#!/usr/bin/perl
# Clean all non-deterministric fields in a newc cpio file
#
# Items fixed:
# Files are sorted by name
# Inode numbers are based on the hash of the filename
# File timestamp is set to 1970-01-01T00:00:00
# uid/gid are set to root
# check field is zeroed
# nlinks is set to zero, since the filesystem manages it
#
use warnings;
use strict;
use Data::Dumper;
use Digest::MD5 'md5_hex';

#	   struct cpio_newc_header {
#		   char    c_magic[6]; -6
#		   char    c_ino[8]; -- set to a monotonic value 0
#		   char    c_mode[8]; 8
#		   char    c_uid[8]; 16
#		   char    c_gid[8];  24
#		   char    c_nlink[8]; 32
#		   char    c_mtime[8]; 40 -- set to zero
#		   char    c_filesize[8]; 48
#		   char    c_devmajor[8]; 56
#		   char    c_devminor[8]; 64
#		   char    c_rdevmajor[8]; 72
#		   char    c_rdevminor[8]; 80
#		   char    c_namesize[8]; 88
#		   char    c_check[8]; 96
#	   }; // 104
# followed by namesize bytes of name (padded to be a multiple of 4)
# followed dby filesize bytes of file (padded to be a multiple of 4)

# Read the entire file at once
undef $/;

# Generate a map of all of the files in the cpio archive
# This will also merge multiple cpio files
my %entries;
my $trailer;

while(<>)
{
	for(my $i = 0 ; $i < length $_ ; )
	{
		my $magic = substr($_, $i, 6);
		if ($magic ne "070701")
		{
			die "$ARGV: offset $i: invalid magic '$magic'\n";
		}

		my $namesize = substr($_, $i + 6+88, 8);
		my $filesize = substr($_, $i + 6+48, 8);

		if ($namesize =~ /[^0-9A-Fa-f]/)
		{
			die "$ARGV: offset $i: invalid characters in namesize '$namesize'\n";
		}

		if ($filesize =~ /[^0-9A-Fa-f]/)
		{
			die "$ARGV: offset $i: invalid characters in filesize '$filesize'\n";
		}

		# Convert them to hex
		$namesize = hex $namesize;
		$filesize = hex $filesize;

		#print STDERR "name: '$namesize', filesize: '$filesize'\n";

		my $name = substr($_, $i + 6+104, $namesize);
		#print STDERR Dumper($name);

		# Align the header size to be a multiple of four bytes
		my $entry_size = (6+104 + $namesize + 3) & ~3;
		$entry_size += ($filesize + 3) & ~3;

		my $entry = substr($_, $i, $entry_size);
		$i += $entry_size;

		if ($name =~ /^TRAILER!!!/)
		{
			$trailer = $entry;
			last;
		}

		$entries{$name} = $entry;
	}

	die "$ARGV: No trailer!\n" unless $trailer;
}

# Apply the cleaning to each one
for my $filename (sort keys %entries)
{
	my $entry = $entries{$filename};
	my $zero = sprintf "%08x", 0;

	# inodes are hashed to be deterministic
	# and hopefully not colliding
	my $md5 = md5_hex($filename);
	my $d0 = hex substr($md5,  0, 8) ;
	my $d1 = hex substr($md5,  8, 8) ;
	my $d2 = hex substr($md5, 16, 8) ;
	my $d3 = hex substr($md5, 24, 8) ;
	my $hash = sprintf "%08x", $d0 ^ $d1 ^ $d2 ^ $d3;
	
	#warn "$filename: $md5 -> $hash\n";
	substr($entry, 6 + 0, 8) = $hash;

	# set timestamps to zero
	substr($entry, 6 + 40, 8) = $zero;

	# remove group/user permissions, leaving only
	# the owner bits intact.
	my $mode = hex substr($entry, 6 + 8, 8);
	$mode &= ~0077;
	#$mode |= $mode >> 3 | $mode >> 6;
	substr($entry, 6 + 8, 8) = sprintf "%08X", $mode;

	# set uid/gid to zero
	substr($entry, 6 + 16, 8) = $zero;
	substr($entry, 6 + 24, 8) = $zero;

	# zero out the nlinks, since it is managed by the real fs
	substr($entry, 6 + 32, 8) = $zero;

	# set the device major/minor to zero
	substr($entry, 6 + 56, 8) = $zero;
	substr($entry, 6 + 64, 8) = $zero;

	# set check to zero
	substr($entry, 6 + 96, 8) = $zero;

	$entries{$filename} = $entry;
}


# Print them in sorted order
for my $filename (sort keys %entries)
{
	my $entry = $entries{$filename};
	print $entry;
}

# Output the trailer to mark the end of the archive
print $trailer;
__END__