fix(tmux): pass -u on the remaining attach paths

[Jonesxq]

What problem does this solve?

When agent-deck starts under a supervisor without a UTF-8 locale, four tmux attach clients outside the web terminal bridge can still start in non-UTF-8 mode and replace Unicode glyphs with _. Fixes #1781 and completes the follow-up to #1779.

Why this change

Pass tmux's global -u option before attach-session in the normal and read-only PTY attaches, ControlPipe, and KeySender. This makes each attach client use UTF-8 independently of its inherited locale. An AST regression test now rejects any production attach spawn that omits -u or places it after the subcommand; the existing exact-argv tests and comments were updated with the same contract.

User impact

Terminal output on the remaining PTY and control-mode attach paths preserves Unicode when agent-deck is launched by locale-light environments such as launchd or systemd. Environments that already provide a UTF-8 locale keep the same behavior. The change adds no process, handshake, poll, or output-processing step.

Evidence

The merged #1779 records real output for the same tmux mechanism: without -u, raw bytes for ⏵⏵ were 5f 5f (__); with -u, they were e2 8f b5 e2 8f b5 (⏵⏵). Issue #1781 identifies the four remaining attach sites that share that locale exposure.

Test-first baseline / revert-check:

$ go test internal/tmux/attach_utf8_argv_lint_test.go internal/tmux/tmux_exec_lint_helpers_test.go internal/tmux/tmux_exec_lint_test.go -run '^TestTmuxAttachSpawnsForceUTF8$' -count=1 -v
4 tmux attach spawn(s) do not force UTF-8:
  internal/tmux/controlpipe.go:113
  internal/tmux/keysender.go:101
  internal/tmux/pty.go:733
  internal/tmux/pty.go:737
FAIL

After adding -u, the same command passes.

Final checks:

$ go test internal/tmux/attach_utf8_argv_lint_test.go internal/tmux/control_mode_argv_lint_test.go internal/tmux/tmux_exec_lint_helpers_test.go internal/tmux/tmux_exec_lint_test.go -count=1
ok      command-line-arguments

$ GOOS=linux GOARCH=amd64 CGO_ENABLED=0 go test -c -o <temporary-path>/tmux.test ./internal/tmux
exit 0

$ GOOS=linux GOARCH=amd64 CGO_ENABLED=0 go vet ./internal/tmux
exit 0

$ gofmt -l internal/tmux/attach_utf8_argv_lint_test.go internal/tmux/control_mode_argv_lint_test.go internal/tmux/controlpipe.go internal/tmux/keysender.go internal/tmux/pty.go internal/tmux/pty_socket_test.go
(no output)

$ git diff --check
exit 0

The full sandboxed go test ./... suite, runtime tmux reproduction, and before/after performance timing were not run: the local WSL environment has no Go toolchain, and runtime WSL execution was unavailable for this run. The performance-sensitive scope is limited to one extra global option on existing one-time attach process launches.

AI disclosure

• Human-written
• AI-assisted (I directed and reviewed it)
• AI-authored (a model wrote most of it)

Model(s), if AI helped: gpt-5-codex

Prompt / session log (optional): Not provided.

What actually bothered you

My human asked: "完成一下 #1781" ("Complete issue #1781.")

Checklist

• Targeted diff: one problem, no unrelated changes
• Tests added or updated for new behavior
• Test suite passes sandboxed: HOME=$(mktemp -d) XDG_CONFIG_HOME= XDG_DATA_HOME= XDG_CACHE_HOME= go test ./... - not run locally; see Evidence
• If this touches a hot path (list, status, session output, startup, tmux layer): before/after timing evidence included - not measured; this only adds a global option to existing attach launches
• CHANGELOG.md untouched (entries are added at landing)
• AI-assisted? Disclosed above, with validation evidence, and I can answer questions about the code
• "Allow edits from maintainers" is enabled

Summary by CodeRabbit

• Bug Fixes

  • Improved tmux session and control-mode attachments to consistently use UTF-8 mode.
  • Ensured headless tmux control behavior avoids opening /dev/tty for detached callers.
  • Updated attach command construction for both interactive and read-only modes for more reliable non-ASCII handling.

