fix(es/compat): Add missing visit children for destructuring

[stormslowly]

Description:

add missing visit_mut_children_with in deconstructing transform

BREAKING CHANGE:
NO

Related issue (if exists):

ref: #9647

[changeset-bot]

🦋 Changeset detected

Latest commit: ebab9b3

The changes in this PR will be included in the next version bump.

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[codspeed-hq]

CodSpeed Performance Report

Merging #9658 will not alter performance

_{Comparing stormslowly:fix/deconstructing_skip_visiting_children (ebab9b3) with main (32116a0)}

Summary

✅ 194 untouched benchmarks

[magic-akari]

I believe the fix is correct, but I can't reproduce the issue in the swc playground.

I suspect our transformer has redundant elsewhere.

[stormslowly]

in my opinion, in swc playground run the for_of transform then destructing , so it can't be reproduced there.

@magic-akari you can try comment out the visit children call, then run the new added case, panic will appear

[kdy1]

If so, the order of transform is wrong. The compatibility transforms are expected to run in the correct order, and the correct order is.. well it's not documented but esxxxx functions encode the correct orders

[magic-akari]

If so, the order of transform is wrong.

swc/crates/swc_ecma_compat_es2015/src/for_of.rs

Lines 274 to 283 in 9c35520

	ForHead::Pat(pat) => prepend_stmt(
	&mut body.stmts,
	AssignExpr {
	span: DUMMY_SP,
	left: pat.try_into().unwrap(),
	op: op!("="),
	right: step.clone().make_member(quote_ident!("value")).into(),
	}
	.into_stmt(),
	),

In the for_of pass, pat has moved from the for-head to the for-body, which leads to the related visitors in the destructing always being skipped.
With the current sequence, we consistently fail to find this bug. (This issue would be revealed in an environment that supports for...of but not destructuring, via SWC env target rather than es version target)

Ideally, these two transformations should be independent, allowing us to retain destructuring (the pat structure) when downgrading from for...of.
Similarly, we should be able to downgrade destructuring while keeping for...of intact.

I have not yet confirmed whether the current situation matches the description above.

[magic-akari]

Overall, this fix is in the right direction, let's merge this fix first.