gc: admit two provably-leaf helpers, and measure that it buys nothing

[proggeramlug]

Stacked on #7368 (needs its report fix to measure anything).

The audit

js_gc_register_global_root was the single most frequent non-leaf callee across the probe suite — 148 call sites — and it is provably GC-leaf. Its entire body:

pub extern "C" fn js_gc_register_global_root(ptr: i64) {
    if !root.is_null() {
        unsafe { runtime_write_barrier_root_heap_word(*root); }
    }
    GLOBAL_ROOTS.with(|roots| { roots.borrow_mut().push(root); });
}

The first call is exactly what js_write_barrier_root_heap_word — already CannotCollect — wraps in one line. It shades one header and calls push_mark_seed, which is a TLS Vec::push. No trace, no sweep, no trigger.

The Vec::push is the part worth pausing on, because CLAUDE.md lists a "malloc count threshold" as a GC trigger. It does not apply: that counter is MALLOC_STATE.objects.len(), a registry of Perry GC objects, and the #[global_allocator] is plain mimalloc/System with no GC hook. A raw Rust allocation cannot arm a trigger — the case the module doc already carves out ("may … allocate through the system allocator and still be safe").

js_typed_feedback_maybe_dump_trace joins its already-admitted family siblings: env read, JSON serialise, file write; empty body without diagnostics.

And it buys nothing — measured, same tree, A/B

probe	safepoints	roots	total bytes	__text
06_string_retention	105 → 100	27 → 27	0	−4 B
09_try_catch_roots	343 → 339	259 → 259	0	−4 B
11_collect_at_depth	120 → 117	36 → 36	0	−4 B

Root counts are identical. The 40 safepoints removed across the suite were all rootless, and a rootless safepoint costs essentially nothing.

Which is exactly what docs/engine-plan.md already told us:

Statepoints have no fixed cost. A function with nothing live across a safepoint pays nothing at all — the axis is not "statepoints are bigger", it is "roots are bigger".

I pulled the safepoint-count lever and confirmed it does not move the size axis. That is the useful output here: sequencing step 2's "reduce root density" has to attack live-root sets, not safepoint counts. I'd rather have that recorded with numbers than have the next person spend a week on the same lever.

Why land it at all

Three reasons, none of them "it's faster":

1. The classification was wrong. js_gc_register_global_root is leaf and was marked Unknown. Accuracy in this table is load-bearing in the other direction too.
2. The tests block the dangerous next step. allocating_helpers_are_not_cannot_collect pins js_nanbox_string out of the allowlist. At 120 call sites it's the obvious next candidate and reads as pure bit manipulation — but its null-pointer guard calls js_string_from_bytes to allocate an empty string rather than boxing null. I nearly admitted it before reading the body. Marking a collecting helper as leaf means the collector misses roots: silent heap corruption, the worst failure direction in this system.
3. register_global_root_tracks_the_barrier_it_wraps ties the two classifications together, so a future demotion of the barrier can't leave its wrapper claiming to be leaf.

If you'd rather not carry a null-result change, the tests and the audit comments are the parts worth keeping — say so and I'll strip it to those.

Verification

Probe suite 11/11 byte-identical to the Node oracle under PERRY_RS4GC=1 PERRY_GC_FORCE_EVACUATE=1 PERRY_GC_VERIFY_EVACUATION=1. cargo test -p perry-codegen --lib green.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 0d895545-6921-4a72-926f-f7c9cc5d992e

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}