fix(exec): strip shims from PATH on Unix to prevent infinite recursion by jdx · Pull Request #8276 · jdx/mise

jdx · 2026-02-21T14:54:49Z

Summary

When not_found_auto_install preserves the shims directory in PATH, mise x -- tool can enter an infinite loop on Unix, eventually hitting E2BIG ("Argument list too long")
Commit 1: Strip shims from PATH during program resolution in Unix exec_program (parity with the existing Windows fix from 3ec95cf9a). This prevents recursion through shims.
Commit 2: Resolve the program via ts.which_bin() in Exec::run() before exec. This prevents recursion through wrapper scripts (e.g., .devcontainer/bin/tool) that call mise x -- tool — the wrapper precedes mise install paths in PATH, so without resolution the wrapper is found again instead of the real binary.
Add a Unix e2e regression test (test_exec_shim_recursion) mirroring the existing Windows test

Test plan

cargo check passes
cargo test — all 481 tests pass
All pre-commit lints pass (shfmt, shellcheck, cargo-fmt, etc.)
Tested on Coder workspace: coder ssh jdx.devcontainer -- 'claude --version' returns 2.1.50 (Claude Code) immediately (previously hung in infinite recursion)
./e2e/run_test cli/test_exec_shim_recursion passes
Existing exec e2e tests still pass (test_exec_chdir, test_exec_latest, test_shims_fallback)

🤖 Generated with Claude Code

Note

Medium Risk
Touches core exec/path resolution behavior; mistakes could change which binary gets executed or break command lookup on Unix in edge PATH configurations.

Overview
Fixes a Unix-only mise x infinite-recursion failure mode by ensuring the executed program resolves to the real tool binary rather than re-hitting shims or wrapper scripts earlier in PATH.

Exec::run() now pre-resolves bare program names to the mise-managed binary via ts.which_bin() (Unix only), and Unix exec_program now resolves the executable using a PATH with the shims directory removed for lookup only while still passing the full environment to the child process. Adds an e2e regression test (e2e/cli/test_exec_shim_recursion) that asserts a real binary is chosen over a fake shim when shims appear first in PATH.

^{Written by Cursor Bugbot for commit d53b221. This will update automatically on new commits. Configure here.}

gemini-code-assist · 2026-02-21T14:55:04Z

Summary of Changes

Hello @jdx, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request resolves a critical infinite recursion bug affecting mise x -- tool commands on Unix systems. By intelligently filtering the shims directory from the PATH during internal program lookup, it ensures that the correct executable is identified, preventing system resource exhaustion and improving the robustness of tool execution.

Highlights

Infinite Recursion Fix: Addressed an infinite recursion bug in mise x -- tool on Unix where shims in PATH could cause the wrapper script to be repeatedly executed instead of the real binary, leading to an E2BIG error.
PATH Handling for Program Resolution: Implemented a mechanism to strip the shims directory from PATH during program resolution within mise x, ensuring the real binary is found. The full PATH (including shims) is still passed to the child process.
Regression Test Added: Added a new Unix end-to-end regression test (test_exec_shim_recursion) to prevent future regressions of this issue, mirroring an existing Windows test.

Changelog

e2e/cli/test_exec_shim_recursion
- Added a new end-to-end test to verify that mise x correctly resolves tools and avoids infinite recursion when shims are present in the PATH.
src/cli/exec.rs
- Modified the exec command logic to filter out the shims directory from the PATH when resolving the program to execute, preventing recursive calls to shims.
- Ensured that the full PATH (including shims) is still passed to the child process for its own subprocesses.

Activity

Author confirmed cargo check passes.
Author confirmed cargo test passes all 481 tests.
Author confirmed all pre-commit lints pass.
Author confirmed ./e2e/run_test cli/test_exec_shim_recursion passes.
Author confirmed existing exec e2e tests still pass (test_exec_chdir, test_exec_latest, test_shims_fallback).
Author confirmed testing on Coder workspace: coder ssh jdx.devcontainer -- 'claude --version'.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	`/gemini review`	Performs a code review for the current pull request in its current state.
Pull Request Summary	`/gemini summary`	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	`/gemini help`	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

gemini-code-assist

Code Review

The pull request correctly addresses the infinite recursion issue on Unix by filtering the shims directory from PATH during program resolution. This brings the Unix implementation in line with the existing Windows fix. I've suggested a few refinements to the resolution logic to improve robustness and ensure consistency with the Windows implementation.

