#! /usr/bin/perl -w
#
# Written by Adam Byrtek <alpha@debian.org>, 2002
# Rewritten by David Sterba <dave@jikos.cz>, 2009
#
# Extfs to handle patches in context and unified diff format.
# Known issues: When name of file to patch is modified during editing,
# hunk is duplicated on copyin. It is unavoidable.

use bytes;
use strict;
use POSIX;
use File::Temp 'tempfile';

# standard binaries
my $lzip = 'lzip';
my $lz4 = 'lz4';
my $lzma = 'lzma';
my $xz = 'xz';
my $bzip = 'bzip2';
my $gzip = 'gzip';
my $fileutil = 'file -b';

# date parsing requires Date::Parse from TimeDate module
my $parsedates = eval 'require Date::Parse';

# regular expressions
my $unified_header=qr/^--- .*\t.*\n\+\+\+ .*\t.*\n$/;
my $unified_extract=qr/^--- ([^\t]+).*\n\+\+\+ ([^\t]+)\s*(.*)\n/;
my $unified_header2=qr/^--- .*\n\+\+\+ .*\n$/;
my $unified_extract2=qr/^--- ([^\s]+).*\n\+\+\+ ([^\s]+)\s*(.*)\n/;
my $unified_contents=qr/^([+\-\\ \n]|@@ .* @@)/;
my $unified_hunk=qr/@@ -(\d+)(,(\d+))? \+(\d+)(,(\d+))? @@.*\n/;

my $context_header=qr/^\*\*\* .*\t.*\n--- .*\t.*\n$/;
my $context_extract=qr/^\*\*\* ([^\t]+).*\n--- ([^\t]+)\s*(.*)\n/;
my $context_header2=qr/^\*\*\* .*\n--- .*\n$/;
my $context_extract2=qr/^\*\*\* ([^\s]+).*\n--- ([^\s]+)\s*(.*)\n/;
my $context_contents=qr/^([!+\-\\ \n]|-{3} .* -{4}|\*{3} .* \*{4}|\*{15})/;

my $ls_extract_id=qr/^[^\s]+\s+[^\s]+\s+([^\s]+)\s+([^\s]+)/;
my $basename=qr|^(.*/)*([^/]+)$|;

sub patchfs_canonicalize_path ($) {
my ($fname) = @_;
$fname =~ s,/+,/,g;
$fname =~ s,(^|/)(?:\.?\./)+,$1,;
return $fname;
}

# output unix date in a mc-readable format
sub timef
{
my @time=localtime($_[0]);
return sprintf '%02d-%02d-%02d %02d:%02d', $time[4]+1, $time[3],
$time[5]+1900, $time[2], $time[1];
}

# parse given string as a date and return unix time
sub datetime
{
# in case of problems fall back to 0 in unix time
# note: str2time interprets some wrong values (eg. " ") as 'today'
if ($parsedates && defined (my $t=str2time($_[0]))) {
return timef($t);
}
return timef(time);
}

# print message on stderr and exit
sub error
{
print STDERR $_[0], "\n";
exit 1;
}

# (compressed) input
sub myin
{
my ($qfname)=(quotemeta $_[0]);

$_=`$fileutil $qfname`;
if (/^'*lz4/) {
return "$lz4 -dc $qfname";
} elsif (/^'*lzip/) {
return "$lzip -dc $qfname";
} elsif (/^'*lzma/) {
return "$lzma -dc $qfname";
} elsif (/^'*xz/) {
return "$xz -dc $qfname";
} elsif (/^'*bzip/) {
return "$bzip -dc $qfname";
} elsif (/^'*gzip/) {
return "$gzip -dc $qfname";
} else {
return "cat $qfname";
}
}

# (compressed) output
sub myout
{
my ($qfname,$append)=(quotemeta $_[0],$_[1]);
my ($sep) = $append ? '>>' : '>';

$_=`$fileutil $qfname`;
if (/^'*lz4/) {
return "$lz4 -c $sep $qfname";
} elsif (/^'*lzip/) {
return "$lzip -c $sep $qfname";
} elsif (/^'*lzma/) {
return "$lzma -c $sep $qfname";
} elsif (/^'*xz/) {
return "$xz -c $sep $qfname";
} elsif (/^'*bzip/) {
return "$bzip -c $sep $qfname";
} elsif (/^'*gzip/) {
return "$gzip -c $sep $qfname";
} else {
return "cat $sep $qfname";
}
}

