Regression in @babel/plugin-transform-named-capturing-groups-regex with unicode regexps

[Jessidhia]

Bug Report

Current Behavior
Likely a regression from #10430, but the contents of u regexps are getting transformed to being non-u without updating a regexp's flags. The problem becomes a more serious bug when the s flag is also present, as the transformed . no longer matches newlines.

Input Code

• REPL link

const re = /<(?<tag>\d)+>.*?<\/\k<tag>>/su
console.log(re.test('<0>xxx\nyyy</0>')) // should be true

• Link using an older version of the plugin, without the regression

Expected behavior/code
A clear and concise description of what you expected to happen (or code).

Output with @7.4.5 (ignoring helpers, just the regexp):

/<(\d)+>.*?<\/\1>/su

Output with @7.6.3:

/<([0-9])+>(?:[\0-\t\x0B\f\x0E-\u2027\u202A-\uD7FF\uE000-\uFFFF]|[\uD800-\uDBFF][\uDC00-\uDFFF]|[\uD800-\uDBFF](?![\uDC00-\uDFFF])|(?:[^\uD800-\uDBFF]|^)[\uDC00-\uDFFF])*?<\/\1>/su

Note how the regexp, which didn't even have to be transformed from being a unicode regexp (and didn't remove the u flag after doing it), specifically skips over both \x0A and \x0D, the \n and \r characters, which are supposed to be matched because the regexp has the s flag.

Babel Configuration (.babelrc, package.json, cli command)

{
  "plugins": ["plugin-transform-named-capturing-groups-regex"]
}

Environment

• Babel version(s): 7.6.4
• Node/npm version: v12.10.0 / yarn 1.19.1
• OS: macOS 10.15
• Monorepo: yes
• How you are using Babel: babel-jest

Possible Solution
Need some way to tell regexpu to not do the unicode downleveling; or if unicode transformation is "inevitable" the dotAll flag needs to be respected.

[nicolo-ribaudo]

Need to check if it's fixed by #10447

[JLHwung]

Will be fixed in #10447, see #10447 (comment) for current approach.

Need some way to tell regexpu to not do the unicode downleveling

True, I have proposed useDotAllFlag support to regexpu-core: mathiasbynens/regexpu-core#34