perf_hooks: implement eventLoopUtilization()

[cirospaciari]

Stacked on #34509 → #34424. Review those first; this PR is the last commit only.

performance.eventLoopUtilization() returned hardcoded {idle:0, active:0, utilization:0} and worker.performance's was a notImplemented stub. So test-worker-eventlooputil didn't fail — it hung forever, spinning on if (elu().idle <= 0) return setTimeout(r, 5).

Design

The loop already knew when it was about to park (will_idle_inside_event_loop), so the accounting is two clock reads on ticks that were going to make a sleeping syscall anyway — a busy tick pays nothing. Windows gets it from libuv's own metrics, enabled with uv_loop_configure(UV_METRICS_IDLE_TIME) (node does this unconditionally: node.cc:1587, node_worker.cc:175). Cross-thread reads are safe by construction: the counter is atomic, loop_start is fixed before the VM is published, and vm_lock only closes the TOCTOU on vm itself.

Two bugs worth recording

1. us_internal_loop_data_t is us_loop_t's FIRST member and is MIRRORED in Rust (src/uws_sys/InternalLoopData.rs). Adding a field without updating the mirror shifted num_polls, so us_loop_run_bun_tick took its num_polls == 0 early return and the loop stopped parking. No compile error. I twice concluded this was an architectural wall before a one-line printf of num_polls (1 without the field, 0 with it) named it. Anyone touching that struct should know.

2. A counter only folded in when a park ends reads stale mid-park, over-reporting active — 49.8 against the required 50. libuv has the identical problem and solves it the identical way: publish the park's entry time and let the reader add the in-progress interval (uv_metrics_idle_time, uv-common.c:1042-1056).

node parity (verified against the v26.3.0 binary)

case	node	bun
main-script (before the loop turns)	{0,0,0}	{0,0,0}
2-arg identical	utilization NaN	NaN
no-arg after the loop turns	0 < utilization < 1	same

The unguarded divisions are deliberate — node returns NaN for a zero total, so collapsing it to 0 would diverge. Read order is idle-then-now, matching node. The shared math lives in internal/perf/event_loop_utilization, exactly as node shares it between perf_hooks and worker_threads.

Also fixed

MessagePort listeners were invoked with this === undefined where node passes the port — injectFakeEmitter's wrapper had the receiver and dropped it. Scoped to MessagePort (its only caller); Worker uses a real EventEmitter and was already correct.

Verification

test-worker-eventlooputil: hung → 10/10, byte-identical to node, clean under BUN_JSC_validateExceptionChecks. perf_hooks 8 pass/0 fail. worker_threads unchanged at its 2 known failures. Windows target compiles.

Known divergence: node reports {0,0,0} during synchronous main-script evaluation because its loopStart milestone is still unset; Bun's loop_start is fixed at VM init. Modelling that needs node's real milestone, not a proxy — I tried iteration_nr and it's wrong, because a worker's script runs after its loop starts while the main script runs before, and it regressed the test 10/10 → 0/10.

[robobun]

^{Updated 4:13 PM PT - Jul 17th, 2026}

@autofix-ci[bot], your commit b24e939 is building: #74776

[github-actions]

Found 2 issues this PR may fix:

1. perf_hooks eventLoopUtilization()/heapUsed stubs silently break @fastify/under-pressure load shedding #34068 - PR replaces the hardcoded {idle: 0, active: 0, utilization: 0} stub with real idle-time tracking, fixing the silent breakage of @fastify/under-pressure load shedding
2. node:worker_threads does not implement Worker.performance and weird stub behaviour #32609 - PR implements worker.performance.eventLoopUtilization() with real cross-thread ELU reads, replacing the notImplemented stub

If this is helpful, copy the block below into the PR description to auto-close these issues on merge.

Fixes #34068
Fixes #32609

🤖 Generated with Claude Code

[cirospaciari]

Folding this into #34424 — three stacked PRs was more confusing than it was worth. Same commits, same tests, one review.

[github-actions]

This PR may be a duplicate of:

1. perf_hooks: implement performance.eventLoopUtilization() #32618 - Also implements performance.eventLoopUtilization() in perf_hooks with event loop idle time tracking

🤖 Generated with Claude Code