fix: silence /dev/tty redirect errors when no tty is attached

[kevmi]

Problem

On Windows / Git Bash the Stop hook runs without a controlling terminal, so every turn prints:

/c/Users/<user>/.claude/plugins/cache/claude-code-warp/warp/2.2.0/scripts/emit-terminal-sequence.sh: line 82: /dev/tty: No such device or address

The 2>/dev/null was meant to suppress this, but shell redirections are applied left to right. In > /dev/tty 2>/dev/null, the shell opens /dev/tty while stderr is still the inherited one, so the failed-open message escapes before 2>/dev/null takes effect.

Fix

Swap the order to 2>/dev/null > /dev/tty so stderr is muted before the open is attempted.

$ ( printf 'x' > /dev/tty 2>/dev/null ) ; echo "exit=$?"
bash: line 1: /dev/tty: No such device or address
exit=1

$ ( printf 'x' 2>/dev/null > /dev/tty ) ; echo "exit=$?"
exit=1

Exit status is unchanged, so the if in emit_terminal_sequence still falls through to the JSON path correctly.

Verification

All three branches of emit_terminal_sequence exercised with no tty attached:

CLAUDE_CODE_VERSION	Output	stderr
unset (unknown)	{"terminalSequence":"HELLO"}	clean
2.5.0 (new)	{"terminalSequence":"HELLO"}	clean
2.0.0 (old, tty path)	none, exit 0	clean

bash -n passes on both files. Same one-line change applied to legacy/warp-notify.sh, which has the identical pattern.

🤖 Generated with Claude Code

https://claude.ai/code/session_01PWTCPGfRHRHJxiuXUjxP6X

[warp-factories]

Overview

Correctly fixes the redirection-order bug: swapping to 2>/dev/null > /dev/tty mutes stderr before the /dev/tty open is attempted, so the "No such device or address" message no longer escapes when no controlling terminal is attached, with exit status unchanged (verified below). No blocking issues found.

Concerns

• Neither changed script has a regression test. plugins/warp/tests/test-hooks.sh exercises emit_terminal_sequence but never asserts stderr is empty when /dev/tty open fails, so this exact bug (and any regression of it) would not be caught by CI. A test that runs the "unknown version" fallback (or the old-version branch) with stderr captured and asserts it's empty would cover this.

Verdict

Checks: build n/a (shell scripts), tests pass (existing suite unaffected by this change), CI not yet run (fork PR pending workflow approval), visual proof n/a

Found: 0 critical, 1 important, 0 suggestions, 0 nits

Reproduced the claimed behavior in a detached (no controlling terminal) shell: old ordering leaks bash: line 1: /dev/tty: No such device or address to stderr with exit 1; new ordering produces empty stderr with the same exit 1.