# select diff filename conforming with rules found in diff.info
sub diff_filename
{
my ($fsrc,$fdst)= @_;
# TODO: can remove these two calls later
$fsrc = patchfs_canonicalize_path ($fsrc);
$fdst = patchfs_canonicalize_path ($fdst);
if (!$fdst && !$fsrc) {
error 'Index: not yet implemented';
} elsif (!$fsrc || $fsrc eq '/dev/null') {
return ($fdst,'PATCH-CREATE/');
} elsif (!$fdst || $fdst eq '/dev/null') {
return ($fsrc,'PATCH-REMOVE/');
} elsif (($fdst eq '/dev/null') && ($fsrc eq '/dev/null')) {
error 'Malformed diff, missing a sane filename';
} else {
# fewest path name components
if ($fdst=~s|/|/|g < $fsrc=~s|/|/|g) {
return ($fdst,'');
} elsif ($fdst=~s|/|/|g > $fsrc=~s|/|/|g) {
return ($fsrc,'');
} else {
# shorter base name
if (($fdst=~/$basename/o,length $2) < ($fsrc=~/$basename/o,length $2)) {
return ($fdst,'');
} elsif (($fdst=~/$basename/o,length $2) > ($fsrc=~/$basename/o,length $2)) {
return ($fsrc,'');
} else {
# shortest names
if (length $fdst < length $fsrc) {
return ($fdst,'');
} else {
return ($fsrc,'');
}
}
}
}
}

# IN: diff "archive" name
# IN: file handle for output; STDIN for list, tempfile else
# IN: filename to watch (for: copyout, rm), '' for: list
# IN: remove the file?
# true - ... and print out the rest
# false - ie. copyout mode, print just the file
sub parse($$$$)
{
my $archive=quotemeta shift;
my $fh=shift;
my $file=shift;
my $rmmod=shift;
my ($state,$fsize,$time);
my ($f,$fsrc,$fdst,$prefix);
my ($unified,$context);
my ($skipread, $filetoprint, $filefound);
my ($h_add,$h_del,$h_ctx); # hunk line counts
my ($h_r1,$h_r2); # hunk ranges
my @outsrc; # if desired ...
my @outdst;
my $line;
my %fmap_size=();
my %fmap_time=();

import Date::Parse if ($parsedates && $file eq '');

$line=1;
$state=0; $fsize=0; $f='';
$filefound=0;
while ($skipread || ($line++,$_=<I>)) {
$skipread=0;
if($state == 0) { # expecting comments
$unified=$context=0;
$unified=1 if (/^--- /);
$context=1 if (/^\*\*\* /);
if (!$unified && !$context) {
$filefound=0 if($file ne '' && $filetoprint);
# shortcut for rmmod xor filefound
# - in rmmod we print if not found
# - in copyout (!rmmod) we print if found
print $fh $_ if($rmmod != $filefound);
next;
}

if($file eq '' && $filetoprint) {
$fmap_size{"$prefix$f"}+=$fsize;
$fmap_time{"$prefix$f"}=$time;
}

# start of new file
$_ .=<I>; # steal next line, both formats
$line++;
if($unified) {
if(/$unified_header/o) {
($fsrc,$fdst,$time) = /$unified_extract/o;
} elsif(/$unified_header2/o) {
($fsrc,$fdst,$time) = /$unified_extract2/o;
} else {
error "Can't parse unified diff header";
}
} elsif($context) {
if(/$context_header/o) {
($fsrc,$fdst,$time) = /$context_extract/o;
} elsif(/$context_header2/o) {
($fsrc,$fdst,$time) = /$context_extract2/o;
} else {
error "Can't parse context diff header";
}
} else {
error "Unrecognized diff header";
}
$fsrc=patchfs_canonicalize_path($fsrc);
$fdst=patchfs_canonicalize_path($fdst);
if(wantarray) {
push @outsrc,$fsrc;
push @outdst,$fdst;
}
($f,$prefix)=diff_filename($fsrc,$fdst);
$filefound=($f eq $file);

$f="$f.diff";
$filetoprint=1;
$fsize=length;
print $fh $_ if($rmmod != $filefound);

