<i18n dev> RFR: 8350646: Calendar.Builder.build() Throws ArrayIndexOutOfBoundsException

[Naoto Sato]

On Tue, 25 Feb 2025 23:58:21 GMT, Justin Lu <jlu at openjdk.org> wrote:

> Please review this PR which prevents an `AIOOBE` from leaking out when `java.util.Calendar.Builder` is used to build a Japanese calendar with an era value too large.
> 
> Note that we don't check under `BEFORE_MEIJI`/0 as historically Japanese calendar ignores negative values during normalization. See `JapaneseImperialCalendar` L2018: `date.setEra(era > 0 ? eras[era] : null);`. 
> 
> We also check against `eras.length` over `REIWA`/5 due to the possibility of additional eras via the property override. (Which ensures `SupplementalJapaneseEraTestRun.java` passes, which also tests against building (and succeeding) with an era value of 6.

Looks good to me. I would expect a test case with `jdk.calendar.japanese.supplemental.era` defined with something, and 7 would throw an IAE.

src/java.base/share/classes/java/util/JapaneseImperialCalendar.java line 1862:

> 1860:             // BEFORE_MEIJI to be ignored during normalization
> 1861:             // We check against eras.length over Reiwa ERA due to possibility
> 1862:             // of additional eras via "jdk.calendar.japanese.supplemental.era"

The possibility is not only that emergency property, but by design Japanese calendar eras increase when the current emperor abdicates.

-------------

PR Review: https://git.openjdk.org/jdk/pull/23789#pullrequestreview-2645256501
PR Review Comment: https://git.openjdk.org/jdk/pull/23789#discussion_r1972020190