gemini-code-assist · 2026-02-21T14:56:59Z

src/cli/exec.rs

+    let program = if program.to_string_lossy().contains('/') {
+        // Already a path, no need to resolve
+        program
+    } else {
+        let cwd = crate::dirs::CWD.clone().unwrap_or_default();
+        let lookup_path = env.get(&*env::PATH_KEY).map(|path_val| {
+            let shims_dir = &*crate::dirs::SHIMS;
+            let filtered: Vec<_> = std::env::split_paths(&OsString::from(path_val))
+                .filter(|p| p != shims_dir)
+                .collect();
+            std::env::join_paths(&filtered).unwrap()
+        });
+        match which::which_in(&program, lookup_path, cwd) {
+            Ok(resolved) => resolved.into_os_string(),
+            Err(_) => program, // Fall back to original if resolution fails
+        }
+    };


The program resolution logic for Unix can be improved for robustness and consistency with the Windows implementation.

Path Expansion: Using crate::file::replace_path(p) when filtering PATH ensures that entries containing ~ are correctly handled, which is important for robustness and matches the Windows implementation.

Error Handling: Removing the fallback to the original program name when resolution fails is safer. If which_in cannot find the executable outside of the shims directory, falling back to the original name will likely cause execvp to find the shim again, leading back to the infinite recursion. Using ? to propagate the error is consistent with the Windows implementation and provides a clear 'command not found' error instead of potentially entering a loop.

let program = if program.to_string_lossy().contains('/') { // Already a path, no need to resolve program } else { let cwd = crate::dirs::CWD.clone().unwrap_or_default(); let lookup_path = env.get(&*env::PATH_KEY).map(|path_val| { let shims_dir = &*crate::dirs::SHIMS; let filtered: Vec<_> = std::env::split_paths(&OsString::from(path_val)) .filter(|p| crate::file::replace_path(p) != shims_dir) .collect(); std::env::join_paths(&filtered).unwrap() }); which::which_in(&program, lookup_path, cwd)?.into_os_string() };

Copilot

Pull request overview

This PR fixes an infinite recursion bug on Unix systems where mise x -- tool can enter an infinite loop when not_found_auto_install preserves the shims directory in PATH. The fix mirrors the existing Windows implementation by filtering the shims directory from PATH during program resolution while still passing the full PATH to the child process.

Changes:

Apply shims directory filtering to Unix exec_program function (matching Windows fix from commit 3ec95cf)
Add Unix e2e regression test test_exec_shim_recursion to verify shims are not recursively executed

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
src/cli/exec.rs	Adds shim directory filtering logic to Unix exec_program function to prevent infinite recursion when resolving program paths
e2e/cli/test_exec_shim_recursion	Adds regression test that verifies mise exec resolves real tool binaries instead of shims when shims appear first in PATH

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

Copilot · 2026-02-21T14:59:55Z

src/cli/exec.rs

+            let shims_dir = &*crate::dirs::SHIMS;
+            let filtered: Vec<_> = std::env::split_paths(&OsString::from(path_val))
+                .filter(|p| p != shims_dir)


The path comparison here does not expand tilde (~) paths, which could cause the shims directory to not be filtered correctly if it's specified with a tilde. The Windows implementation (lines 247-248) uses crate::file::replace_path(p) to expand tilde before comparison. Consider adding the same expansion here for consistency and correctness.

Suggested change

let shims_dir = &*crate::dirs::SHIMS;

let filtered: Vec<_> = std::env::split_paths(&OsString::from(path_val))

.filter(|p| p != shims_dir)

let shims_dir = crate::file::replace_path(&*crate::dirs::SHIMS);

let filtered: Vec<_> = std::env::split_paths(&OsString::from(path_val))

.filter(|p| crate::file::replace_path(p) != shims_dir)

Copilot · 2026-02-21T14:59:55Z

src/cli/exec.rs

+        match which::which_in(&program, lookup_path, cwd) {
+            Ok(resolved) => resolved.into_os_string(),
+            Err(_) => program, // Fall back to original if resolution fails
+        }


This fallback behavior differs from the Windows implementation (line 253), which propagates the error using the ? operator. Falling back to the original program could mask issues where the tool doesn't exist at all, making it harder to debug. Consider whether this should propagate the error like Windows does, or if there's a specific reason for the different behavior that should be documented in a comment.

