fix(runtime): setPrototypeOf cycle walk must not advance past chain end (#5866)

* fix(runtime): __proto__ cycle walk must treat undefined getPrototypeOf as chain-end

js_object_get_prototype_of returns undefined (not spec's object-or-null)
for some exotic receivers; the OrdinarySetPrototypeOf cycle walk fed that
back into the next advance and threw "Cannot convert undefined or null to
object". comment-json's __extends feature-test ({__proto__: []}) hit this
at Next.js server boot. Undefined can never be part of a genuine cycle,
so it ends the walk like null.

Claude-Session: https://claude.ai/code/session_01RcePwqv92QidGakrfYvf3v

* fix(runtime): setPrototypeOf cycle walk must not advance the hare past chain end

The Floyd walk in js_object_set_prototype_of only null-guarded the hare's
second step. On any acyclic proto chain longer than one link (every
function/class proto: fn -> Function.prototype -> Object.prototype -> null)
the hare reaches null first and the next iteration called advance(null) ->
js_object_get_prototype_of(null) -> TypeError 'Cannot convert undefined or
null to object'. comment-json's __extends hit this on every transpiled
subclass, killing Next.js server boot on main (bisected to 1f4a2c5,
#5763). Freeze the hare at null; the tortoise still completes the
obj-membership walk, and a cyclic chain never yields a null hare.

Claude-Session: https://claude.ai/code/session_01RcePwqv92QidGakrfYvf3v

* test(runtime): #5763 regression — e2e guards for setPrototypeOf chain-end fixes

Four cases vs node --experimental-strip-types: acyclic chains of length
2/4, the __extends statics link (3-link function proto chain), the
comment-json {__proto__: []} feature-test evaluation (throw-only guard;
the instanceof gap is separate and pre-existing), and a genuine-cycle
negative.

* chore(release): v0.5.1214 — version bump + changelog for #5866

---------

Co-authored-by: Ralph Küpper <ralph@skelpo.com>

Author: proggeramlug

CHANGELOG.md

@@ -1,3 +1,7 @@
+## v0.5.1214 — setPrototypeOf cycle walk must not advance past chain end (#5866)
+
+Fixes the #5763 regression that broke `Object.setPrototypeOf(obj, proto)` for every acyclic prototype chain of ≥2 links ("Cannot convert undefined or null to object"): the OrdinarySetPrototypeOf Floyd walk advanced the hare past the chain end, and fed exotic `undefined` prototype reports back into the next advance. Every transpiled `__extends` (`setPrototypeOf(Derived, Base)` — a 3-link function proto chain) threw; comment-json's `__extends` + its `{ __proto__: [] }` feature test killed Next.js server boot. The hare now freezes at null (the tortoise completes the membership walk alone, so cycle detection is unchanged) and `undefined` ends the walk like null. e2e regression suite: `crates/perry/tests/issue_5763_setprototypeof_chain_end.rs` (chains of length 2/4, the `__extends` statics link, the feature-test evaluation, and a genuine-cycle negative), byte-for-byte vs `node --experimental-strip-types`.
+
 ## v0.5.1213 — representation-aware type lowering + native-ABI material evidence gate (#5466)
 
 Lands the type-lowering track on main (builds on #5291; integrates #5462's

CLAUDE.md

@@ -8,7 +8,7 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 Perry is a native TypeScript compiler written in Rust that compiles TypeScript source code directly to native executables. It uses SWC for TypeScript parsing and LLVM for code generation.
 
-**Current Version:** 0.5.1213
+**Current Version:** 0.5.1214
 
 
 ## TypeScript Parity Status