fix: PATHEXT resolution on Windows for JS/TS and Python tools by FlorianBruniaux · Pull Request #492 · rtk-ai/rtk

FlorianBruniaux · 2026-03-10T23:21:06Z

Summary

Add which crate (v8) for cross-platform binary resolution honoring PATHEXT on Windows
Two new helpers in utils.rs: resolve_binary() and tool_exists() replacing all Command::new("which") usages
Apply resolve_binary() to all JS/TS and Python tool entry points, plus the 3 central passthrough/TOML/proxy entry points in main.rs
Remove 4 duplicate local which_command() functions (mypy, pytest, pip, ccusage)

Root cause

On Windows, JS/TS tools (vitest, eslint, tsc, prettier, pnpm, next, playwright, prisma...) are installed as .CMD/.BAT/.PS1 wrappers. std::process::Command::new("vitest") doesn't resolve PATHEXT, so all these commands fail with "program not found". Additionally, Command::new("which") doesn't exist on Windows — it's where.

Test plan

cargo fmt --all && cargo clippy --all-targets — 0 errors
cargo test — 762 tests pass
4 new unit tests for resolve_binary() and tool_exists() helpers
Manual smoke test: rtk git status works
Windows validation by issue reporter (@F0rty-Tw0)

Fixes #212

🤖 Generated with Claude Code

Copilot

Pull request overview

This PR fixes Windows command spawning failures by resolving tool binaries in a cross-platform way that honors PATHEXT (e.g., .CMD/.BAT shims for Node/Python tools), replacing prior reliance on the external which executable.

Changes:

Add which crate and introduce utils::resolve_binary() + utils::tool_exists() as the canonical binary resolution/check mechanism.
Update JS/TS, Python, and central passthrough/proxy entry points to spawn commands via resolve_binary() (and replace ad-hoc which checks).
Remove duplicated per-command which_command() helpers and update docs/version references to 0.28.2.

Reviewed changes

Copilot reviewed 21 out of 22 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
src/utils.rs	Adds `resolve_binary()` / `tool_exists()` helpers and unit tests; updates `package_manager_exec()` to use them
src/tsc_cmd.rs	Uses `tool_exists()` + `resolve_binary()` for `tsc`/`npx`
src/tree.rs	Replaces `which tree` check and spawns `tree` via resolved binary
src/ruff_cmd.rs	Spawns `ruff` via resolved binary
src/pytest_cmd.rs	Uses `tool_exists()` + `resolve_binary()` and removes local `which_command()`
src/prisma_cmd.rs	Uses `tool_exists()` + `resolve_binary()` for `prisma`/`npx`
src/pnpm_cmd.rs	Spawns `pnpm` via resolved binary across subcommands/passthrough
src/playwright_cmd.rs	Spawns `pnpm`/`yarn`/`npx` via resolved binary
src/pip_cmd.rs	Uses `tool_exists()` + `resolve_binary()` and removes local `which_command()`
src/npm_cmd.rs	Spawns `npm` via resolved binary
src/next_cmd.rs	Uses `tool_exists()` + `resolve_binary()` for `next`/`npx`
src/mypy_cmd.rs	Uses `tool_exists()` + `resolve_binary()` and removes local `which_command()`
src/main.rs	Applies `resolve_binary()` to fallback execution, proxy mode, and some `npx` paths
src/lint_cmd.rs	Spawns python linters via resolved binary
src/golangci_cmd.rs	Spawns `golangci-lint` via resolved binary
src/format_cmd.rs	Spawns non-JS formatters via resolved binary
src/ccusage.rs	Uses `tool_exists()` + `resolve_binary()` for `ccusage`/`npx` fallback
README.md	Updates expected `rtk --version` output
Cargo.toml	Bumps version and adds `which = "8"` dependency
Cargo.lock	Locks `which` dependency
CLAUDE.md	Updates expected `rtk --version` output
ARCHITECTURE.md	Updates referenced RTK version

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-03-10T23:27:58Z

src/utils.rs

+    #[test]
+    fn test_resolve_binary_known() {
+        // cargo exists on all dev machines
+        let resolved = resolve_binary("cargo");
+        assert!(!resolved.is_empty());
+    }
+
+    #[test]
+    fn test_resolve_binary_unknown() {
+        let resolved = resolve_binary("nonexistent_tool_xyz_99999");
+        assert_eq!(
+            resolved,
+            std::ffi::OsString::from("nonexistent_tool_xyz_99999")
+        );
+    }
+
+    #[test]
+    fn test_tool_exists_positive() {
+        assert!(tool_exists("cargo"));
+    }
+
+    #[test]
+    fn test_tool_exists_negative() {
+        assert!(!tool_exists("nonexistent_tool_xyz_99999"));
+    }


The new unit tests assume cargo is discoverable on PATH (via tool_exists("cargo")). That’s not guaranteed in all test runners (e.g., if cargo test is invoked via an absolute path and PATH doesn’t include the cargo bin dir), which can make these tests flaky. Consider making the tests self-contained by creating a temporary executable/shim in a temp dir, prepending that dir to PATH (and on Windows using a .cmd file to exercise PATHEXT), then asserting tool_exists() and resolve_binary() behave as expected.

