Up until now I have been using std::string in my C++ applications for embedded system (routers, switches, telco gear, etc.).
For the next project, I am considering to switch from std::string to std::wstring for Unicode support. This would, for example, allow end-users to use Chinese characters in the command line interface (CLI).
What complications / headaches / surprises should I expect? What, for example, if I use a third-party library which still uses std::string?
Since support for international strings isn’t that strong of a requirement for the type of embedded systems that I work on, I would only do it if it isn’t going to cause major headaches.
Note that many communications protocols require 8-bit characters (or 7-bit characters, or other varieties), so you will often need to translate between your internal wchar_t/wstring data and external encodings.
UTF-8 encoding is useful when you need to have an 8-bit representation of Unicode characters. (See How Do You Write Code That Is Safe for UTF-8? for some more info.) But note that you may need to support other encodings.
More and more third-party libraries are supporting Unicode, but there are still plenty that don’t.
I can’t really tell you whether it is worth the headaches. It depends on what your requirements are. If you are starting from scratch, then it will be easier to start with std::wstring than converting from std::string to std::wstring later.