Proxy health check hangs indefinitely if proxy fails to bind to port 8877

[nunzio57]

What happened?

When GEMINI_SANDBOX_PROXY_COMMAND is set and the proxy process starts
but never binds to port 8877 (e.g. wrong port, slow startup, missing
dependency), the health check loop in sandbox.ts runs forever:

until timeout 0.25 curl -s http://localhost:8877; do sleep 0.25; done

The CLI hangs with no error message and the user must Ctrl+C manually.

This affects both the Seatbelt (macOS) and Docker sandbox paths.

Steps to reproduce

GEMINI_SANDBOX_PROXY_COMMAND="sleep infinity" GEMINI_SANDBOX=docker gemini -s
# CLI hangs forever with no output

Expected behavior

The health check should time out after a reasonable period and throw
a clear error message.

CLI version

0.30.0-nightly.20260210

Platform

Linux (WSL2 Ubuntu 24.04)

[nunzio57]

/assign

[gemini-cli]

👋 @nunzio57, thanks for your interest in this issue! We're reserving self-assignment for issues that have been marked with the help wanted label. Feel free to check out our list of issues that need attention.

[aishop-lab]

Related to this — I found that sandboxEnv proxy env vars (like HTTP_PROXY, HTTPS_PROXY) weren't being passed to the seatbelt sandbox spawn() call either (#19187, fix in #20788). The proxy plumbing in sandbox.ts has a few rough edges.

The timeout fix in #20821 handles the case where the proxy starts but doesn't bind to the port. There's also the case where the proxy process crashes immediately — the proxyProcess.on('close') handler throws FatalSandboxError from inside an event callback, which is an unhandled exception rather than a clean error. Might be worth hardening that path as a follow-up to make the whole proxy startup flow more predictable.