Suggested change

match which::which_in(&program, lookup_path, cwd) {

Ok(resolved) => resolved.into_os_string(),

Err(_) => program, // Fall back to original if resolution fails

}

which::which_in(&program, lookup_path, cwd)?.into_os_string()

Copilot · 2026-02-21T14:59:56Z

src/cli/exec.rs

+            let filtered: Vec<_> = std::env::split_paths(&OsString::from(path_val))
+                .filter(|p| p != shims_dir)
+                .collect();
+            std::env::join_paths(&filtered).unwrap()


This unwrap() could panic if join_paths fails (e.g., if a path contains a null byte on Unix). The Windows implementation (line 251) also uses unwrap() here, but it would be more robust to handle this error case, perhaps by propagating the error or logging a warning and falling back to the original PATH.

Suggested change

std::env::join_paths(&filtered).unwrap()

match std::env::join_paths(&filtered) {

Ok(joined) => joined,

Err(err) => {

eprintln!(

"warning: failed to rebuild PATH without shims ({}); using original PATH",

err

);

OsString::from(path_val)

}

}

cursor

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is ON, but a Cloud Agent failed to start.}

src/cli/exec.rs

github-actions · 2026-02-21T15:10:44Z

Hyperfine Performance

`mise x -- echo`

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`mise-2026.2.17 x -- echo`	21.9 ± 0.2	21.4	24.2	1.00
`mise x -- echo`	22.8 ± 0.2	22.3	25.5	1.04 ± 0.02

`mise env`

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`mise-2026.2.17 env`	21.5 ± 0.6	20.9	27.8	1.00
`mise env`	21.6 ± 0.2	20.7	23.4	1.00 ± 0.03

`mise hook-env`

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`mise-2026.2.17 hook-env`	22.0 ± 0.2	21.0	24.9	1.00
`mise hook-env`	22.2 ± 0.2	21.7	22.9	1.01 ± 0.01

`mise ls`

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`mise-2026.2.17 ls`	19.6 ± 0.2	18.9	21.4	1.00
`mise ls`	19.9 ± 0.3	19.4	22.2	1.01 ± 0.02

`xtasks/test/perf`

Command	mise-2026.2.17	mise	Variance
install (cached)	121ms	121ms	+0%
ls (cached)	74ms	74ms	+0%
bin-paths (cached)	78ms	78ms	+0%
task-ls (cached)	787ms	793ms	+0%

When `not_found_auto_install` preserves the shims directory in PATH, `mise x -- tool` can enter an infinite loop: execvp finds a wrapper script (or shim) instead of the real binary, the wrapper calls `mise x -- tool` again, and PATH grows until hitting E2BIG. Apply the same fix that Windows already has (3ec95cf): filter the shims directory from PATH when resolving the program via which::which_in, while still passing the full PATH (with shims) to the child process. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

The shim-stripping fix prevents recursion through shims, but wrapper scripts in non-shim directories (e.g., .devcontainer/bin/tool) can still cause infinite loops. These wrappers call `mise x -- tool`, and since .devcontainer/bin precedes mise install paths in PATH, execvp finds the wrapper again. Fix by resolving the program via ts.which_bin() before exec. If the program is provided by a mise-managed tool, use the full install path, bypassing any wrappers or shims in PATH. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

The which_bin pre-resolution converts a bare command name (e.g. "prettier") to its full installed path before calling exec_program. On Windows, this bypasses the which::which_in PATHEXT resolution that correctly finds .cmd/.exe wrappers, causing "not a valid Win32 application" (OS error 193) when the pre-resolved .cmd path is passed directly to duct::cmd. Make the which_bin resolution #[cfg(unix)] only since the underlying problem (execvp wrapper recursion) is Unix-specific. On Windows, exec_program's which::which_in with shim-filtered PATH already handles the recursion case correctly. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

