fix(cli): eliminate extra Ctrl+C needed to exit local spawn session

[aulorbe]

Summary

Exiting a local spawn session with Ctrl+C requires pressing it three times instead of two. After the agent process exits on the second Ctrl+C, the parent process hangs briefly — long enough that the user presses Ctrl+C again.

Root cause

Two things add delay between the agent exiting and the process actually terminating:

1. spawnInteractive runs stty sane — a synchronous nodeSpawnSync call that blocks right after the agent exits, during the exact moment the user expects the process to be dying.

2. runOrchestration continues post-session cleanup — reconnect logic, tunnel teardown, and history pull all run before process.exit(), even on SIGINT where the user just wants out.

Fix

• ssh.ts: On exit code 130 or SIGINT signal, skip terminal reset (stty sane, ANSI escape writes) and return immediately.
• orchestrate.ts: On exit code 130, call process.exit(130) immediately after the interactive session returns, skipping all post-session cleanup.

Test plan

• bunx @biomejs/biome check src/shared/ssh.ts src/shared/orchestrate.ts — zero errors
• Full test suite — 2201 pass, 3 pre-existing flaky failures unrelated to this change
• Run spawn cursor local, press Ctrl+C twice — should exit cleanly without needing a third press

[aulorbe]

• fix(cli): eliminate extra Ctrl+C needed to exit local spawn session #3430 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[aulorbe]

Summary

Exiting a local spawn session with Ctrl+C requires pressing it three times instead of two. After the agent process exits on the second Ctrl+C, the parent process hangs briefly — long enough that the user presses Ctrl+C again.

Root cause

Two things add delay between the agent exiting and the process actually terminating:

1. spawnInteractive runs stty sane — a synchronous nodeSpawnSync call that blocks right after the agent exits, during the exact moment the user expects the process to be dying.

2. runOrchestration continues post-session cleanup — reconnect logic, tunnel teardown, and history pull all run before process.exit(), even on SIGINT where the user just wants out.

Fix

• ssh.ts: On exit code 130 or SIGINT signal, skip terminal reset (stty sane, ANSI escape writes) and return immediately.
• orchestrate.ts: On exit code 130, call process.exit(130) immediately after the interactive session returns, skipping all post-session cleanup.

Test plan

• bunx @biomejs/biome check src/shared/ssh.ts src/shared/orchestrate.ts — zero errors
• Full test suite — 2201 pass, 3 pre-existing flaky failures unrelated to this change
• Run spawn cursor local, press Ctrl+C twice — should exit cleanly without needing a third press

[aulorbe]

• fix(cli): eliminate extra Ctrl+C needed to exit local spawn session #3430 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[aulorbe]

Summary

Exiting a local spawn session with Ctrl+C requires pressing it three times instead of two. After the agent process exits on the second Ctrl+C, the parent process hangs briefly — long enough that the user presses Ctrl+C again.

Root cause

Two things add delay between the agent exiting and the process actually terminating:

1. spawnInteractive runs stty sane — a synchronous nodeSpawnSync call that blocks right after the agent exits, during the exact moment the user expects the process to be dying.

2. runOrchestration continues post-session cleanup — reconnect logic, tunnel teardown, and history pull all run before process.exit(), even on SIGINT where the user just wants out.

Fix

• ssh.ts: On exit code 130 or SIGINT signal, skip terminal reset (stty sane, ANSI escape writes) and return immediately.
• orchestrate.ts: On exit code 130, call process.exit(130) immediately after the interactive session returns, skipping all post-session cleanup.

Test plan

• bunx @biomejs/biome check src/shared/ssh.ts src/shared/orchestrate.ts — zero errors
• Full test suite — 2201 pass, 3 pre-existing flaky failures unrelated to this change
• Run spawn cursor local, press Ctrl+C twice — should exit cleanly without needing a third press

[aulorbe]

• fix(cli): eliminate extra Ctrl+C needed to exit local spawn session #3430 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[aulorbe]

Summary

Exiting a local spawn session with Ctrl+C requires pressing it three times instead of two. After the agent process exits on the second Ctrl+C, the parent process hangs briefly — long enough that the user presses Ctrl+C again.

Root cause

Two things add delay between the agent exiting and the process actually terminating:

1. spawnInteractive runs stty sane — a synchronous nodeSpawnSync call that blocks right after the agent exits, during the exact moment the user expects the process to be dying.

2. runOrchestration continues post-session cleanup — reconnect logic, tunnel teardown, and history pull all run before process.exit(), even on SIGINT where the user just wants out.

Fix

• ssh.ts: On exit code 130 or SIGINT signal, skip terminal reset (stty sane, ANSI escape writes) and return immediately.
• orchestrate.ts: On exit code 130, call process.exit(130) immediately after the interactive session returns, skipping all post-session cleanup.

Test plan

• bunx @biomejs/biome check src/shared/ssh.ts src/shared/orchestrate.ts — zero errors
• Full test suite — 2201 pass, 3 pre-existing flaky failures unrelated to this change
• Run spawn cursor local, press Ctrl+C twice — should exit cleanly without needing a third press

[aulorbe]

• fix(cli): eliminate extra Ctrl+C needed to exit local spawn session #3430 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[aulorbe]

Summary

Exiting a local spawn session with Ctrl+C requires pressing it three times instead of two. After the agent process exits on the second Ctrl+C, the parent process hangs briefly — long enough that the user presses Ctrl+C again.

