ZipInputStream.readEnd currently expects 8-byte size and csize fields in the data descriptor header if and only if the actually numbers of compressed or uncompressed bytes exceeds the Zip64 limit of 0XFFFFFFFF.

This prevents parsing of small zip64 entries, such as the ones created using:

echo hello | zip -fd > hello.zip

The APPNOTE.txt specification indicates that such files are valid, so ZipInputStream should be able to parse them:

When extracting, if the zip64 extended information extra 
field is present for the file the compressed and 
uncompressed sizes will be 8 byte values.

The solution is to update ZipOutputStream.readEnd to check for the presence of Zip64 extra information fields in the LOC.