Branch: refs/heads/main
Home: https://github.com/WebKit/WebKit
Commit: 284afdacfb77ee797e4874af70539b8b8b3b16d9
https://github.com/WebKit/WebKit/commit/284afdacfb77ee797e4874af70539b8b8b3b16d9
Author: Yijia Huang <[email protected]>
Date: 2026-08-10 (Mon, 10 Aug 2026)
Changed paths:
A JSTests/stress/temporal-lunisolar-extreme-year-arithmetic.js
A JSTests/stress/temporal-plainyearmonth-toplaindate-day-range.js
M Source/JavaScriptCore/runtime/TemporalPlainYearMonthPrototype.cpp
M Source/JavaScriptCore/runtime/temporal/core/CalendarICUBridge.cpp
Log Message:
-----------
[JSC][Temporal] Fix non-ISO date field resolution at the range edges
https://bugs.webkit.org/show_bug.cgi?id=321414
rdar://184478785
Reviewed by Yusuke Suzuki.
toPlainDate's non-ISO branch narrowed day to uint8_t, so day mod 256 decided
the answer,
and both clamps in nonISOCalendarDateToISO were one-sided, letting the wrapped
0 build a
live PlainDate with day 00:
PlainYearMonth.from({ year: 2567, month: 1, calendar: "buddhist" })
.toPlainDate({ day: 256 }) // 2024-01-00, want 2024-01-31
day is saturated with clampTo<uint8_t> now and both clamps are two-sided. V8 and
SpiderMonkey agree.
calendarDateAdd and calendarDateUntil queried ICU past the range where its
astronomical
output is trustworthy, though the chinese and dangi getters already report
plain ISO fields
there, so add({years: 1}) on {calendar: "chinese", year: 270000} threw on a
date that
constructs and reads fine. Both take the ISO path past the threshold now. That
range is a
deliberate divergence, so the values are not expected to match V8; only the
absence of a
failure is.
calendarDateUntil passed the calendar's arithmetic year to
resolveMonthCodeToOrdinal, whose
year argument is the related year -- for chinese and dangi a different number
on ICU 76
(4661 against 2024, 4353 against 2020) and the same one on ICU 78, so a local
run cannot
tell them apart. The snapshot carries the related year separately now, computed
from the
year's start instant rather than read from an ICU year field, and only the
source endpoint
computes it because only the year loop consumes it. Comparisons still use the
arithmetic
year, so both operands of one always come from the same field. Nothing branches
on an ICU
version.
Tests: JSTests/stress/temporal-lunisolar-extreme-year-arithmetic.js
JSTests/stress/temporal-plainyearmonth-toplaindate-day-range.js
Canonical link: https://commits.webkit.org/318940@main
To unsubscribe from these emails, change your notification settings at
https://github.com/WebKit/WebKit/settings/notifications
