[v12.x] icu: broken date locale string format on zh-CN

[legendecas]

• Version: v12.17.0-12.18.4
• Platform: all
• Subsystem: icu

What steps will reproduce the bug?

$ LANG=zh_CN.UTF-8 node -p 'new Date().toLocaleString()'
2020-9-28 3:33:35 ├F10: PM┤

How often does it reproduce? Is there a required condition?

Always.

What is the expected behavior?

node v12.16.3 (small-icu, English only)

$ LANG=zh_CN.UTF-8 node -p 'new Date().toLocaleString()'
2020-9-28 15:34:08

node v14.12.0 (full-icu)

$ LANG=zh_CN.UTF-8 node -p 'new Date().toLocaleString()'
2020/9/28 下午3:35:06

Both cases are correct.

What do you see instead?

Invalid and meaningless symbol characters "├F10: PM┤" in localized strings.

Additional information

Reverting #33337 (5c0232a, 2d76ae7) can fix the issue.

[richardlau]

Have you installed the full-icu module? You may need to reinstall it so that it installs the full ICU data to match the version of ICU used in Node.js.

cc @nodejs/i18n-api

[legendecas]

@richardlau v12.x is built with small-icu by default (doc). Is this issue possibly related to full-icu?

[richardlau]

I wouldn't expect the Node.js 14 behaviour without full ICU. I can't remember if Node.js will ignore a full ICU install for the wrong version of ICU or would error/silently ignore it.

[legendecas]

Yes, Node.js 14 behavior is expected to working with full-icu. The issue is not if the new Date().toLocaleString() result is localized or not. In v12.16.3, the locale result should be identical to the default English result (or some locales supported in icu-small, human-readable). However, starting from v12.17.0 there are unexpected symbols in the localized strings.

[srl295]

the symbol ├F10: PM┤ sounds like absolutely no data was found, not even English.

[srl295]

repro'ed. Sounds like the data slicing had a failure. Some data is present, but not other data.

[richardlau]

cc @nodejs/i18n-api Any idea what we should do with this one?

[srl295]

@richardlau sorry, i have not looked into this any further

[targos]

Closing this issue because v12.x goes EOL at the end of this month and no more releases are planned.