Wikimedia is updating to libicu57, so we should update the version of Unicode used in utfnormal and Scribunto to match. These patches should be deployed after libicu finishes being upgraded.
|Resolved||Legoktm||T188480 Update MediaWiki to use Unicode 8.0.0 (utfnormal, Scribunto, etc.)|
|Resolved||MoritzMuehlenhoff||T177498 Provide a forward port of ICU 52 for stretch / Investigate best ICU update strategy|
|Resolved||None||T189295 ICU 57 migration for wikis using non-default collation|
|Resolved||Quiddity||T189486 Announcing ICU 57 transition to the community|
|Resolved||Ladsgroup||T190965 Remove uca-fa from beta cluster|
We already updated css-sanitizer for the new utfnormal version, but it looks like we also need to update RemexHtml and Equivset: https://packagist.org/packages/wikimedia/utfnormal/dependents
As a side note, the Unicode situation in our environment is complicated:
- libucu57, and thus PHP's intl extension, will use Unicode 8.0.0.
- But the Unicode support in PCRE functions is not tied to that. The current version of libpcre on terbium is 8.35, while stretch has 8.39. The PCRE changelog entries mentioning Unicode version updates indicate that these use 6.3.0 and 7.0.0, respectively.
- The mbstring library in HHVM uses a really old version of Unicode, maybe 3.2 assuming they copied it from Zend PHP before this commit. PHP7.0's mbstring uses Unicode 8.0.0, and 7.2 uses Unicode 10.