Root cause

Two things add delay between the agent exiting and the process actually terminating:

1. spawnInteractive runs stty sane — a synchronous nodeSpawnSync call that blocks right after the agent exits, during the exact moment the user expects the process to be dying.

2. runOrchestration continues post-session cleanup — reconnect logic, tunnel teardown, and history pull all run before process.exit(), even on SIGINT where the user just wants out.

Fix

• ssh.ts: On exit code 130 or SIGINT signal, skip terminal reset (stty sane, ANSI escape writes) and return immediately.
• orchestrate.ts: On exit code 130, call process.exit(130) immediately after the interactive session returns, skipping all post-session cleanup.

Test plan

• bunx @biomejs/biome check src/shared/ssh.ts src/shared/orchestrate.ts — zero errors
• Full test suite — 2201 pass, 3 pre-existing flaky failures unrelated to this change
• Run spawn cursor local, press Ctrl+C twice — should exit cleanly without needing a third press

[aulorbe]

• fix(cli): eliminate extra Ctrl+C needed to exit local spawn session #3430 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[aulorbe]

Summary

Exiting a local spawn session with Ctrl+C requires pressing it three times instead of two. After the agent process exits on the second Ctrl+C, the parent process hangs briefly — long enough that the user presses Ctrl+C again.

Root cause

Two things add delay between the agent exiting and the process actually terminating:

1. spawnInteractive runs stty sane — a synchronous nodeSpawnSync call that blocks right after the agent exits, during the exact moment the user expects the process to be dying.

2. runOrchestration continues post-session cleanup — reconnect logic, tunnel teardown, and history pull all run before process.exit(), even on SIGINT where the user just wants out.

Fix

• ssh.ts: On exit code 130 or SIGINT signal, skip terminal reset (stty sane, ANSI escape writes) and return immediately.
• orchestrate.ts: On exit code 130, call process.exit(130) immediately after the interactive session returns, skipping all post-session cleanup.

Test plan

• bunx @biomejs/biome check src/shared/ssh.ts src/shared/orchestrate.ts — zero errors
• Full test suite — 2201 pass, 3 pre-existing flaky failures unrelated to this change
• Run spawn cursor local, press Ctrl+C twice — should exit cleanly without needing a third press

[aulorbe]

• fix(cli): eliminate extra Ctrl+C needed to exit local spawn session #3430 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[aulorbe]

Summary

Exiting a local spawn session with Ctrl+C requires pressing it three times instead of two. After the agent process exits on the second Ctrl+C, the parent process hangs briefly — long enough that the user presses Ctrl+C again.

Root cause

Two things add delay between the agent exiting and the process actually terminating:

1. spawnInteractive runs stty sane — a synchronous nodeSpawnSync call that blocks right after the agent exits, during the exact moment the user expects the process to be dying.

2. runOrchestration continues post-session cleanup — reconnect logic, tunnel teardown, and history pull all run before process.exit(), even on SIGINT where the user just wants out.

Fix

• ssh.ts: On exit code 130 or SIGINT signal, skip terminal reset (stty sane, ANSI escape writes) and return immediately.
• orchestrate.ts: On exit code 130, call process.exit(130) immediately after the interactive session returns, skipping all post-session cleanup.

Test plan

• bunx @biomejs/biome check src/shared/ssh.ts src/shared/orchestrate.ts — zero errors
• Full test suite — 2201 pass, 3 pre-existing flaky failures unrelated to this change
• Run spawn cursor local, press Ctrl+C twice — should exit cleanly without needing a third press

[aulorbe]

• fix(cli): eliminate extra Ctrl+C needed to exit local spawn session #3430 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[aulorbe]

Summary

Exiting a local spawn session with Ctrl+C requires pressing it three times instead of two. After the agent process exits on the second Ctrl+C, the parent process hangs briefly — long enough that the user presses Ctrl+C again.

Root cause

Two things add delay between the agent exiting and the process actually terminating:

1. spawnInteractive runs stty sane — a synchronous nodeSpawnSync call that blocks right after the agent exits, during the exact moment the user expects the process to be dying.

2. runOrchestration continues post-session cleanup — reconnect logic, tunnel teardown, and history pull all run before process.exit(), even on SIGINT where the user just wants out.

Fix

• ssh.ts: On exit code 130 or SIGINT signal, skip terminal reset (stty sane, ANSI escape writes) and return immediately.
• orchestrate.ts: On exit code 130, call process.exit(130) immediately after the interactive session returns, skipping all post-session cleanup.

Test plan

• bunx @biomejs/biome check src/shared/ssh.ts src/shared/orchestrate.ts — zero errors
• Full test suite — 2201 pass, 3 pre-existing flaky failures unrelated to this change
• Run spawn cursor local, press Ctrl+C twice — should exit cleanly without needing a third press

[aulorbe]

• fix(cli): eliminate extra Ctrl+C needed to exit local spawn session #3430 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.