use v8 copy constructor for v8 global

[anonrig]

Create a proper independent persistent handle via V8's copy constructor. The previous implementation would cause a double-free when both the original and clone were dropped.

[ask-bonk]

I'm Bonk, and I've done a quick review of your PR.

This fixes a double-free bug in Global<T> cloning by replacing a bitwise pointer copy with V8's proper v8::Global<T>(Isolate*, const Global&) copy constructor, and removes the Clone trait impl in favor of an inherent method that takes &mut Lock.

The fix is correct — the old implementation created two v8::Global handles that shared the same persistent handle slot, so dropping both would double-free. One observation:

1. Low: No regression test — This was a serious memory safety bug (double-free). It would be good to add a test in src/rust/jsg-test/ that exercises Global<T>::clone() and verifies both original and clone can be independently dropped and dereferenced. Not blocking, but worth tracking.

[ask-bonk]

I'm Bonk, and I've done a quick review of your PR.

This fixes a double-free bug in Global<T> cloning by replacing a bitwise pointer copy with V8's proper v8::Global<T>(Isolate*, const Global&) copy constructor, and removes the Clone trait impl in favor of an inherent method that takes &mut Lock.

The fix is correct. The C++ side properly uses global_as_ref_from_ffi to get a const reference to the original v8::Global<v8::Value> and constructs a new independent persistent handle via v8::Global<v8::Value>(isolate, original). The Rust side correctly removes the Clone trait (which couldn't provide an isolate) and replaces it with an inherent clone(&self, lock: &mut Lock) method. The #[must_use] annotation and the explanatory comment are good additions.

One low-severity observation posted: no regression test for the double-free fix. Not blocking.

github run