If C++ has advantages over C, then I don’t see a reason to limit yourself to C in libraries (unless those advantages aren’t helpful for libraries).
I do agree that writing libraries is fundamentally different (and fundamentally harder) than writing apps. I still don’t feel comfortable passing smart pointers across a C++ API (in or out), so I’m fine with a rule that the API should generally be C-friendly, but you give up a lot when you limit yourself to C, and some of what you give up would be very useful inside the library.
I do agree that writing libraries is fundamentally different (and fundamentally harder) than writing apps. I still don’t feel comfortable passing smart pointers across a C++ API (in or out), so I’m fine with a rule that the API should generally be C-friendly, but you give up a lot when you limit yourself to C, and some of what you give up would be very useful inside the library.