Polyfill: fix Chinese/Dangi calendar at extreme date ranges

[jessealama]

Attempt to fix #3081 by using the 19-year Metonic cycle to shift extreme ISO years into ICU4C's safe range before calling Intl.DateTimeFormat, then adjusting the calendar year back. This produces approximate but non-throwing results for dates outside ICU4C's supported range, which should be sufficient for the test262 extreme-dates tests that only check that no error is thrown. Happy to adjust the approach if there's a better way to handle this.

[codecov]

Codecov Report

❌ Patch coverage is 97.95918% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 98.09%. Comparing base (86a7b46) to head (f9ef3df).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
polyfill/lib/calendar.mjs	97.91%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3277      +/-   ##
==========================================
- Coverage   98.09%   98.09%   -0.01%     
==========================================
  Files          22       22              
  Lines       10457    10504      +47     
  Branches     1808     1813       +5     
==========================================
+ Hits        10258    10304      +46     
- Misses        182      184       +2     
+ Partials       17       16       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ptomato]

Cool, thanks!

This fix means you can probably also add "chinese" and "dangi" to the list of calendars in test/thorough/calendardaymath.js and remove the comment explaining why they need to be skipped.

I like the solution of shifting the year into range. I had a half-started branch for this issue where I did something similar only instead of shifting the year, I made a fake Metonic cycle going into the past and future. I like your solution better.

One thing from my branch that might be helpful to have here, I added a method to helperChinese:

// https://unicode-org.atlassian.net/browse/ICU-23286
isVulnerableTo70000Bug() {
  if (this.vulnerableTo70000Bug === undefined) {
    const formatter = this.getFormatter();
    try {
      Call(IntlDateTimeFormatPrototypeFormatToParts, formatter, [2146851043199999 + 1]);
    } catch {
      this.vulnerableTo70000Bug = true;
    }
    this.vulnerableTo70000Bug = false;
  }
  return this.vulnerableTo70000Bug;
},

(And only do the shifting if this flag is true.) This way, if a runtime ships a fix to the bug, we won't override their data.

[jessealama]

Cool, thanks!

This fix means you can probably also add "chinese" and "dangi" to the list of calendars in test/thorough/calendardaymath.js and remove the comment explaining why they need to be skipped.

I like the solution of shifting the year into range. I had a half-started branch for this issue where I did something similar only instead of shifting the year, I made a fake Metonic cycle going into the past and future. I like your solution better.

One thing from my branch that might be helpful to have here, I added a method to helperChinese:

// https://unicode-org.atlassian.net/browse/ICU-23286
isVulnerableTo70000Bug() {
  if (this.vulnerableTo70000Bug === undefined) {
    const formatter = this.getFormatter();
    try {
      Call(IntlDateTimeFormatPrototypeFormatToParts, formatter, [2146851043199999 + 1]);
    } catch {
      this.vulnerableTo70000Bug = true;
    }
    this.vulnerableTo70000Bug = false;
  }
  return this.vulnerableTo70000Bug;
},

(And only do the shifting if this flag is true.) This way, if a runtime ships a fix to the bug, we won't override their data.

I've incorporated this helper; thanks! I've also ungated the Chinese and Dangi calendars, so they won't be skipped now.

[ptomato]

Thanks!