@jdx

### 🚀 Features - **(install)** auto-lock all platforms after tool installation by @jdx in [#8277](#8277) ### 🐛 Bug Fixes - **(config)** respect --yes flag for config trust prompts by @jdx in [#8288](#8288) - **(exec)** strip shims from PATH on Unix to prevent infinite recursion by @jdx in [#8276](#8276) - **(install)** validate --locked before --dry-run short-circuit by @altendky in [#8290](#8290) - **(release)** refresh PATH after mise up in release-plz by @jdx in [#8292](#8292) - **(schema)** replace unevaluatedProperties with additionalProperties by @jdx in [#8285](#8285) - **(task)** avoid duplicated stderr on task failure in replacing mode by @jdx in [#8275](#8275) - **(task)** use process groups to kill child process trees on Unix by @jdx in [#8279](#8279) - **(task)** run depends_post tasks even when parent task fails by @jdx in [#8274](#8274) - **(task)** suggest similar commands when mistyping a CLI subcommand by @jdx in [#8286](#8286) - **(task)** execute monorepo subdirectory prepare steps from root by @jdx in [#8291](#8291) - **(upgrade)** don't force-reinstall already installed versions by @jdx in [#8282](#8282) - **(watch)** restore terminal state after watchexec exits by @jdx in [#8273](#8273) ### 📚 Documentation - clarify that MISE_CEILING_PATHS excludes the ceiling directory itself by @jdx in [#8283](#8283) ### Chore - replace gen-release-notes script with communique by @jdx in [#8289](#8289) ### New Contributors - @altendky made their first contribution in [#8290](#8290) ## 📦 Aqua Registry Updates #### New Packages (4) - [`Skarlso/crd-to-sample-yaml`](https://github.com/Skarlso/crd-to-sample-yaml) - [`kunobi-ninja/kunobi-releases`](https://github.com/kunobi-ninja/kunobi-releases) - [`swanysimon/markdownlint-rs`](https://github.com/swanysimon/markdownlint-rs) - [`tmux/tmux-builds`](https://github.com/tmux/tmux-builds) #### Updated Packages (2) - [`firecow/gitlab-ci-local`](https://github.com/firecow/gitlab-ci-local) - [`k1LoW/runn`](https://github.com/k1LoW/runn)

@jdx

## [2026.2.18](https://github.com/jdx/mise/compare/v2026.2.17..v2026.2.18) - 2026-02-21 ### 🚀 Features - **(install)** auto-lock all platforms after tool installation by @jdx in [#8277](jdx/mise#8277) ### 🐛 Bug Fixes - **(config)** respect --yes flag for config trust prompts by @jdx in [#8288](jdx/mise#8288) - **(exec)** strip shims from PATH on Unix to prevent infinite recursion by @jdx in [#8276](jdx/mise#8276) - **(install)** validate --locked before --dry-run short-circuit by @altendky in [#8290](jdx/mise#8290) - **(release)** refresh PATH after mise up in release-plz by @jdx in [#8292](jdx/mise#8292) - **(schema)** replace unevaluatedProperties with additionalProperties by @jdx in [#8285](jdx/mise#8285) - **(task)** avoid duplicated stderr on task failure in replacing mode by @jdx in [#8275](jdx/mise#8275) - **(task)** use process groups to kill child process trees on Unix by @jdx in [#8279](jdx/mise#8279) - **(task)** run depends_post tasks even when parent task fails by @jdx in [#8274](jdx/mise#8274) - **(task)** suggest similar commands when mistyping a CLI subcommand by @jdx in [#8286](jdx/mise#8286) - **(task)** execute monorepo subdirectory prepare steps from root by @jdx in [#8291](jdx/mise#8291) - **(upgrade)** don't force-reinstall already installed versions by @jdx in [#8282](jdx/mise#8282) - **(watch)** restore terminal state after watchexec exits by @jdx in [#8273](jdx/mise#8273) ### 📚 Documentation - clarify that MISE_CEILING_PATHS excludes the ceiling directory itself by @jdx in [#8283](jdx/mise#8283) ### Chore - replace gen-release-notes script with communique by @jdx in [#8289](jdx/mise#8289) ### New Contributors - @altendky made their first contribution in [#8290](jdx/mise#8290) ### 📦 Aqua Registry Updates #### New Packages (4) - [`Skarlso/crd-to-sample-yaml`](https://github.com/Skarlso/crd-to-sample-yaml) - [`kunobi-ninja/kunobi-releases`](https://github.com/kunobi-ninja/kunobi-releases) - [`swanysimon/markdownlint-rs`](https://github.com/swanysimon/markdownlint-rs) - [`tmux/tmux-builds`](https://github.com/tmux/tmux-builds) #### Updated Packages (2) - [`firecow/gitlab-ci-local`](https://github.com/firecow/gitlab-ci-local) - [`k1LoW/runn`](https://github.com/k1LoW/runn) ## [2026.2.17](https://github.com/jdx/mise/compare/v2026.2.16..v2026.2.17) - 2026-02-19 ### 🚀 Features - **(prepare)** update mtime of outputs after command is run by @halms in [#8243](jdx/mise#8243) ### 🐛 Bug Fixes - **(install)** use backend bin paths for per-tool postinstall hooks by @jdx in [#8234](jdx/mise#8234) - **(use)** write to config.toml instead of config.local.toml by @jdx in [#8240](jdx/mise#8240) - default legacy .mise.backend installs to non-explicit by @jean-humann in [#8245](jdx/mise#8245) ### 🚜 Refactor - **(config)** consolidate flat task_* settings into nested task.* by @jdx in [#8239](jdx/mise#8239) ### Chore - **(prepare)** refactor common code into ProviderBase by @halms in [#8246](jdx/mise#8246) ### 📦 Aqua Registry Updates #### Updated Packages (1) - [`namespacelabs/foundation/nsc`](https://github.com/namespacelabs/foundation/nsc)

