Problems uncompressing compressed jar files

Question

Hi all,
I am seeing an unusual issue where some entries don't have full header info. By this I mean, according to the ZIP spec, it is possible that the header info is stored before the file data, so you CAN get the right info. But sometimes, it can be stored as an EXT header, where it is some place AFTER the compressed file data.
So most of the time we are getting good size/comrpessed size info and can properly get the entry out of the zip file. But sometimes, we get -1 values and are not able to get the entry at all.
We are working on allowing a classloader to get classes out of an "embeded" zip/jar file. In other words, a jar file can contain embeded zip/jar files (only 1 level deep), and our classloader not only finds the classes of the main jar file (using the URLClassLoader capabilities, but if a class being looked for is not found in the main jar file, we then try to find it in any embeded jar/zip files. Finding the classes is fine, trying to create .class files out of the compressed data from an embeded jar/zip is the problem.
We are getting NegativeArraySizeExceptions thrown.
URLConnection connection = resURL.openConnection();
byte[] classData = new byte[connection.getContentLength()];
InputStream input = connection.getInputStream();
The resURL points to an embeded jar/zip within the outer jar file. The exception is thrown in the 2nd line above, when we try to create the byte[] array to hold the class data. The getContentLength() returns -1.
Any help would be appreciated. Thanks.

Answer 1

ZipInputStream only reads the local file header (LFH) of an entry if getNextEntry() is called, when the size and compressed size are not in the LFH becourse the sizes were unknow when the LFH was written the sizes are set to -1 and a flag is set to indicate that the entry has an EXT header after the data which contains the sizes.
So I think the only thing you could do at the moment it to write an extension of ZipInputStream that checks if the flag is set and then looks for the EXT header to get the sizes but the difficult part is to get the data, ZipInputStream uses a PushbackInputStream to read the data but that InputStream doesn't support the mark method.
Good luck.