[TooonyChen]

Summary

Until now the server monitor sent the app's only keepalive. With the host's exec channel off, through a jump host, or after the stats script hung, a quiet session sent nothing, and nothing noticed when a connection died without a word. This PR adds a keepalive to every SSH connection that does not depend on the monitor.

• sshx.KeepAlive(ctx, client, KeepAliveConfig{Interval, Timeout, MaxMissed}) sends keepalive@openssh.com with want-reply every interval. Any answer counts, success or failure. Each wait has its own timeout, run in a goroutine, because SendRequest has no timeout. While one request is still waiting, no new one is sent: global requests wait for their answers one at a time anyway. After MaxMissed (3) misses in a row it closes the client. The loop ends when ctx ends or the client closes (client.Wait).
• sshx.Options.KeepAlive: Dial starts the keepalive on the client it returns, and dialJump gives the same config to each jump host's client. A jump host whose keepalive closes it takes down the host's connection through it. With an HTTP or SOCKS proxy, the host's own keepalive crosses the proxy.
• When the keepalive closes a client, the shell's channel gets EOF and the terminal's existing OnExit sets statusClosed / errConnectionLost, so the session shows "Connection lost" and Reconnect works. The test checks this path. No new status code was needed.
• Setting keepAliveSeconds (Options › General › Connections, "Send a keepalive every N seconds", 0 to 3600). Each wait times out after min(interval, 15 s).
  • Storage: the field is an int with a sentinel value. Missing or 0 means the default of 30, and the options' 0 ("off") is stored as -1. This follows the repo's existing rule that zero means default (wheelLines, termFontSize, sync.interval). It keeps Settings a comparable value (tests compare with ==, and a *int would compare addresses). It also keeps the sync rule that a setting at its default has no key. The downside is that a hand-written 0 in settings.json means 30. configuration.md says so.
  • Page: General, not Network monitor. The "Network monitor" page is about the Network tab (IP locations, trace route), and General already holds app-wide behavior.
  • Synced: added to syncedSettings, like the other behavior settings.
  • It applies to connections made after the change. The Options intro in features.md now lists it as an exception to "takes effect at once".
• monitor.Run: runOutput now gives up after 10 s. It closes that exec session and keeps the connection, so a stuck df no longer stops the monitor loop.
• internal/sshtest: counts keepalives (KeepAlives), can stop answering them (StopAnswering), and can accept exec requests that never finish (HangExec).
• Docs: features.md (Sessions and tabs, Server monitor, Options › General) and configuration.md (keepAliveSeconds), in both languages.

CHANGELOG.md only takes version sections, so here is a ready bullet for the next release:

• English: Sessions no longer drop when left idle: every connection, and each jump host's, sends an SSH keepalive every 30 seconds even with the server monitor off, and a connection that stops answering shows "Connection lost" so you can reconnect. Change the interval, or turn it off with 0, in Options › General.
• 简体中文：会话空闲时不再被断开：每个连接（以及途经的跳板机）每 30 秒发送一次 SSH 保活消息，服务器监控关闭时也一样；连接没有回应时显示“连接已断开”，可以重新连接。间隔可在“选项 › 常规”里修改，设为 0 关闭。

Closes #161

Testing

New tests:

• internal/app TestKeepAliveIdleSession: a session to sshtest with the monitor off gets keepalives at the interval. After StopAnswering, the session becomes statusClosed / "Connection lost". Fails on main in both parts: 0 keepalives, want 3, and, with that step removed, status "Connected" after the server stopped answering.
• internal/monitor TestRunScriptHangs: with HangExec, the monitor keeps publishing stats. Fails on main: 0 stats in 3 s with the script hanging.
• TestKeepAliveJumpHost (app) and TestDialKeepAliveJumpHost (sshx): keepalives reach both the jump host and the target. In the sshx test, a jump host that stops answering closes the host's client.
• TestKeepAlive and TestKeepAliveEnds (sshx): the interval, closing after 3 misses, and stopping when ctx ends or the client closes. TestKeepAliveOff: the setting at 0 sends none.
• TestKeepAliveOption: the Options control, its 30 default, 0 saved as -1, and back to default (nothing saved). TestKeepAliveSetting (config): the mapping and sync.

The tests use 10 to 50 ms intervals and add about 1.5 s in total.

Checks: go fmt ./... changes nothing, and go build ./..., go vet ./... (also GOOS=windows and GOOS=darwin) and go tool mygo vet pass. go test -race ./... passes except TestArchiveTreeMenu, which also fails in this sandbox on main because it tars the real /tmp. I looked at screenshots of the General page in English and Simplified Chinese.

Not tested: a real NAT, firewall or proxy dropping an idle connection, and a real OpenSSH server. The in-process server stands in for both.

---

Generated by Claude Code

[chatgpt-codex-connector]

Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-09T23:09:12.091207Z	c4113c0	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.