Remove the `which_bin` pre-resolution that was introduced in #8276 to prevent wrapper script recursion. This pre-resolution bypasses PATH entirely for mise-managed tools, breaking `_.python.venv` and similar configs where a virtualenv binary should take precedence over the mise-managed install. The shim stripping in `exec_program` (also from #8276) already prevents the main recursion vector (shims in PATH). The wrapper script edge case should not be solved by bypassing PATH for all programs. Fixes #8340 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Addresses Cursor Bugbot feedback on PR #8342. Simulates the .devcontainer/bin/tool scenario from #8276 where a wrapper script that calls `mise x -- tool` sits before mise tool paths in the system PATH. The test passes because `exec_program` resolves programs via the mise-computed PATH (where tool install paths are prepended before the original system PATH), so the real binary is found before the wrapper. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

) ## Summary - Remove the `ts.which_bin()` pre-resolution from `Exec::run()` that was introduced in #8276 - This pre-resolution bypassed PATH entirely for mise-managed tools, breaking `_.python.venv` and similar configs where a virtualenv binary should take precedence over the mise-managed install - The shim stripping in `exec_program` (also from #8276) already prevents shim recursion, which was the primary problem that PR solved ## Root Cause PR #8276 added two fixes for infinite recursion: 1. Strip shims from PATH in `exec_program` — prevents shim recursion (**kept**) 2. Pre-resolve bare program names via `which_bin` — prevents wrapper script recursion (**removed**) Fix #2 resolves e.g. `python` → `~/.local/share/mise/installs/python/3.9.1/bin/python` before PATH is computed. Since the result is an absolute path, `exec_program` skips PATH lookup entirely, so `.venv/bin/python` (added by `_.python.venv`) is never found. Fixes #8340 ## Test plan - [x] `cargo check` — no warnings - [x] `mise run test:e2e test_exec` — all exec e2e tests pass - [x] `test_exec_shim_recursion` still passes (shim stripping in `exec_program` handles this) 🤖 Generated with [Claude Code](https://claude.com/claude-code)  --- > [!NOTE] > **Medium Risk** > Changes how `mise x` resolves executables, which can alter which binary runs in edge PATH setups. Added E2E tests reduce regression risk, but behavior changes could impact users relying on the old forced mise-path resolution. > > **Overview** > `mise x` no longer pre-resolves bare command names to a mise-managed tool path via `Toolset::which_bin` in `Exec::run`, allowing normal PATH lookup to pick up higher-precedence overrides like virtualenv binaries. > > Adds two E2E regression tests: one ensuring PATH-prepended “venv” bins win over mise installs during `mise x`, and another ensuring a PATH-preceding wrapper that re-invokes `mise x` doesn’t hang (infinite recursion) and still executes the real tool. > > <sup>Written by [Cursor Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit b241665. This will update automatically on new commits. Configure [here](https://cursor.com/dashboard?tab=bugbot).</sup>  --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

