#!/usr/bin/perl
use strict;
use warnings;
# A tool to build a perl release tarball
# Very basic but functional - if you're on a unix system.
#
# If you're on Win32 then it should still work, but various Unix command-line
# tools will need to be available somewhere. An obvious choice is to install
# Cygwin and ensure its 'bin' folder is on the PATH in the shell where you run
# this script. The Cygwin 'bin' folder needs to precede the Windows 'system32'
# folder so that Cygwin's 'find' command is found in preference to the Windows
# 'find' command. In addition to the commands installed by default, your Cygwin
# installation will need to contain at least the 'cpio' and '7z' commands.
# Finally, ensure that the 'awk' and '7z' commands
# are copies of 'gawk.exe' and 'lib\p7zip\7z.exe' respectively,
# rather than the links to them that only work in a Cygwin bash shell which
# they are by default.
#
# No matter how automated this gets, you'll always need to read
# and re-read pumpkin.pod and release_managers_guide.pod to
# check for things to be done at various stages of the process.
#
# Tim Bunce, June 1997
use ExtUtils::Manifest qw(fullcheck);
$ExtUtils::Manifest::Quiet = 1;
use Getopt::Std;
use Digest::SHA;
$|=1;
sub usage { die <<EOF; }
usage: $0 [ -r rootdir ] [-s suffix ] [ -x ] [ -n ]
    -r rootdir   directory under which to create the build dir and tarball
                 defaults to '..'
    -s suffix    suffix to append to the perl-x.y.z dir and tarball name
                 defaults to the concatenation of the local_patches entry
                 in patchlevel.h (or blank, if none)
    -x           make a .xz file in addition to a .gz file
    -n           do not make any tarballs, just the directory
    -c cleanup   perform a cleanup before building: clean git repo and target
                 directory/tarballs
    -e           Make the outputs be translated into EBCDIC.  (They can then
                 be downloaded directly to an EBCDIC platform without needing
                 any further translation.)
EOF
my %opts;
getopts('exncr:s:', \%opts) or usage;
@ARGV && usage;
my $relroot = defined $opts{r} ? $opts{r} : "..";
die "Must be in root of the perl source tree.\n"
        unless -f "./MANIFEST" and -f "patchlevel.h";
open PATCHLEVEL, '<', 'patchlevel.h' or die;
my @patchlevel_h = <PATCHLEVEL>;
close PATCHLEVEL;
my $patchlevel_h = join "", grep { /^#\s*define/ } @patchlevel_h;
print $patchlevel_h;
my $revision = $1 if $patchlevel_h =~ /PERL_REVISION\s+(\d+)/;
my $patchlevel = $1 if $patchlevel_h =~ /PERL_VERSION\s+(\d+)/;
my $subversion = $1 if $patchlevel_h =~ /PERL_SUBVERSION\s+(\d+)/;
die "Unable to parse patchlevel.h" unless $subversion >= 0;
my $vers = sprintf("%d.%d.%d", $revision, $patchlevel, $subversion);
# fetch list of local patches
my (@local_patches, @lpatch_tags, $lpatch_tags);
@local_patches = grep { !/^\s*,?NULL/ && ! /,"uncommitted-changes"/ }
                 grep { /^static.*local_patches/../^};/ }
                 @patchlevel_h;
@lpatch_tags   = map  {  /^\s*,"(\w+)/ } @local_patches;
$lpatch_tags   = join "-", @lpatch_tags;
my $perl = "perl-$vers";
my $reldir = "$perl";
$lpatch_tags = $opts{s} if defined $opts{s};
$reldir .= "-$lpatch_tags" if $lpatch_tags;
print "\nMaking a release for $perl in $relroot/$reldir\n\n";
cleanup($relroot, $reldir) if $opts{c};
print "Cross-checking the MANIFEST...\n";
my ($missfile, $missentry) = fullcheck();
@$missentry
    = grep {$_ !~ m!^\.(?:git|github|mailmap)! and $_ !~ m!(?:/|^)\.gitignore!} @$missentry;
if (@$missfile ) {
    warn "Can't make a release with MANIFEST files missing:\n";
    warn "\t".$_."\n" for (@$missfile);
}
if (@$missentry ) {
    warn "Can't make a release with files not listed in MANIFEST\n";
    warn "\t".$_."\n" for (@$missentry);