$state=1;
} elsif($state == 1) { # expecting diff hunk headers, end of file or comments
if($unified) {
my ($a,$b,$c,$d);
($a,$b,$h_r1,$c,$d,$h_r2)=/$unified_hunk/o;
if(!defined($a) || !defined($c)) {
# hunk header does not come, a comment inside
# or maybe a new file, state 0 will decide
$skipread=1;
$state=0;
next;
}
$fsize+=length;
print $fh $_ if($rmmod != $filefound);
$h_r1=1 if(!defined($b));
$h_r2=1 if(!defined($d));
$h_add=$h_del=$h_ctx=0;
$state=2;
} elsif($context) {
if(!/$context_contents/o) {
$skipread=1;
$state=0;
next;
}
print $fh $_ if($rmmod != $filefound);
$fsize+=length;
}
} elsif($state == 2) { # expecting hunk contents
if($h_del + $h_ctx == $h_r1 && $h_add + $h_ctx == $h_r2) {
# hooray, end of hunk
# we optimistically ended with a hunk before but
# the line has been read already
$skipread=1;
$state=1;
next;
}
print $fh $_ if($rmmod != $filefound);
$fsize+=length;
my ($first)= /^(.)/;
if(ord($first) == ord('+')) { $h_add++; }
elsif(ord($first) == ord('-')) { $h_del++; }
elsif(ord($first) == ord(' ')) { $h_ctx++; }
elsif(ord($first) == ord('\\')) { 0; }
elsif(ord($first) == ord('@')) { error "Malformed hunk, header came too early"; }
else { error "Unrecognized character in hunk"; }
}
}
if($file eq '' && $filetoprint) {
$fmap_size{"$prefix$f"}+=$fsize;
$fmap_time{"$prefix$f"}=$time;
}

# use uid and gid from file
my ($uid,$gid)=(`ls -l $archive`=~/$ls_extract_id/o);

# flush all file names with cumulative file size
while(my ($fn, $fs) = each %fmap_size) {
printf $fh "-rw-r--r-- 1 %s %s %d %s %s\n", $uid, $gid, $fs, datetime($fmap_time{$fn}), $fn;
}

close($fh) if($file ne '');
return \(@outsrc, @outdst) if wantarray;
}

# list files affected by patch
sub list($) {
parse($_[0], *STDOUT, '', 0);
close(I);
}

# extract diff from patch
# IN: diff file to find
# IN: output file name
sub copyout($$) {
my ($file,$out)=@_;

$file=~s/^(PATCH-(CREATE|REMOVE)\/)?(.*)\.diff$/$3/;
$file = patchfs_canonicalize_path ($file);

open(FH, ">$out") or error("Cannot open output file");
parse('', *FH, $file, 0);
}

# remove diff(s) from patch
# IN: archive
# IN: file to delete
sub rm($$) {
my $archive=shift;
my ($tmp,$tmpname)=tempfile();

@_=map {scalar(s/^(PATCH-(CREATE|REMOVE)\/)?(.*)\.diff$/$3/,$_)} @_;

# just the first file for now
parse($archive, $tmp, $_[0], 1);
close I;

# replace archive
system("cat \Q$tmpname\E | " . myout($archive,0))==0
or error "Can't write to archive";
system("rm -f -- \Q$tmpname\E");
}

# append diff to archive
# IN: diff archive name
# IN: newly created file name in archive
# IN: the real source file
sub copyin($$$) {
# TODO: seems to be tricky. what to do?
# copyin of file which is already there may:
# * delete the original and copy only the new
# * just append the new hunks to the same file
# problems: may not be a valid diff, unmerged hunks
# * try to merge the two together
# ... but we do not want write patchutils again, right?
error "Copying files into diff not supported";
return;

my ($archive,$name,$src)=@_;

# in case we are appending another diff, we have
# to delete/merge all the files
open(DEVNULL, ">/dev/null");
open I, myin($src).'|';
my ($srclist,$dstlist)=parse($archive, *DEVNULL, '', 0);
close(I);
close(DEVNULL);
foreach(@$srclist) {
print("SRC: del $_\n");
}
foreach(@$dstlist) {
print("DST: del $_\n");
}
return;

# remove overwritten file
open I, myin($archive).'|';
rm ($archive, $name);
close I;

my $cmd1=myin("$src.diff");
my $cmd2=myout($archive,1);
system("$cmd1 | $cmd2")==0
or error "Can't write to archive";
}

my $fin = $ARGV[1];

# resolve symlink
while (-l $fin) {
$fin = readlink $fin;
}

if ($ARGV[0] eq 'list') {
open I, myin($fin).'|';
list ($fin);
exit 0;
} elsif ($ARGV[0] eq 'copyout') {
open I, myin($fin)."|";
copyout ($ARGV[2], $ARGV[3]);
exit 0;
} elsif ($ARGV[0] eq 'rm') {
open I, myin($fin)."|";
rm ($fin, $ARGV[2]);
exit 0;
} elsif ($ARGV[0] eq 'rmdir') {
exit 0;
} elsif ($ARGV[0] eq 'mkdir') {
exit 0;
} elsif ($ARGV[0] eq 'copyin') {
copyin ($fin, $ARGV[2], $ARGV[3]);
exit 0;
}
exit 1;