wp_trim_words fails with certain charsets in the Excerpt

[SirLouen]

Bug Report

Following the ​GB74925 we have noticed that the excerpts were not being adapted accordingly in the Editor. It was also noticed that the current regex trim was not applying for certain more edge character types of cases like

Environment

• WordPress: 7.0-alpha-61215-src
• PHP: 8.2.29
• Server: nginx/1.29.4
• Database: mysqli (Server: 8.4.7 / Client: mysqlnd 8.2.29)
• Browser: Chrome 144.0.0.0
• OS: Windows 10/11
• Theme: Twenty Twenty-Three 1.6
• MU Plugins: None activated
• Plugins:
  • Gutenberg 22.4.1

Testing Instructions

1. Ideally use the 2023 theme
2. Create a new post and add the following excerpt:

Fabio　vel　iudice　vincam,　sunt	in	culpa	qui　officia.
Inmensae　subtilitatis,　obscuris　et　malesuada　fames.
Ambitioni　dedisse　scripsisse　iudicaretur.
Nec　dubitamus　multa　iter　quae　et　nos　invenerat.
Petierunt　uti　sibi　concilium　totius　Galliae　in　diem　certam　indicere.

3. Go to Appearance > Editor
4. Click on the canvas
5. Select the excerpt of the newly created Post
6. Adjust the Excerpt Max Number of Words (you need Gutenberg trunk to test this; if you don't use trunk, it will not work either in the Editor).
7. Set, for example, 30 words, and Save
8. Go to the front end.
9. 🐞 The number of words is not the same as in the editor.

Expected Behaviour

• The number of words of the excerpt should be the same in the Editor and in the front end.

Additional Information

Apparently, it appears that the wp_trim_words is not trimming all types of spacing characters accordingly; hence, the word count is inconsistent. @wildworks proposed that the /[\n\r\t ]+/ must be fixed for this purpose.

It appears that this was first introduced in [13942]/#10376

[SirLouen]

Pushing back to Future Release, as we don't really have a patch for it yet.

This updates the word boundary handling on UTF-8 sites to use a Unicode-aware whitespace pattern, so ideographic spaces (U+3000), non-breaking spaces (U+00A0), and other Unicode whitespace characters are treated as word separators. This matches the behavior already used in the Gutenberg editor.

For non-UTF-8 sites, the previous regex is kept as a fallback.

Props sirlouen, wildworks.
Fixes #64552.