Java accessors generated by protoc for string fields

For any field x of type string, protoc seems to generate two java accessors:

• getX() that returns a String
• getXBytes() that returns a ByteString

While the getXBytes() method may be useful, I am surprised that this behavior doesn't seem to documented in the official protoc Java documentation.

Moreover, this introduces a potential name conflict issue. Take for example the following proto definition:

syntax = "proto3";

message Message {
  string x = 2;
  string x_bytes = 3;
}

Here is the corresponding interface generated by protoc

  public interface MessageOrBuilder extends
      // @@protoc_insertion_point(interface_extends:Message)
      com.google.protobuf.MessageOrBuilder {

    /**
     * <code>string x = 2;</code>
     * @return The x.
     */
    java.lang.String getX();
    /**
     * <code>string x = 2;</code>
     * @return The bytes for x.
     */
    com.google.protobuf.ByteString
        getXBytes();

    /**
     * <code>string x_bytes = 3;</code>
     * @return The xBytes.
     */
    java.lang.String getXBytes();
    /**
     * <code>string x_bytes = 3;</code>
     * @return The bytes for xBytes.
     */
    com.google.protobuf.ByteString
        getXBytesBytes();
  }
  /**

This is invalid java because two methods are named getXBytes.

Interestingly, protoc does seem to handle similar name collisions for repeated fields, as described in the documentation: "If another non-repeated field has a name that conflicts with one of the repeated field’s generated methods, both field names will have their protobuf field number appended to the end.".

This seems like an inconsistency (or a bug). Did I miss something here? Is there a recommended way to handle such name conflicts?

Any insights or clarification would be greatly appreciated!