java/util/zip/ZipOutputStream.java has an implementation of UTF8 encoding
that does not take into account surrogates:

private static byte[] getUTF8Bytes(String s) {
  char[] c = s.toCharArray();
  int len = c.length;
  // Count the number of encoded bytes...
  int count = 0;
  for (int i = 0; i < len; i++) {
      int ch = c[i];
      if (ch <= 0x7f) {
  	count++;
      } else if (ch <= 0x7ff) {
  	count += 2;
      } else {
  	count += 3;
      }
  }
  // Now return the encoded bytes...
  byte[] b = new byte[count];
  int off = 0;
  for (int i = 0; i < len; i++) {
      int ch = c[i];
      if (ch <= 0x7f) {
  	b[off++] = (byte)ch;
      } else if (ch <= 0x7ff) {
  	b[off++] = (byte)((ch >> 6) | 0xc0);
  	b[off++] = (byte)((ch & 0x3f) | 0x80);
      } else {
  	b[off++] = (byte)((ch >> 12) | 0xe0);
  	b[off++] = (byte)(((ch >> 6) & 0x3f) | 0x80);
  	b[off++] = (byte)((ch & 0x3f) | 0x80);
      }
  }
  return b;
}
-----------------------------------------------------------
Also, Norbert Lindenberg noted:

I did notice another thing that looks fishy: 
src/share/native/java/util/zip/ZipFile.c has calls to the JNI routines 
GetStringUTFLength and GetStringUTFRegion, apparently also to handle 
file names. These are probably wrong, because JNI uses modified UTF-8 
and zip/jar files should use standard UTF-8.