}
if ("@$missentry" =~ m/\.orig\b/) {
    # Handy listing of find command and .orig files from patching work.
    # I tend to run 'xargs rm' and copy and paste the file list.
    my $cmd = "find . -name '*.orig' -print";
    print "$cmd\n";
    system($cmd);
}
die "Aborted.\n" if @$missentry or @$missfile;
print "\n";
# VMS no longer has hardcoded version numbers descrip.mms
print "Creating $relroot/$reldir release directory...\n";
die "$relroot/$reldir release directory already exists [consider using -c]\n"   if -e "$relroot/$reldir";
die "$relroot/$reldir.tar.gz release file already exists [consider using -c]\n" if -e "$relroot/$reldir.tar.gz";
die "$relroot/$reldir.tar.xz release file already exists [consider using -c]\n" if $opts{x} && -e "$relroot/$reldir.tar.xz";
mkdir("$relroot/$reldir", 0755) or die "mkdir $relroot/$reldir: $!\n";
print "\n";
print "Copying files to release directory...\n";
# ExtUtils::Manifest maniread does not preserve the order
my $cmd = "awk '{print \$1}' MANIFEST | cpio -pdm $relroot/$reldir";
system($cmd) == 0
    or die "$cmd failed";
print "\n";
chdir "$relroot/$reldir" or die $!;
my @exe = map   { my ($f) = split; glob($f) }
          grep  { $_ !~ /\A#/ && $_ !~ /\A\s*\z/ }
          map   { split "\n" }
          do    { local (@ARGV, $/) = 'Porting/exec-bit.txt'; <> };
if ($opts{e}) {
    require './regen/charset_translations.pl';
    # Translation tables, so far only to 1047
    my @charset = grep { /1047/ } get_supported_code_pages();
    my $charset = $charset[0];
    my $a2e = get_a2n($charset);
    die "$0 must be run on an ASCII platform" if ord("A") != 65;
    print "Translating to EBCDIC...\n";
    open my $mani_fh, "<", "MANIFEST" or die "Can't read copied MANIFEST: $!";
    my @manifest = <$mani_fh>;  # Slurp in whole thing before the file gets trashed
    close $mani_fh or die "Couldn't close MANIFEST: $!";
    while (defined ($_ = shift @manifest)) {
        chomp;
        my $file = $_ =~ s/\s.*//r; # Rmv description to get just the file
                                    # name
        local $/; # slurp mode
        open my $fh, "+<:raw", $file or die "Can't read copied $file: $!";
        my $text = <$fh>;
        my $xlated = "";
        my $utf16_high = 0;
        my $utf16_low = 0;
        my $potential_BOM = substr($text, 0, 2);
        if ($potential_BOM eq "\xFE\xFF") {
            $utf16_high = 0;
            $utf16_low = 1;
            print STDERR "$file is UTF-16BE\n";
        }
        elsif ($potential_BOM eq "\xFF\xFE") {
            $utf16_high = 1;
            $utf16_low = 0;
            print STDERR "$file is UTF-16LE\n";
        }
        if ($utf16_high || $utf16_low) {
            my $len = length $text;
            die "Odd length in UTF-16 files: $file" if $len % 2;
            # Look 2 bytes at a time
            for (my $i = 0; $i < $len; $i+=2) {
                my $cur = substr($text, $i, 2);
                # If the code point's high byte is 0, it means the code point
                # itself is 00-FF, so want native value of it.
                if (substr($cur, $utf16_high, 1) eq "\0") {
                    # Just substitute the translated native value
                    my $low_byte = substr($cur, $utf16_low, 1);
                    $low_byte = chr $a2e->[ord $low_byte];
                    substr($cur, $utf16_low, 1) = $low_byte;
                }
                $xlated .= $cur;
            }
        }
        elsif (-B $file) {   # Binary files aren't translated
            print STDERR "$file is binary\n";
            close $fh or die "Couldn't close $file: $!";
            next;
        }
        else {
            if (   ! utf8::decode($text)
                || $text =~ / ^ [[:ascii:][:cntrl:]]* $ /x)
            {
                # Here, either $text isn't legal UTF-8; or it is, but it
                # consists entirely of one of the 160 ASCII and control
                # characters whose EBCDIC representation is the same whether
                # UTF-EBCDIC or not.  This means we just translate
                # byte-by-byte from Latin1 to EBCDIC.
                $xlated = ($text =~ s/(.)/chr $a2e->[ord $1]/rsge);