jdx#8276) ## Summary - When `not_found_auto_install` preserves the shims directory in PATH, `mise x -- tool` can enter an infinite loop on Unix, eventually hitting `E2BIG` ("Argument list too long") - **Commit 1**: Strip shims from PATH during program resolution in Unix `exec_program` (parity with the existing Windows fix from `3ec95cf9a`). This prevents recursion through shims. - **Commit 2**: Resolve the program via `ts.which_bin()` in `Exec::run()` before exec. This prevents recursion through wrapper scripts (e.g., `.devcontainer/bin/tool`) that call `mise x -- tool` — the wrapper precedes mise install paths in PATH, so without resolution the wrapper is found again instead of the real binary. - Add a Unix e2e regression test (`test_exec_shim_recursion`) mirroring the existing Windows test ## Test plan - [x] `cargo check` passes - [x] `cargo test` — all 481 tests pass - [x] All pre-commit lints pass (shfmt, shellcheck, cargo-fmt, etc.) - [x] Tested on Coder workspace: `coder ssh jdx.devcontainer -- 'claude --version'` returns `2.1.50 (Claude Code)` immediately (previously hung in infinite recursion) - [ ] `./e2e/run_test cli/test_exec_shim_recursion` passes - [ ] Existing exec e2e tests still pass (`test_exec_chdir`, `test_exec_latest`, `test_shims_fallback`) 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

@jdx

### 🚀 Features - **(install)** auto-lock all platforms after tool installation by @jdx in [jdx#8277](jdx#8277) ### 🐛 Bug Fixes - **(config)** respect --yes flag for config trust prompts by @jdx in [jdx#8288](jdx#8288) - **(exec)** strip shims from PATH on Unix to prevent infinite recursion by @jdx in [jdx#8276](jdx#8276) - **(install)** validate --locked before --dry-run short-circuit by @altendky in [jdx#8290](jdx#8290) - **(release)** refresh PATH after mise up in release-plz by @jdx in [jdx#8292](jdx#8292) - **(schema)** replace unevaluatedProperties with additionalProperties by @jdx in [jdx#8285](jdx#8285) - **(task)** avoid duplicated stderr on task failure in replacing mode by @jdx in [jdx#8275](jdx#8275) - **(task)** use process groups to kill child process trees on Unix by @jdx in [jdx#8279](jdx#8279) - **(task)** run depends_post tasks even when parent task fails by @jdx in [jdx#8274](jdx#8274) - **(task)** suggest similar commands when mistyping a CLI subcommand by @jdx in [jdx#8286](jdx#8286) - **(task)** execute monorepo subdirectory prepare steps from root by @jdx in [jdx#8291](jdx#8291) - **(upgrade)** don't force-reinstall already installed versions by @jdx in [jdx#8282](jdx#8282) - **(watch)** restore terminal state after watchexec exits by @jdx in [jdx#8273](jdx#8273) ### 📚 Documentation - clarify that MISE_CEILING_PATHS excludes the ceiling directory itself by @jdx in [jdx#8283](jdx#8283) ### Chore - replace gen-release-notes script with communique by @jdx in [jdx#8289](jdx#8289) ### New Contributors - @altendky made their first contribution in [jdx#8290](jdx#8290) ## 📦 Aqua Registry Updates #### New Packages (4) - [`Skarlso/crd-to-sample-yaml`](https://github.com/Skarlso/crd-to-sample-yaml) - [`kunobi-ninja/kunobi-releases`](https://github.com/kunobi-ninja/kunobi-releases) - [`swanysimon/markdownlint-rs`](https://github.com/swanysimon/markdownlint-rs) - [`tmux/tmux-builds`](https://github.com/tmux/tmux-builds) #### Updated Packages (2) - [`firecow/gitlab-ci-local`](https://github.com/firecow/gitlab-ci-local) - [`k1LoW/runn`](https://github.com/k1LoW/runn)

…x#8342) ## Summary - Remove the `ts.which_bin()` pre-resolution from `Exec::run()` that was introduced in jdx#8276 - This pre-resolution bypassed PATH entirely for mise-managed tools, breaking `_.python.venv` and similar configs where a virtualenv binary should take precedence over the mise-managed install - The shim stripping in `exec_program` (also from jdx#8276) already prevents shim recursion, which was the primary problem that PR solved ## Root Cause PR jdx#8276 added two fixes for infinite recursion: 1. Strip shims from PATH in `exec_program` — prevents shim recursion (**kept**) 2. Pre-resolve bare program names via `which_bin` — prevents wrapper script recursion (**removed**) Fix #2 resolves e.g. `python` → `~/.local/share/mise/installs/python/3.9.1/bin/python` before PATH is computed. Since the result is an absolute path, `exec_program` skips PATH lookup entirely, so `.venv/bin/python` (added by `_.python.venv`) is never found. Fixes jdx#8340 ## Test plan - [x] `cargo check` — no warnings - [x] `mise run test:e2e test_exec` — all exec e2e tests pass - [x] `test_exec_shim_recursion` still passes (shim stripping in `exec_program` handles this) 🤖 Generated with [Claude Code](https://claude.com/claude-code)  --- > [!NOTE] > **Medium Risk** > Changes how `mise x` resolves executables, which can alter which binary runs in edge PATH setups. Added E2E tests reduce regression risk, but behavior changes could impact users relying on the old forced mise-path resolution. > > **Overview** > `mise x` no longer pre-resolves bare command names to a mise-managed tool path via `Toolset::which_bin` in `Exec::run`, allowing normal PATH lookup to pick up higher-precedence overrides like virtualenv binaries. > > Adds two E2E regression tests: one ensuring PATH-prepended “venv” bins win over mise installs during `mise x`, and another ensuring a PATH-preceding wrapper that re-invokes `mise x` doesn’t hang (infinite recursion) and still executes the real tool. > > <sup>Written by [Cursor Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit b241665. This will update automatically on new commits. Configure [here](https://cursor.com/dashboard?tab=bugbot).</sup>  --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Copilot AI review requested due to automatic review settings February 21, 2026 14:54

