When using certain fonts (in my case, Noto CJK, both pan-CJK and subset
versions), the output looks right but when copied from the generated PDF file,
everything becomes garbage. For example:
\version "2.22.0"
\markup {
\override #'(font-name . "Noto Serif CJK SC")
{ English 中文 にほんご }
}
The output looks right, but when copied it becomes "&OHMJTI ╚倁 מ؆ׇ". If I use Noto Sans CJK SC, it will be "&OHMJTI ╈乄 מ؆ׇ" (different from the previous one). Noto Serif JP (not pan-CJK) then "&OHMJTI ╚倀 מ؆ׇ". (It seems that LilyPond is smart when selecting the language depending on your system environment, so we may have to use the subset version to select explicitly.)