refactor(es/compat): Use special span instead of passing static_blocks_mark

[magic-akari]

Description:

Let's make an assumption in the code: the use of dummy span as a private class field name is generated by us, it is there simply because it is necessary to maintain the order of side effect execution, its name is not referenced elsewhere, so we can safely remove its name.

This is a breaking change for Rust users.

BREAKING CHANGE:

Related issue (if exists):

[changeset-bot]

🦋 Changeset detected

Latest commit: 586b38b

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 #9725 will degrade performances by 8.27%

_{Comparing magic-akari:refactor/static-blocks (586b38b) with main (aa0f784)}

Summary

❌ 2 regressions
✅ 192 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	main	magic-akari:refactor/static-blocks	Change
❌	es/full/base/fixer	116.8 µs	127.4 µs	-8.27%
❌	es/visitor/base-perf/boxing_unboxed_clone	2.6 µs	2.6 µs	-3.32%

[kdy1]

What if plugin authors create a private name with dummy spans?

I think some plugins related to decorators may generate some private names, and if it's the case, most of the plugin authors will simply use dummy span (at least initially)

[magic-akari]

Unfortunately, we have already used a similar trick here

swc/crates/swc_ecma_compat_es2015/src/classes/mod.rs

Lines 631 to 636 in a417ff4

	// Black magic to detect injected constructor.
	let is_constructor_default = constructor.span.is_dummy();
	if is_constructor_default {
	debug!("Dropping constructor parameters because the constructor is injected");
	constructor.params = Vec::new();
	}

Or here

swc/crates/swc_ecma_compat_es2015/src/classes/constructor.rs

Lines 23 to 28 in 49a8b3e

	// Black magic to detect injected constructor.
	let is_constructor_default = constructor.span.is_dummy();
	if is_constructor_default {
	debug!("Dropping constructor parameters because the constructor is injected");
	constructor.params.take();
	}

[kdy1]

Hmm. Makes sense. I'll review this PR later today.

[magic-akari]

Or should we add another special span?

By the way: We added PURE_SP last time, but we didn't add the documentation properly.

---

I personally prefer to use dummy span directly.

[kdy1]

I didn't think deeply enough while writing the DUMMY_SP hack because there was no Wasm plugin API, and I introduced static_block_mark because there's a custom plugin API at the moment of writing.

We may introduce another special flag in bytepos like SYNTHESIZED, one more for the Span type and use it, like you suggested.

Also, can you add a bit of documentation?