Test262: twelve intl402 exclusions name tests that pass, and the thirteenth candidate keeps its entry

[lahma]

Jint.Tests.Test262/AGENTS.md says an ExcludedFiles entry must match a failing test and no passing one. Twelve intl402 entries no longer did. A thirteenth candidate the issue named still fails, and keeps its entry.

What was measured

Every one of the thirteen candidates was removed together, then run on its own, both parse modes, against this branch's engine. Twelve passed 2/2; one failed 2/2 and went back in.

exclusion removed	result	closed by
DateTimeFormat/prototype/formatRangeToParts/chinese-calendar-dates.js	0 failed, 2 passed	#3507
DateTimeFormat/prototype/formatToParts/chinese-calendar-dates.js	0 failed, 2 passed	#3507
DateTimeFormat/prototype/formatRangeToParts/dangi-calendar-dates.js	0 failed, 2 passed	already passing before #3405
DateTimeFormat/prototype/formatToParts/dangi-calendar-dates.js	0 failed, 2 passed	already passing before #3405
Temporal/PlainDate/from/roundtrip-from-property-bag.js	0 failed, 2 passed	#3482
Temporal/PlainDateTime/from/roundtrip-from-property-bag.js	0 failed, 2 passed	#3482
Temporal/PlainYearMonth/from/roundtrip-from-property-bag.js	0 failed, 2 passed	#3482
Temporal/ZonedDateTime/from/roundtrip-from-property-bag.js	0 failed, 2 passed	#3482
Temporal/PlainYearMonth/from/extreme-dates.js	0 failed, 2 passed	#3482
Temporal/PlainDate/prototype/with/basic-islamic-umalqura.js	0 failed, 2 passed	already passing before #3405
Temporal/PlainDateTime/prototype/with/basic-islamic-umalqura.js	0 failed, 2 passed	already passing before #3405
Temporal/ZonedDateTime/prototype/with/basic-islamic-umalqura.js	0 failed, 2 passed	already passing before #3405
Temporal/ZonedDateTime/prototype/withCalendar/extreme-dates.js	2 failed, 0 passed — kept	—

Attribution is bisected, not inferred. Each row's fix was pinned by running the same twelve-removal against the commit and its parent:

• Temporal: a date past a lunisolar calendar's table is still reckoned in that calendar #3482 (a date past a lunisolar calendar's table is still reckoned in that calendar) — the five roundtrip-from-property-bag / extreme-dates files fail 2/2 at its parent 5b2f8f12 and pass 2/2 at d8ab98c3. Both commits sit on the same pinned suite SHA, so the flip is the fix and not the #3277 corpus bump, which lands after it.
• Temporal: chinese and dangi read the same on every target framework #3507 (chinese and dangi read the same on every target framework) — the two chinese-calendar-dates files fail 2/2 at its parent 9f0afc3f (Temporal: a date a calendar reports fields for is a date it reckons arithmetic in #3502) and pass 2/2 at eb3bf010.
• The two dangi-calendar-dates files and the three with/basic-islamic-umalqura files already passed at a487ee06 (Temporal: a calendar the provider claims is a calendar Temporal accepts, and adding one is three overrides #3405), before the recent calendar work began. They are simply stale — added by Complete Intl API with remaining formatters and tests #2260 and Enable Intl.Era-monthcode and implement 7 new calendar systems (+2,498 test262 passes) #2344 respectively and never re-checked.

The one that stays

intl402/Temporal/ZonedDateTime/prototype/withCalendar/extreme-dates.js fails deterministically, on one assertion:

persian minimum supported date: eraYear result: Expected SameValue(«-272443», «-272442») to be true

Its year of -272442 is already right; only eraYear is off by one, and the whole file is "UTC", so no time zone or ICU data is involved. The issue scored it as passing; Jint/Native/Temporal/ is byte-identical between the issue's base 0b44c60a and main, so it was failing there too — the file is one of three near-identical withCalendar/extreme-dates.js siblings and its two neighbours stay excluded. Its entry now carries a comment naming the assertion, so the next reading of the group does not have to re-derive it.

The ICU question the four calendar-dates files raise

#3591's comment flags that whether chinese/dangi pass may depend on the machine's ICU data, which is why they were kept out of #3594 and want a PR whose cross-platform matrix decides them. That is this PR: linux, linux-ARM, macos and windows are the arbiter. If any leg reddens on those four, they come back out and the other eight stand on their own.

Verification

test262: 102,573 / 0 failed / 115 skipped of 102,688, up 24 passes from main's 102,549 / 0 / 139 at 1df62639 — exactly the twelve files × two parse modes, with the total unchanged. Full dotnet build -c Release and dotnet test -c Release are green on net472, net8.0 and net10.0, Jint.Tests.PublicInterface included.

Test-only: no engine change, so no migration-guide entry and no 4.x backport to consider.

Closes #3591

🤖 Generated with Claude Code

https://claude.ai/code/session_014W5mbjGhyvgAS4pivXoc4S