Summary

Revert the file.encoding initialization back to JDK11's behavior.

Problem

With the fix to JDK-4947890, the system property file.encoding changed to be initialized to sprops->encoding, which used to be sprops->sun_jnu_encoding on non macOS platforms. This became an issue on Windows with system locale and user locale set to be different. For example, a user with a Japanese Windows (system locale = Japanese(Japan)) and user's locale English (United States) sees the following behavioral difference:

JDK11:
file.encoding = MS932
sun_jnu_encoding = MS932

JDK17:
file.encoding = Cp1252
sun_jnu_encoding = MS932

java.nio.charset.Charset#defaultCharset() relies on this file.encoding system property, so the users encounter inconsistent default charset with the said condition on Windows.

Solution

Initialize file.encoding based on sun_jnu_encoding on non macOS platforms as prior to the fix to JDK-4947890.

Specification

N/A. This is a behavioral change only.