astral-sh/ruff · error · anyhow::Error

Working directory does not exist

Error message

Working directory does not exist

What it means

Ruff resolves configuration relative to the current working directory. If `std::env::current_dir()` fails — typically because the process's working directory was deleted or is inaccessible — `resolve` bails with this error since no configuration or file discovery can proceed reliably.

Source

Thrown at crates/ruff/src/resolve.rs:25

use ruff_workspace::configuration::Configuration;
use ruff_workspace::pyproject::{self, find_fallback_target_version};
use ruff_workspace::resolver::{
    ConfigurationOrigin, ConfigurationTransformer, PyprojectConfig, PyprojectDiscoveryStrategy,
    resolve_root_settings,
};

use ruff_python_ast as ast;

use crate::args::ConfigArguments;

/// Resolve the relevant settings strategy and defaults for the current
/// invocation.
pub fn resolve(
    config_arguments: &ConfigArguments,
    stdin_filename: Option<&Path>,
) -> Result<PyprojectConfig> {
    let Ok(cwd) = std::env::current_dir() else {
        bail!("Working directory does not exist")
    };

    // First priority: if we're running in isolated mode, use the default settings.
    if config_arguments.isolated {
        let config = config_arguments.transform(Configuration::default());
        let settings = config.into_settings(&cwd)?;
        debug!("Isolated mode, not reading any pyproject.toml");
        return Ok(PyprojectConfig::new(
            PyprojectDiscoveryStrategy::Fixed,
            settings,
            None,
        ));
    }

    // Second priority: the user specified a `pyproject.toml` file. Use that
    // `pyproject.toml` for _all_ configuration, and resolve paths relative to the
    // current working directory. (This matches ESLint's behavior.)
    if let Some(pyproject) = config_arguments.config_file() {

View on GitHub (pinned to 26f38c119c)

Solutions

1. cd to an existing directory (e.g. the project root) before running ruff.
2. Recreate the deleted working directory or restore the project checkout.
3. Fix the wrapper/CI script so it runs ruff from a valid path, e.g. `cd "$PROJECT_DIR" && ruff check .`.

Example fix

// before: shell sits in a deleted directory
ruff check .
// after
cd /path/to/project && ruff check .

# Shell: bail early if the CWD is gone before invoking ruff
cd . 2>/dev/null || { echo "working directory no longer exists"; exit 1; }

// In Rust wrappers: check the CWD before spawning ruff
match std::env::current_dir() {
    Ok(dir) => run_ruff(&dir),
    Err(_) => eprintln!("Working directory does not exist; cd to a valid path first"),
}

When it happens

Trigger: Invoking ruff from a shell whose current directory has been deleted (or is unreadable), so `std::env::current_dir()` returns Err at the start of `resolve`.

Common situations: Running ruff in CI or containers after a build step removed the directory the shell sits in; long-lived shells whose project folder was renamed/deleted; sandboxed environments revoking access to the CWD.

Related errors

• No files found under the given path
• Expected {}
• Unknown option: {key}
• The argument `--config={}` cannot be used with `--isolated`
• You cannot specify more than one configuration file on the c

AI-assisted analysis of astral-sh/ruff@26f38c119c (2026-09-05). Data as JSON: /api/errors/3e48051e2e60c040. Report an issue: GitHub.