#!/usr/bin/perl

open( Z, "test.zip" );

sub error {
    print "The file is corrupt.\n";
    exit;
}

sub readstr {
    my $received;
    error() if eof Z;
    $received .= getc(Z) . " " for ( 1 .. $_[0] );
    return $received;
}

sub readint {
    my $received = 0;
    error() if eof Z;
    $received = $received * 255 + ord( getc(Z) ) for ( 1 .. $_[0] );
    return $received;
}

while ( !eof Z ) {
    my $head     = readstr(4);    # PK^C^D
    my $versions = readstr(4);
    # ...
    my $filenamelength = readint(2);
    # ...
    # Parse the rest of this file
    # Until we get an error or go on to the next file header
}

close(Z);
[download]

You've hit on the key -- that the files are not connected. Looking at the code, Archive::Zip appears to always first access the central directory information, which is at the end of the file. For files that are not fully sent, that never works since it's the last part of the file that's missing. It makes sense to build the code around the central directory -- it's surely faster than parsing the entire zip file to get the pieces that are available. So I think a recovery method would have to try and piece things together the slow way as you state.

All my attempts with existing Archive::Zip methods result in errors and no data.

Code, errors? Did you try Low-level member data reading?

I figured the key is probably in the low level routines. I didn't get too far with those (yet anyway). The issue is that the low level routines are member oriented. So I first have to know what members are in the zip file. In order to get members, you first have to read an existing zip file. The read method fails. After that failure none of the member methods return anything.

Next I was going to take a corrupt zip I was able to read using one of the many Windows zip readers that handles corrupt files and try pulling members it tells me exist using the low level methods -- bypass the necessary step of having to first identify the members for now.