test(javascript): fail where webpack's output is bigger than swc's recorded one - #22389

Summary

The minify-corpora suite already holds webpack's minifier to terser's bytes and runs outputs that state what they print. swc's recorded output.js was only a report (JS_MINIFY_REPORT), never a check. Now each swc test's output under its own options must be no bigger than swc's (gzip decides, raw breaks a tie, as docs/performance.md weighs size), unless test/helpers/swcSmaller.js lists it under what swc does that webpack does not yet. A listed test that stops being bigger fails until retired, so the list only shrinks and a size regression is a failure.

The list groups all 422 current cases by transform, biggest first, as a work list:

Example, fixture/issues/11102:

// webpack (gzip +66 bytes)
export const zoom=(range=>{const isUnder=accuracy=>range<=accuracy;return isUnder(0)||isUnder(50)||isUnder(100)?15:11})(75);
// swc
export const zoom=15;

JS_MINIFY_REPORT=<file> now tallies standing: worse 422, the same 354, better 125. The reader also maps swc's camelCase topLevel in mangle.json to terser's toplevel; it was dropped as unknown, which made one test look worse.

Builds on the idea of #22355 (swc's tests in the minifier suite), which main since covered in #22352.

What kind of change does this PR introduce?

test

Did you add tests for your changes?

Yes, test/specCases/minify-corpora.spectest.js and test/helpers/swcSmaller.js. With an empty list the suite fails on all 423 cases; the topLevel fix made one listed test equal and the suite failed until it was retired. yarn test:minify-corpora passes, 232 tests in 336 s against 350 s on main.

Does this PR introduce a breaking change?

No, only tests and docs change.

If relevant, what needs to be documented once your changes are merged or what have you already documented?

docs/syntax.md and TESTING_DOCS.md describe the check and the list.

Use of AI

AI wrote the check, triaged each case where swc's output is smaller by comparing both outputs, and checked the five "swc is wrong" cases by running or reading them. The groups were reviewed against the diffs.

🤖 Generated with Claude Code

https://claude.ai/code/session_01ViD41i2KdwwGxrDk7yrjeD

Generated by Claude Code

Summary by CodeRabbit

• Tests

  • Expanded JavaScript minifier corpus checks to compare compressed output sizes against recorded SWC results, using raw size to break ties.
  • Added documented exceptions for cases where webpack output is larger, and checks to ensure those exceptions remain applicable.
  • Improved comparison reports to group results as worse, equal, or better.

• Documentation

  • Updated testing guidance to describe output validation, size comparisons, and exceptions.