Update closure-compiler. NFC

[sbc100]

No description provided.

[kripken]

The code diffs look fine, but most of the size changes are small but worse. Maybe worth investigating one to see what changed?

[sbc100]

files.zip

[sbc100]

I attached the JS generated for the test_minimal_runtime_code_size_hello_embind test. The size differs by just one byte, but if I clang-format the files they don't differ at all.

Any idea how to find which is the differing byte here?

[sbc100]

Ah, I found an interesting hunk in the code generated by test_codesize_mem_O3_standalone_narg_flto:

@@ -139,7 +139,7 @@
   if (b.instantiateWasm)
     return new Promise(d => {b.instantiateWasm(c, (f, e) => {d(a(f, e))})});
   z ??= b.locateFile ? b.locateFile("a.out.wasm", n) : n + "a.out.wasm";
-  return a((await C(c)).instance)
+  return function(d) { return a(d.instance) }(await C(c))

[sbc100]

Looks like maybe some subtle async/await difference ?

[kripken]

-  return a((await C(c)).instance)
+  return function(d) { return a(d.instance) }(await C(c))

Hmm these are identical in every way I think, except that a stack trace would show that nameless thunk function..? Weird closure doesn't remove it, since it does inlining etc. so it is ok with changes to stack traces in principle.

[sbc100]

Closing favor of #26869