test: deflake python venv go backend e2e

[jdx]

Summary

• increase the deadlock regression test watchdog from 60s to a configurable default of 180s
• run Go cache permission cleanup through an EXIT trap so timed-out runs do not leave read-only files behind

Context

The failing CI run timed out in core/test_python_venv_with_go_backend_deadlock_slow while cold-installing go:github.com/charmbracelet/gum@0.17.0. The test still catches real hangs, but gives ordinary Go module downloads/builds more room.

Verification

• bash -n e2e/core/test_python_venv_with_go_backend_deadlock_slow
• shellcheck e2e/core/test_python_venv_with_go_backend_deadlock_slow

Full targeted e2e was attempted locally but did not reach the test because the local build failed first: aws-lc-sys/bindgen could not find libclang.

This PR was generated by an AI coding assistant.

---

Note

Low Risk
Only changes an e2e shell test timeout and cleanup; no production code paths.

Overview
Deflakes the slow Python-venv + Go-backend deadlock e2e by giving mise install more time on cold CI (default 180s via MISE_E2E_INSTALL_TIMEOUT, was a fixed 60s).

Go module cache cleanup (chmod -R +w ~/go) now runs on EXIT so a timeout kill still leaves the cache writable for teardown instead of only running after a successful run.

^{Reviewed by Cursor Bugbot for commit a7f2f4a. Bugbot is set up for automated code reviews on this repo. Configure here.}

[gemini-code-assist]

Code Review

This pull request refactors the cleanup process in the test_python_venv_with_go_backend_deadlock_slow end-to-end test by introducing a cleanup_go_cache function registered as an EXIT trap and parameterizing the installation timeout. The review feedback suggests making the cleanup function more robust by using '$HOME/go' instead of ~/go and appending '|| true' to the chmod command to prevent potential test failures under set -e.

[gemini-code-assist]

Since set -euo pipefail is enabled, any failure in the chmod command (for example, if files are concurrently modified or if there are permission restrictions) will cause the cleanup function to exit with an error. In an EXIT trap, this can result in the entire script exiting with a non-zero status, causing false-positive test failures in CI.

Additionally, using "$HOME/go" is more robust and standard in shell scripts than the unquoted tilde ~/go.

Consider appending || true to the chmod command and using "$HOME/go".

cleanup_go_cache() {
  # Required to properly cleanup as go installs read-only sources
  if [ -d "$HOME/go" ]; then
    chmod -R +w "$HOME/go" || true
  fi
}

[greptile-apps]

Greptile Summary

This PR deflakes the Python venv + Go backend deadlock e2e test by giving cold Go module downloads more room to finish and ensuring the go-cache write-permission cleanup runs even when the test is cut short by a timeout.

• The hardcoded 60 s timeout is replaced with a configurable MISE_E2E_INSTALL_TIMEOUT variable (default 180 s), matching the observed CI failure where a cold gum@0.17.0 build exceeded the old limit.
• The trailing chmod -R +w ~/go that only ran on success is promoted to an EXIT trap, so read-only Go source files are cleaned up regardless of how the script exits (timeout, assertion failure, or success).

Confidence Score: 5/5

This is a small, targeted e2e test fix with no production code changes; the two modifications are straightforward and correct.

The EXIT trap correctly fires when set -e terminates the shell after timeout kills mise install, so the go-cache cleanup now runs in all exit paths. The timeout variable is safely quoted and defaults to a reasonable value. No logic from the original test is altered.

No files require special attention.

Important Files Changed

Filename	Overview
e2e/core/test_python_venv_with_go_backend_deadlock_slow	Moves go cache cleanup to an EXIT trap and makes the install timeout configurable via MISE_E2E_INSTALL_TIMEOUT (default 180s, up from 60s)

_{Reviews (1): Last reviewed commit: "test: deflake python venv go backend e2e" | Re-trigger Greptile}

[github-actions]

Hyperfine Performance

mise x -- echo

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
mise-2026.5.16 x -- echo	22.5 ± 0.4	21.7	24.2	1.00 ± 0.03
mise x -- echo	22.4 ± 0.4	21.6	25.1	1.00

mise env

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
mise-2026.5.16 env	22.1 ± 0.9	21.2	29.3	1.01 ± 0.04
mise env	21.9 ± 0.4	21.1	26.0	1.00

mise hook-env

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
mise-2026.5.16 hook-env	22.8 ± 0.4	22.1	24.4	1.00
mise hook-env	23.0 ± 1.1	22.1	38.7	1.01 ± 0.05

mise ls

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
mise-2026.5.16 ls	19.3 ± 0.4	18.6	24.3	1.01 ± 0.02
mise ls	19.1 ± 0.3	18.5	21.0	1.00

xtasks/test/perf

Command	mise-2026.5.16	mise	Variance
install (cached)	171ms	167ms	+2%
ls (cached)	86ms	84ms	+2%
bin-paths (cached)	89ms	86ms	+3%
task-ls (cached)	157ms	155ms	+1%