fix(daemon): resolve the shell in-process and cache successful resolution

[mu-hashmi]

TL;DR: the daemon resolves the shell by reading /etc/shells in-process (no subprocess), caches a successful resolution for its lifetime, and can never return an empty shell.

Description

Every toolbox exec, code-run, session create, and PTY spawn resolved the shell by spawning sh -c \"grep '^[^#]' /etc/shells\" - an extra fork+execve on the daemon's five hottest call sites, and the first thing to fail under degraded filesystem conditions (during the fd-exhaustion incident the probe died before the real command, silently downgrading users to sh).

Now GetShell():

• reads /etc/shells directly and preserves the exact preference order (/usr/bin/zsh > /bin/zsh > /usr/bin/bash > /bin/bash > $SHELL if non-empty > sh);
• caches a successful resolution behind a mutex (daemon serves concurrent requests);
• never caches failure - an unreadable file falls back for that call only ($SHELL, then sh) and retries next call;
• never returns empty: $SHELL set-but-empty falls through to sh (review-caught edge; regression-tested).

Behavior note: a shell installed after daemon start is not picked up until restart (previously each exec re-probed). Accepted: daemon lifetime == sandbox session.

Review guide

1. apps/daemon/pkg/common/get_shell.go - the whole change.
2. get_shell_test.go - 8 cases document the contract.

Lowest-risk PR of the six; no generated files, no API surface.

Commit map

Commit	What	Why it exists
c9ee8ef	in-process resolution + success-only cache	implementation (plan)
dac41ca	never return empty when /etc/shells unreadable	cubic review (valid)

Related PRs

Independent. Siblings from the same incident: #5001, #5002.

Validation

• 8 unit tests: preference order, comment filtering, $SHELL fallback, missing-file fallback, empty-$SHELL regression, success cached (file mutated after first call - answer unchanged), failure not cached (file appears later - next call picks it up). Runs on darwin and in CI.
• gofmt clean, golangci-lint v2.6.2 zero issues.
• All CI checks green on dac41ca.

Closes #4997

[nx-cloud]

View your CI Pipeline Execution ↗ for commit dac41ca

Command	Status	Duration	Result
nx e2e:cleanup daytona-e2e	✅ Succeeded	<1s	View ↗
nx run-many --target=test:e2e --all --nxBail=true	✅ Succeeded	6m 39s	View ↗
nx e2e daytona-e2e	✅ Succeeded	55s	View ↗
nx run-many --target=build --projects=api,runne...	✅ Succeeded	18s	View ↗

💡 Verify your cache is correct by running tasks in a sandbox. Read docs ↗

---

☁️ Nx Cloud last updated this comment at 2026-06-11 04:10:25 UTC

[Copilot]

Pull request overview

This PR updates the daemon’s shell selection logic to avoid spawning a subprocess on hot paths by reading /etc/shells in-process and caching the first successful resolution for the daemon lifetime (while not caching failures to allow later retries).

Changes:

• Replaced sh -c "grep ... /etc/shells" probing with in-process /etc/shells parsing and preference-based selection.
• Added mutex-guarded caching so successful shell resolution is computed once and reused across requests.
• Added unit tests covering preference order, comment filtering, env/missing-file fallback, and caching semantics.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
apps/daemon/pkg/common/get_shell.go	In-process /etc/shells parsing, fallback behavior, and mutex-guarded cache for resolved shell.
apps/daemon/pkg/common/get_shell_test.go	New unit tests validating shell resolution order, comment filtering, and caching/retry behavior.

Comments suppressed due to low confidence (1)

apps/daemon/pkg/common/get_shell.go:77

• os.LookupEnv("SHELL") returns shellSet=true even when the value is an empty string. In that case shellFallback() currently returns "", which can later cause exec.Command(shell) (e.g. in SpawnTTY) to fail. Treat empty $SHELL as unset and fall back to sh.

func shellFallback() string {
	if shellEnv, shellSet := os.LookupEnv("SHELL"); shellSet {
		return shellEnv
	}

	return "sh"
}

---

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

[cubic-dev-ai]

1 issue found across 2 files

<sub>Reply with feedback, questions, or to request a fix.

Fix all with cubic | Re-trigger cubic</sub>