Character encoding for strings in and out of libcgraph


I understand that by default, as specified here, DOT assumes the UTF-8 encoding and is like HTML in this regard. I wonder if the strings passed in to libcgraph, such as the names of nodes, need to be UTF-8 encoded as well, or is the locale’s default encoding expected? In other words, is calling agnode(graph, "foobar", true) okay, or should one do agnode(graph, u8"foobar", true)?

Also, does agwrite() write the graph in UTF-8, or does it write it in the locale’s default encoding (“the execution character set”)? What does agread() assume?