Polyfill: Make Intl.DateTimeFormat replacement more conformant

Previously we did not care very much about fidelity for the replacement
Intl.DateTimeFormat object, since the point of this code is to run the
Temporal tests. Occasionally people want to run the Intl.DateTimeFormat
tests and are surprised that it doesn't work.

Things we have to do to get the Intl.DateTimeFormat tests to pass:

- Change ASCIILowercase not to use a RegExp, so that it doesn't clobber
  legacy RegExp variables like RegExp.$_. (Probably this means that for a
  100% conformant polyfill, you can't use any RegExp anywhere. I'm not
  bothering with that.)

- Change DateTimeFormat to be a class, DateTimeFormatImpl. Methods defined
  in a class body don't have a 'prototype' property and can't be called as
  constructors, whereas regular functions do.
  (However, we still need a regular function to construct DateTimeFormat,
  because it needs to be able to be invoked without 'new'.)

- Use CreateSlots/GetSlot/SetSlot like the other Temporal objects, instead
  of symbol-valued properties. The symbol-valued properties were
  observably present for user code.

- In the constructor, observably read all the options from the options bag
  in the expected order, and copy them to a null-prototype object with
  which we do subsequent operations unobservably.

- Implement the behaviour of format() already having bound the correct
  this-object.

- Previously we only half-implemented the normative optional behaviour of
  being able to 'bless' a plain object into becoming a DateTimeFormat
  object.
  Instead, just don't implement it at all. (A real production polyfill
  should probably check if the existing Intl.DateTimeFormat has that
  behaviour, and match that.)

Closes: #2471

Author: ptomato

polyfill/ci_codecov_test262.sh

@@ -24,13 +24,9 @@ for subdir in $subdirs; do
   node runtest262.mjs "$subdir/**" || failed=1
 done
 node runtest262.mjs "test262/test/staging/Intl402/Temporal/**/*.js" || failed=1
-node runtest262.mjs "test262/test/intl402/**/*[tT]emporal*.js" || failed=1
-# TODO: remove the line above and uncomment the three lines below to run tests for all localized
-# date formatting, because the Temporal polyfill replaces the entire DateTimeFormat object.
-# See https://github.com/tc39/proposal-temporal/issues/2471 for more info.
-# node runtest262.mjs "test262/test/intl402/DateTimeFormat/**/*.js" || failed=1
-# node runtest262.mjs "test262/test/intl402/Intl/DateTimeFormat/**/*.js" || [ $? = 66 ] || failed=1
-# node runtest262.mjs "test262/test/built-ins/Date/*/toLocale*String/*.js" || failed=1
+node runtest262.mjs "test262/test/intl402/DateTimeFormat/**/*.js" || failed=1
+node runtest262.mjs "test262/test/intl402/Intl/DateTimeFormat/**/*.js" || failed=1
+node runtest262.mjs "test262/test/built-ins/Date/*/toLocale*String/*.js" || failed=1
 
 c8 report --reporter=text-lcov --temp-directory=$NODE_V8_COVERAGE \
   --exclude=polyfill/runtest262.mjs \

polyfill/lib/ecmascript.mjs

@@ -5722,13 +5722,16 @@ export function ASCIILowercase(str) {
   // values. For example, Turkish's "I" character was the source of a security
   // issue involving "file://" URLs. See
   // https://haacked.com/archive/2012/07/05/turkish-i-problem-and-why-you-should-care.aspx/.
-  return Call(StringPrototypeReplace, str, [
-    /[A-Z]/g,
-    (l) => {
-      const code = Call(StringPrototypeCharCodeAt, l, [0]);
-      return StringFromCharCode(code + 0x20);
+  let lowercase = '';
+  for (let ix = 0; ix < str.length; ix++) {
+    const code = Call(StringPrototypeCharCodeAt, str, [ix]);
+    if (code >= 0x41 && code <= 0x5a) {
+      lowercase += StringFromCharCode(code + 0x20);
+    } else {
+      lowercase += StringFromCharCode(code);
     }
-  ]);
+  }
+  return lowercase;
 }
 
 // This function isn't in the spec, but we put it in the polyfill to avoid