sec    1.00     12.1±0.08ms     3.4 MB/sec
linter/all-rules/numpy/ctypeslib.py        1.00      3.2±0.03ms     5.3 MB/sec    1.00      3.1±0.02ms     5.3 MB/sec
linter/all-rules/numpy/globals.py          1.00    321.0±4.98µs     9.2 MB/sec    1.02    328.2±6.32µs     9.0 MB/sec
linter/all-rules/pydantic/types.py         1.00      5.3±0.05ms     4.9 MB/sec    1.00      5.3±0.08ms     4.8 MB/sec
linter/default-rules/large/dataset.py      1.02      6.3±0.05ms     6.5 MB/sec    1.00      6.2±0.04ms     6.6 MB/sec
linter/default-rules/numpy/ctypeslib.py    1.01  1329.1±12.29µs    12.5 MB/sec    1.00  1316.6±12.05µs    12.6 MB/sec
linter/default-rules/numpy/globals.py      1.00    138.9±1.78µs    21.2 MB/sec    1.02    141.6±2.50µs    20.8 MB/sec
linter/default-rules/pydantic/types.py     1.02      2.9±0.03ms     8.9 MB/sec    1.00      2.8±0.02ms     9.1 MB/sec
parser/large/dataset.py                    1.00      4.9±0.05ms     8.3 MB/sec    1.01      4.9±0.05ms     8.3 MB/sec
parser/numpy/ctypeslib.py                  1.00    926.4±7.05µs    18.0 MB/sec    1.00    928.9±8.82µs    17.9 MB/sec
parser/numpy/globals.py                    1.00     95.4±1.06µs    30.9 MB/sec    1.00     95.8±1.38µs    30.8 MB/sec
parser/pydantic/types.py                   1.00      2.1±0.02ms    12.3 MB/sec    1.00      2.1±0.02ms    12.3 MB/sec

[MichaReiser]

I'm a bit confused. Isn't diff --fix and fix_only --fix-only? If so, shouldn't the cli.diff and fix_only branches be identical?

[charliermarsh]

No, those aren't the same. --fix, --fix-only, and --diff are all different options.

[MichaReiser]

Nit: Safe to fallthrough.

Suggested change

	} else {
	return Ok(ExitStatus::Success);
	}
	}

[MichaReiser]

Nit: It may be helpful to add some documentation explaining why/when we exit with a negative exit code for each option.