Latest Results
fix(es/minifier): preserve bindings in copied inline arrows (#12141)
**Description:**
Cheap arrow functions stored in `Finalizer::lits` can be cloned into
multiple sibling scopes while retaining the same resolved binding IDs. A
later compression pass can turn each copy into an IIFE with local
declarations. With mangling enabled, the shared ID can receive a short
name that collides with another binding in one sibling scope, producing
invalid JavaScript.
This change freshens binding IDs when `Finalizer` clones an inline
literal. It applies at both clone sites (identifier replacement and
object-shorthand expansion), which fixes the ownership boundary instead
of coupling hygiene to a later IIFE optimization. Non-binding literals
keep the existing fast path. The helper also replaces the equivalent
existing remapping code for cloned simple callees.
The regression fixture is the exact bundle shape reported by Next.js.
Its empty `expected.stdout` intentionally makes the fixture execute the
generated code, so the pre-fix duplicate declaration fails with a parse
error.
**Validation:**
- `cargo fmt --all`
- `cargo clippy -p swc_ecma_minifier --lib -- -D warnings`
- `cargo test -p swc_ecma_minifier --no-fail-fast`
- `crates/swc_ecma_minifier/scripts/exec.sh` (484 minifier fixtures and
2,128 execution fixtures passed)
- rebuilt the historical SWC native binding consumed by the affected
Next.js release; the reporter's Next.js 16.3.1 app builds, starts, and
renders correctly
- verified passes 1/2/3, mangle on/off, toplevel on/off, object
shorthand, and same-spelling collision variants
- interleaved benchmark over 23 reporter server chunks: invalid outputs
1 → 0; aggregate output 569,415 → 569,403 bytes; candidate time +1.28%
and +1.56% in two runs (A/A noise +0.44% and +0.28%)
The workspace-wide all-target clippy command was also attempted; it is
currently blocked by unrelated pre-existing unknown-lint errors in
`swc_ecma_lints` and `swc_ecma_transforms_react`. The touched minifier
crate's library clippy is clean.
**Related issue:**
- https://github.com/vercel/next.js/issues/97517
- Regression became reachable after
https://github.com/swc-project/swc/pull/11021
---------
Co-authored-by: DongYun Kang <kdy.1997.dev@gmail.com> Latest Branches
0%
fix/parse-unambiguous-followup 0%
fix/unsafe-wrong-ast-path 0%
marcoshernanz:codex/fix-cloned-iife-mangle-bindings © 2026 CodSpeed Technology