] > The Apache Compress Antlib now requires Apache Commons Compress 1.4 and the public domain XZ for Java library if you want to use the XZ format. Commons Compress 1.4 is also required for the GNU, STAR and PAX formats of the tar task. Support for xz compression has been added. The gunzip and bunzip2 tasks as well as the corresponding resources now support uncompressing resources that contain multiple concatenated streams as a single stream. This behavior has to be enabled explicitly via the decompressConcatenated attribute. The tar task now supports the POSIX 1003.1/PAX way of storing long file names used by most modern tar implementations. The tar resources and the untar task read entries bigger than 8 GByte transparently now, if you use Commons Compress 1.4. The tar task writes entries bigger than 8 GByte using POSIX 1003.1/PAX headers or the star extension if you set the format attribute to "star", "gnu" or "pax". See the tar task's documentation of details. The tar and untar tasks as well as the tar related resources now allow the encoding for file names to be specified. The default is the platform's native encoding, it used to be ISO-8859-1 in Commons Compress 1.3 and thus the Compress Antlib 1.1. The Apache Compress Antlib now requires Apache Commons Compress 1.3 which in turn requires Java5 at runtime. This change allows the Compress Antlib to support Zip64 extensions. Support for pack200 compression has been added. The new pack200normalize task allows you to prepare JAR archives so they can be signed and compressed with pack200 and the signature will still be valid after uncompressing them again. Read-only support for the Unix dump format has been added in the form of an <undump> task and corresponding <dumpentry> and <dumpfileset> resources. A new attribute zip64Mode provides finer grained control over whether Zip64 extensions will be used in the <zip> task. The <zip> task has a new attribute zip64Mode that controls whether Zip64 extended information is written to the archive. The <ar> task now optionally suppports writing entries with names longer than 16 characters using the BSD variant of AR archives. The unarchiving tasks and fileset types share a new attribute skipUnreadableEntries that can be used to ignore archive entries that Apache Commons Compress knows it cannot handle (like encrypted ZIP entries or sparse files in GNU tar archives). The preserveLeadingSlashes attribute of the <tar> task now works as expected.