Copilot started reviewing on behalf of jdx February 21, 2026 14:55 View session

gemini-code-assist bot reviewed Feb 21, 2026

View reviewed changes

Copilot AI reviewed Feb 21, 2026

View reviewed changes

cursor bot reviewed Feb 21, 2026

View reviewed changes

src/cli/exec.rs Show resolved Hide resolved

jdx enabled auto-merge (squash) February 21, 2026 15:22

jdx and others added 2 commits February 21, 2026 10:07

jdx force-pushed the fix/unix-exec-shim-recursion branch from 683b3f9 to f75c105 Compare February 21, 2026 16:08

jdx force-pushed the fix/unix-exec-shim-recursion branch from f75c105 to d53b221 Compare February 21, 2026 16:29

jdx merged commit ab63c6b into main Feb 21, 2026
35 checks passed

jdx deleted the fix/unix-exec-shim-recursion branch February 21, 2026 16:42

mise-en-dev mentioned this pull request Feb 21, 2026

chore: release 2026.2.18 #8259

Merged

BrewTestBot mentioned this pull request Feb 21, 2026

mise 2026.2.18 Homebrew/homebrew-core#268760

Merged

jdx mentioned this pull request Feb 25, 2026

fix(exec): respect PATH order for virtualenv resolution in mise x #8342

Merged

3 tasks

BrewTestBot mentioned this pull request Feb 26, 2026

mise 2026.2.21 Homebrew/homebrew-core#269534

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(exec): strip shims from PATH on Unix to prevent infinite recursion#8276

fix(exec): strip shims from PATH on Unix to prevent infinite recursion#8276
jdx merged 3 commits intomainfrom
fix/unix-exec-shim-recursion

jdx commented Feb 21, 2026 •

edited by cursor bot

Loading

Uh oh!

gemini-code-assist bot commented Feb 21, 2026

Uh oh!

gemini-code-assist bot left a comment

Uh oh!

gemini-code-assist bot Feb 21, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Feb 21, 2026

Uh oh!

Copilot AI Feb 21, 2026

Uh oh!

Copilot AI Feb 21, 2026

Uh oh!

cursor bot left a comment

Uh oh!

Uh oh!

github-actions bot commented Feb 21, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

-            std::env::join_paths(&filtered).unwrap()
+            match std::env::join_paths(&filtered) {
+                Ok(joined) => joined,
+                Err(err) => {
+                    eprintln!(
+                        "warning: failed to rebuild PATH without shims ({}); using original PATH",
+                        err
+                    );
+                    OsString::from(path_val)
+                }
+            }

Uh oh!

Conversation

jdx commented Feb 21, 2026 • edited by cursor bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Uh oh!

gemini-code-assist bot commented Feb 21, 2026

Summary of Changes

Highlights

Footnotes

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Feb 21, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Feb 21, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 21, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 21, 2026

Choose a reason for hiding this comment

Uh oh!

cursor bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

github-actions bot commented Feb 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Hyperfine Performance

mise x -- echo

mise env

mise hook-env

mise ls

xtasks/test/perf

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

jdx commented Feb 21, 2026 •

edited by cursor bot

Loading

github-actions bot commented Feb 21, 2026 •

edited

Loading

`mise x -- echo`

`mise env`

`mise hook-env`

`mise ls`

`xtasks/test/perf`