Copilot · 2026-03-10T23:27:58Z

src/tsc_cmd.rs

    // Try tsc directly first, fallback to npx if not found
-    let tsc_exists = Command::new("which")
-        .arg("tsc")
-        .output()
-        .map(|o| o.status.success())
-        .unwrap_or(false);
+    let tsc_exists = tool_exists("tsc");

    let mut cmd = if tsc_exists {
-        Command::new("tsc")
+        Command::new(resolve_binary("tsc"))
    } else {
-        let mut c = Command::new("npx");
+        let mut c = Command::new(resolve_binary("npx"));
        c.arg("tsc");


This pattern does two PATH scans (tool_exists("tsc") and then resolve_binary("tsc"), both calling which::which). Consider avoiding the duplicate lookup by calling the resolver once (e.g., a helper that returns Option<OsString>/PathBuf), and branching on that result.

Add `which` crate (v8) and two central helpers in utils.rs: - `resolve_binary(name)` resolves a binary to its full path, honoring PATHEXT on Windows so `.CMD`/`.BAT`/`.PS1` wrappers are found - `tool_exists(name)` replaces all `Command::new("which")` usages (which does not exist on Windows) with a cross-platform check Apply `resolve_binary()` to: - All JS/TS tool entry points (tsc, next, prisma, playwright, npm, pnpm, ruff, golangci-lint, vitest via package_manager_exec) - Python tool entry points (mypy, pytest, pip/uv) - The 3 central passthrough/TOML/proxy entry points in main.rs - `package_manager_exec()` in utils.rs (covers vitest, prettier, lint, format, biome automatically) Remove 4 local `which_command()` duplicates (mypy_cmd, pytest_cmd, pip_cmd, ccusage) and the `binary_exists()` helper in ccusage. Fixes: #212 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Signed-off-by: Florian BRUNIAUX <florian@bruniaux.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Signed-off-by: Florian BRUNIAUX <florian@bruniaux.com>

F0rty-Tw0 · 2026-03-11T05:46:49Z

Nice work.
What about #269 ? Are we closing it?

pszymkowiak

ok reviewed

pszymkowiak · 2026-03-11T08:40:03Z

Great work @FlorianBruniaux! Clean and well-scoped fix. A few things to address:

P1 — go_cmd.rs not updated
4 Command::new("go") calls at lines 43, 102, 155, 213 were not updated to resolve_binary("go"). Not broken today since go ships as .exe, but breaks the consistency of the fix.

P1 — cargo_cmd.rs not updated
Same situation — 2 Command::new("cargo") calls. Single-line fix for uniformity.

P1 — Double PATH scan pattern
Several modules call tool_exists(name) then immediately resolve_binary(name) in the taken branch — that's two which::which() calls. Appears in next_cmd.rs, tsc_cmd.rs, mypy_cmd.rs, pytest_cmd.rs, prisma_cmd.rs, ccusage.rs. Idiomatic fix:

let mut cmd = match which::which("next") {
    Ok(path) => Command::new(path),
    Err(_) => {
        let mut c = Command::new(resolve_binary("npx"));
        c.arg("next");
        c
    }
};

P2 — test_resolve_binary_known could be stronger
Currently asserts !resolved.is_empty(). Could assert path.is_absolute() to verify actual resolution vs fallback.

Otherwise LGTM — good crate choice, DRY cleanup (4 duplicate which_command() removed), correct fallback semantics. Ready to merge after P1 fixes.

pszymkowiak · 2026-03-11T08:45:02Z

Hey @FlorianBruniaux, closing this in favor of #269 by @F0rty-Tw0 which addresses the same issue and was submitted first (Feb 26). It was already reviewed and LGTM'd. Sorry about the duplicate — we have too many PRs in flight right now and this one slipped through. Thanks for understanding!

Copilot AI review requested due to automatic review settings March 10, 2026 23:21

Copilot started reviewing on behalf of FlorianBruniaux March 10, 2026 23:21 View session

FlorianBruniaux requested a review from pszymkowiak March 10, 2026 23:21

FlorianBruniaux force-pushed the fix/issue-212 branch from ba9f5d8 to 5133e2a Compare March 10, 2026 23:22

Copilot AI reviewed Mar 10, 2026

View reviewed changes

FlorianBruniaux and others added 2 commits March 11, 2026 00:35

chore: bump version references to 0.28.2 for consistency with develop

2aff2fa

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Signed-off-by: Florian BRUNIAUX <florian@bruniaux.com>

FlorianBruniaux force-pushed the fix/issue-212 branch from 5133e2a to 2aff2fa Compare March 10, 2026 23:35

pszymkowiak approved these changes Mar 11, 2026

View reviewed changes

pszymkowiak mentioned this pull request Mar 11, 2026

fix windows PATHEXT resolution for .cmd/.bat wrappers #269

Merged

pszymkowiak closed this Mar 11, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix: PATHEXT resolution on Windows for JS/TS and Python tools#492

fix: PATHEXT resolution on Windows for JS/TS and Python tools#492
FlorianBruniaux wants to merge 2 commits intodevelopfrom
fix/issue-212

FlorianBruniaux commented Mar 10, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 10, 2026

Uh oh!

Copilot AI Mar 10, 2026

Uh oh!

F0rty-Tw0 commented Mar 11, 2026

Uh oh!

pszymkowiak left a comment

Uh oh!

pszymkowiak commented Mar 11, 2026

Uh oh!

pszymkowiak commented Mar 11, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

FlorianBruniaux commented Mar 10, 2026

Summary

Root cause

Test plan

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

F0rty-Tw0 commented Mar 11, 2026

Uh oh!

pszymkowiak left a comment

Choose a reason for hiding this comment

Uh oh!

pszymkowiak commented Mar 11, 2026

Uh oh!

pszymkowiak commented Mar 11, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants