Detect dead connections via keepalive timeout

[EhabY]

_recvAckCheck only fires when there are unacknowledged regular messages. If a connection dies silently while idle, the timeout is never detected.

Add _keepAliveTimeoutCheck: since both sides send keepalives every 5s, receiving no data for 20s means the connection is dead. Called from _sendKeepAlive after each keepalive is sent. This assumes that there is symmetry, if the client sends keepalives then so should the server.

Related to microsoft/vscode-remote-release#11538

[vs-code-engineering]

📬 CODENOTIFY

The following users are being notified based on files changed in this PR:

@bpasero

Matched files:

• src/vs/base/parts/ipc/common/ipc.net.ts
• src/vs/base/parts/ipc/test/node/ipc.net.test.ts

[benoit-nexthop]

We really need those values to be configurable. There are environments where the 5s KeepAliveSendTime or 20s TimeoutTime are not appropriate.

[EhabY]

Agreed, a configurable timeout would be useful. These constants live in the base IPC layer though, so it would need some plumbing to get a setting value down there.

For naming, maybe something under a new prefix like ipc.keepAliveTimeout or connection.keepAliveTimeout? Also this would just affect the client dead-connection check, like TimeoutTime would still non-configurable as it is now no?

[Copilot]

Pull request overview

Adds liveness detection to PersistentProtocol so silent/idle connection deaths are detected even when there are no unacknowledged regular messages (i.e., cases where _recvAckCheck would never run). This supports more reliable remote/IPC reconnection behavior (related to microsoft/vscode-remote-release#11538).

Changes:

• Add _keepAliveTimeoutCheck() and invoke it after each keepalive send to detect “no incoming data for TimeoutTime” cases.
• Add a regression test ensuring keepalive-based timeouts fire when the peer stops sending data while no regular messages are pending.

Show a summary per file

File	Description
src/vs/base/parts/ipc/common/ipc.net.ts	Adds keepalive-driven socket timeout detection to catch silent idle disconnects.
src/vs/base/parts/ipc/test/node/ipc.net.test.ts	Adds a unit test validating the new keepalive timeout behavior when there are no pending regular messages.

Copilot's findings

• Files reviewed: 2/2 changed files
• Comments generated: 3

[Copilot]

_keepAliveTimeoutCheck() can fire a socket-timeout even when there are unacknowledged regular messages (it only checks time since last received data). This changes the existing timeout semantics: e.g. if the last incoming data was >20s ago but a new regular message was sent recently, this method can trigger a timeout earlier than the existing _recvAckCheck() would. If the intent is to catch idle silent deaths, consider guarding this check to only run when there are no unacknowledged regular messages (e.g. _outgoingUnackMsg is empty / unacknowledgedCount===0), or include timeSinceOldestUnacknowledgedMsg in the condition similar to _recvAckCheck().

[EhabY]

The intent is exactly to fire on idle silence: no data for 20s means the link is dead regardless of outgoing state. Guarding on unacknowledgedCount === 0 or ANDing in timeSinceOldestUnacknowledgedMsg reintroduces the gap this PR closes. The whole point is to measure silence from the peer, not time since our last send

[EhabY]

@alexdima this is ready for another review, please let me know if this should/will be tackled also by the Remote SSH extension as well (closed source though)