• Tests

  • Added lint-style checks to fail if required tmux attach/control flags or arguments are missing.
  • Updated tmux command-argument expectations in socket-related tests to include UTF-8 mode.

[github-actions]

👋 Thanks for the contribution — intake looks complete.

Your PR body carries everything the maintainer's validation pipeline reads first: the problem, the reasoning, the human intent behind it, and an AI-disclosure. It will be applied, built, and tested against main, and you'll get a structured result within about a day. Merges are always human.

_{gate marker read: ai=authored model=gpt-5-codex intent=yes}

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 2469d04c-4f3b-4691-a8f5-8d593afb6428

📥 Commits

Reviewing files that changed from the base of the PR and between 1a524fe and 0bb8d4e.

📒 Files selected for processing (3)

• internal/tmux/controlpipe.go
• internal/tmux/keysender.go
• internal/tmux/pty.go

🚧 Files skipped from review as they are similar to previous changes (3)

• internal/tmux/keysender.go
• internal/tmux/controlpipe.go
• internal/tmux/pty.go

---

📝 Walkthrough

Walkthrough

tmux session and control-client attachment commands now pass the global -u flag before attach-session. Socket argv tests and AST-based lint coverage enforce the command shape across the remaining attach paths.

Changes

tmux UTF-8 attachment handling

Layer / File(s)	Summary
Session attach argv updates internal/tmux/pty.go, internal/tmux/pty_socket_test.go	Read-write and read-only session attachments include -u, with socket and empty-socket argv expectations updated.
Control-client attach argv updates internal/tmux/controlpipe.go, internal/tmux/keysender.go	ControlPipe and KeySender use -C -u attach-session, with matching headless-process documentation and error text.
Attach command validation internal/tmux/attach_utf8_argv_lint_test.go, internal/tmux/control_mode_argv_lint_test.go	Repository linting detects attach-session spawns without -u, and control-mode guidance documents the required UTF-8 attach form.

Estimated code review effort: 2 (Simple) | ~15 minutes

Possibly related PRs

• asheshgoplani/agent-deck#1736: Updates tmux control-client spawning and argv validation in the same package.
• asheshgoplani/agent-deck#1779: Adds the global -u flag to other tmux attach-session invocations.

Suggested reviewers: asheshgoplani

🚥 Pre-merge checks | ✅ 7

✅ Passed checks (7 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title uses conventional commits format and accurately describes the tmux attach-path UTF-8 fix.
Linked Issues check	✅ Passed	The PR updates all remaining attach paths listed in #1781 and adds regression coverage for the UTF-8 flag.
Out of Scope Changes check	✅ Passed	The test and comment updates are directly tied to enforcing -u on tmux attach paths and are in scope.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 70.00%.
Remote_parity	✅ Passed	Not applicable: the PR only changes internal/tmux files; no internal/ui/ or cmd/agent-deck code was modified, so RemoteSession parity isn’t in scope.
Test_coverage_per_surface	✅ Passed	Affected surfaces are covered: pty, controlpipe, and keysender now have regression/argv tests, and web already has explicit UTF-8 attach tests. No remote change.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@internal/tmux/pty.go`:
- Around line 732-737: Document the controlling-TTY contracts near the
attach-client symbols: in internal/tmux/pty.go lines 732-737, explain that the
normal PTY attach requires a controlling terminal and describe its failure
behavior per `#1114`; in internal/tmux/controlpipe.go line 36, document the
control-pipe client’s distinct controlling-TTY behavior; and in
internal/tmux/keysender.go line 46, document the KeySender client’s
controlling-TTY behavior. Update comments only, preserving the existing command
implementations.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 4ccb7d2f-ae5f-4bcd-81c4-9d34ea4f50d9

📥 Commits

Reviewing files that changed from the base of the PR and between 580e772 and 1a524fe.

📒 Files selected for processing (6)

• internal/tmux/attach_utf8_argv_lint_test.go
• internal/tmux/control_mode_argv_lint_test.go
• internal/tmux/controlpipe.go
• internal/tmux/keysender.go
• internal/tmux/pty.go
• internal/tmux/pty_socket_test.go