在Retrofit 1 时，会在onError中返回并抛出HttpException，但是在2中，不会再回调onError了，而是会在onNext中，这个任务就需要客户端自己处理。

https://futurestud.io/tutorials/retrofit-2-simple-error-handling

参考如下：

Error Handler in Action

etrofit 2 has a different concept of handling "successful" requests than Retrofit 1. In Retrofit 2, all requests that can be executed (sent to the API) and for which you’re receiving a response are seen as "successful". That means, for these requests the onResponse callback is fired and you need to manually check whether the request is actual successful (status 200-299) or erroneous (status 400-599).

If the request finished successfully, we can use the response object and do whatever we wanted. In case the error actually failed (remember, status 400-599), we want to show the user appropriate information about the issue.

Call<User> call = service.me();  
call.enqueue(new Callback<User>() {  
    @Override
    public void onResponse(Call<User> call, Response<User> response) {
        if (response.isSuccessful()) {
            // use response data and do some fancy stuff :)
        } else {
            // parse the response body …
            APIError error = ErrorUtils.parseError(response);
            // … and use it to show error information

            // … or just log the issue like we’re doing :)
            Log.d("error message", error.message());
        }
    }

    @Override
    public void onFailure(Call<User> call, Throwable t) {
        // there is more than just a failing request (like: no internet connection)
    }
});

As you can see, we use the ErrorUtils class to parse the error body and get an APIError object. Use this object and the contained information to show a meaningful message instead of a generic error message.