This relates to https://issues.apache.org/jira/browse/CXF-2450 .
Basically, unless the encoding option of cxf tools is specified, both cxf and xjc rely on the system encoding setup, defaulting to UTF-8 if nothing is set at system level.
On system with a specific encoding (e.g. Windows installed w/ Italian language), the "file.encoding" system property is set to a value different from UTF-8 (ISO-8859-1 for the italian setup). This can lead to generating sources which can't then be compiled using a more restrictive encoding like UTF-8.
Generally speaking it would hence be good to allow users specifying the desired encoding for the sources to be generated.