            }
            else {
                # Here, $text is legal UTF-8, and the representation of some
                # character(s) in it it matters if is encoded in UTF-EBCDIC or
                # not.  Also, the decode caused $text to now be viewed as
                # UTF-8 characters instead of the input bytes.  We convert to
                # UTF-EBCDIC.
                $xlated = ($text =~ s/(.)/cp_2_utfbytes(ord $1, $charset)/rsge);
            }
        }
        # Overwrite the file with the translation
        truncate $fh, 0;
        seek $fh, 0, 0;
        print $fh $xlated;
        close $fh or die "Couldn't close $file: $!";
    }
}
print "Setting file permissions...\n";
system("find . -type f -print     | xargs chmod 0444");
system("find . -type d -print     | xargs chmod 0755");
system("chmod +x @exe") == 0
    or die "system: $!";
# MANIFEST may be resorted, so needs to be writable
my @writables = qw(
    feature.h
    lib/feature.pm
    keywords.h
    keywords.c
    MANIFEST
    opcode.h
    opnames.h
    pp_proto.h
    proto.h
    embed.h
    embedvar.h
    overload.inc
    overload.h
    mg_vtable.h
    dist/Devel-PPPort/module2.c
    dist/Devel-PPPort/module3.c
    cpan/autodie/t/touch_me
    reentr.c
    reentr.h
    regcharclass.h
    regnodes.h
    warnings.h
    lib/warnings.pm
    win32/GNUmakefile
    win32/Makefile
    win32/config_H.gc
    win32/config_H.vc
    uconfig.h
);
my $out = `chmod u+w @writables 2>&1`;
if ($? != 0) {
    warn $out;
    if ($out =~ /no such file/i) {
        warn "Check that the files above still exist in the Perl core.\n";
        warn "If not, remove them from \@writables in Porting/makerel\n";
    }
    exit 1;
}
warn $out if $out;
chdir ".." or die $!;
exit if $opts{n};
my $src = (-e $perl) ? $perl : 'perl'; # 'perl' in maint branch
my $output_7z;
my $have_7z;
if (! $opts{e}) {
    print "Checking if you have 7z...\n";
    $output_7z = `7z 2>&1`;
    $have_7z = defined $output_7z && $output_7z =~ /7-Zip/;
}
print "Checking if you have advdef...\n";
my $output_advdef = `advdef --version 2>&1`;
my $have_advdef = defined $output_advdef && $output_advdef =~ /advancecomp/;
if (! $opts{e} && $have_7z) {
    print "Creating and compressing the tar.gz file with 7z...\n";
    $cmd = "tar cf - $reldir | 7z a -tgzip -mx9 -bd -si $reldir.tar.gz";
    system($cmd) == 0 or die "$cmd failed";
} else {
    print "Creating and compressing the tar.gz file...\n";
    my $extra_opts = "";
    if ($opts{e}) {
        print "(Using ustar format since is for an EBCDIC box)\n";
        $extra_opts = ' --format=ustar';
    }
    $cmd = "tar cf - $extra_opts $reldir | gzip --best > $reldir.tar.gz";
    system($cmd) == 0 or die "$cmd failed";
    if ($have_advdef) {
        print "Recompressing the tar.gz file with advdef...\n";
        $cmd = "advdef -z -4 $reldir.tar.gz";
        system($cmd) == 0 or die "$cmd failed";
    }
}
if ($opts{x}) {
    print "Creating and compressing the tar.xz file with xz...\n";
    $cmd = "tar cf - $reldir | xz -z -c > $reldir.tar.xz";
    system($cmd) == 0 or die "$cmd failed";
}
print "\n";
system("ls -ld $perl*");
print "\n";
my @files = glob "'$perl*.tar.*'";
for my $file (@files) {
    my $sha = Digest::SHA->new('sha256');
    $sha->addfile($file, 'b');
    print $sha->hexdigest . "  $file\n";
}
sub cleanup {
    my ( $relroot, $reldir ) = @_;
    require File::Path;
    my @cmds = (
        [ qw{make distclean} ],
        [ qw{git clean -dxf} ],
    );
    foreach my $cmd (@cmds) {
        print join( ' ', "Running:", @$cmd, "\n" );
        system @$cmd;
        die "fail to run ".(join(' ', @$cmd) ) unless $? == 0;
    }
    if ( -d "$relroot/$reldir" ) {
        print "Removing directory $relroot/$reldir\n";
        File::Path::rmtree("$relroot/$reldir");
    }
    # always clean both
    my @files = ( "$relroot/$reldir.tar.gz", "$relroot/$reldir.tar.xz" );
    foreach my $f ( @files ) {
        next unless -f $f;
        print "Removing file '$f'\n";
        unlink($f);
    }
    return;
}
1;