Fix metrics test cancellation race

[simongdavies]

Summary

The metrics test cancels its guest from a host callback. Cancellation targets the intended call without depending on scheduling delays. The test asserts ExecutionCanceledByHost and the callback duration metric.

Fixes #1815.

Validation

• The targeted test passes locally on Windows with default features and with function_call_metrics.
• just fmt-apply and just clippy pass.
• macOS HVF execution needs CI confirmation.

[Copilot]

🟡 Changes recommended

The updated test expects 5 metrics but does not currently assert all expected metric keys, which can allow incorrect metric emission to slip through.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR fixes flakiness in the host-side metrics test by canceling guest execution deterministically from a host callback, avoiding timing-based sleeps and asserting the expected cancellation error and emitted duration metrics.

Changes:

• Replace time-delayed cancellation (sleep + thread) with a host-function callback that triggers cancellation at a deterministic point.
• Update the test to call CallHostThenSpin and assert ExecutionCanceledByHost.
• Extend metrics assertions to include the guest function name change and the new host callback duration metric.

File summaries

File	Description
src/hyperlight_host/src/metrics/mod.rs	Makes the metrics test cancellation deterministic via a host callback and adjusts expected metrics/labels accordingly.

Review details

• Files reviewed: 1/1 changed files
• Comments generated: 1
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[ludfjig]

good idea calling kill from a host function (I guess slightly changes behavior in that we don't re-enter the vm rather than interrupt runnign vm, but metric is emitted in either case)