[PATCH] Possible cleanups for local-pull.c
[gitweb.git] / git-cvsimport-script
index e480a50f4888369b1b1f7c27671ea803d4e4279d..e3a8e584a562e8e50c5d7316acc653babafcd8ba 100755 (executable)
 use strict;
 use warnings;
 use Getopt::Std;
+use File::Spec;
+use File::Temp qw(tempfile);
 use File::Path qw(mkpath);
 use File::Basename qw(basename dirname);
 use Time::Local;
-use POSIX qw(strftime);
+use IO::Socket;
+use IO::Pipe;
+use POSIX qw(strftime dup2);
 
 $SIG{'PIPE'}="IGNORE";
 $ENV{'TZ'}="UTC";
 
-our($opt_h,$opt_o,$opt_v);
+our($opt_h,$opt_o,$opt_v,$opt_k,$opt_d,$opt_p,$opt_C,$opt_z,$opt_i,$opt_s,$opt_m,$opt_M);
 
 sub usage() {
        print STDERR <<END;
-Usage: ${\basename $0}  -- fetch/update GIT from CVS
-       [ -h (help) ]
-          [ -v (verbose) ] [ -o name-of-HEAD ]
-       CVS_archive GIT_dest
+Usage: ${\basename $0}     # fetch/update GIT from CVS
+       [ -o branch-for-HEAD ] [ -h ] [ -v ] [ -d CVSROOT ]
+       [ -p opts-for-cvsps ] [ -C GIT_repository ] [ -z fuzz ]
+       [ -i ] [ -k ] [-s subst] [ -m ] [ -M regex] [ CVS_module ]
 END
        exit(1);
 }
 
-getopts("hqvo:") or usage();
+getopts("hivmko:d:p:C:z:s:M:") or usage();
 usage if $opt_h;
 
-@ARGV == 2 or usage();
+@ARGV <= 1 or usage();
+
+if($opt_d) {
+       $ENV{"CVSROOT"} = $opt_d;
+} elsif(-f 'CVS/Root') {
+       open my $f, '<', 'CVS/Root' or die 'Failed to open CVS/Root';
+       $opt_d = <$f>;
+       chomp $opt_d;
+       close $f;
+       $ENV{"CVSROOT"} = $opt_d;
+} elsif($ENV{"CVSROOT"}) {
+       $opt_d = $ENV{"CVSROOT"};
+} else {
+       die "CVSROOT needs to be set";
+}
 $opt_o ||= "origin";
+$opt_s ||= "-";
+my $git_tree = $opt_C;
+$git_tree ||= ".";
+
+my $cvs_tree;
+if ($#ARGV == 0) {
+       $cvs_tree = $ARGV[0];
+} elsif (-f 'CVS/Repository') {
+       open my $f, '<', 'CVS/Repository' or 
+           die 'Failed to open CVS/Repository';
+       $cvs_tree = <$f>;
+       chomp $cvs_tree;
+       close $f;
+} else {
+       usage();
+}
 
-my($cvs_tree, $git_tree) = @ARGV;
+our @mergerx = ();
+if ($opt_m) {
+       @mergerx = ( qr/\W(?:from|of|merge|merging|merged) (\w+)/i );
+}
+if ($opt_M) {
+       push (@mergerx, qr/$opt_M/);
+}
 
 select(STDERR); $|=1; select(STDOUT);
 
 
 package CVSconn;
 # Basic CVS dialog.
-# We're only interested in downloading files.
+# We're only interested in connecting and downloading, so ...
+
+use File::Spec;
+use File::Temp qw(tempfile);
+use POSIX qw(strftime dup2);
 
 sub new {
        my($what,$repo,$subdir) = @_;
@@ -75,7 +119,7 @@ sub conn {
        if($repo =~ s/^:pserver:(?:(.*?)(?::(.*?))?@)?([^:\/]*)(?::(\d*))?//) {
                my($user,$pass,$serv,$port) = ($1,$2,$3,$4);
                $user="anonymous" unless defined $user;
-               my $rr2;
+               my $rr2 = "-";
                unless($port) {
                        $rr2 = ":pserver:$user\@$serv:$repo";
                        $port=2401;
@@ -98,7 +142,6 @@ sub conn {
                }
                $pass="A" unless $pass;
 
-               use IO::Socket;
                my $s = IO::Socket::INET->new(PeerHost => $serv, PeerPort => $port);
                die "Socket to $serv: $!\n" unless defined $s;
                $s->write("BEGIN AUTH REQUEST\n$repo\n$user\n$pass\nEND AUTH REQUEST\n")
@@ -113,21 +156,40 @@ sub conn {
                }
                $self->{'socketo'} = $s;
                $self->{'socketi'} = $s;
-       } else { # local: Fork off our own cvs server.
-               use IO::Pipe;
+       } else { # local or ext: Fork off our own cvs server.
                my $pr = IO::Pipe->new();
                my $pw = IO::Pipe->new();
                my $pid = fork();
                die "Fork: $!\n" unless defined $pid;
+               my $cvs = 'cvs';
+               $cvs = $ENV{CVS_SERVER} if exists $ENV{CVS_SERVER};
+               my $rsh = 'rsh';
+               $rsh = $ENV{CVS_RSH} if exists $ENV{CVS_RSH};
+
+               my @cvs = ($cvs, 'server');
+               my ($local, $user, $host);
+               $local = $repo =~ s/:local://;
+               if (!$local) {
+                   $repo =~ s/:ext://;
+                   $local = !($repo =~ s/^(?:([^\@:]+)\@)?([^:]+)://);
+                   ($user, $host) = ($1, $2);
+               }
+               if (!$local) {
+                   if ($user) {
+                       unshift @cvs, $rsh, '-l', $user, $host;
+                   } else {
+                       unshift @cvs, $rsh, $host;
+                   }
+               }
+
                unless($pid) {
                        $pr->writer();
                        $pw->reader();
-                       use POSIX qw(dup2);
                        dup2($pw->fileno(),0);
                        dup2($pr->fileno(),1);
                        $pr->close();
                        $pw->close();
-                       exec("cvs","server");
+                       exec(@cvs);
                }
                $pw->writer();
                $pr->reader();
@@ -137,7 +199,7 @@ sub conn {
        $self->{'socketo'}->write("Root $repo\n");
 
        # Trial and error says that this probably is the minimum set
-       $self->{'socketo'}->write("Valid-responses ok error Valid-requests Mode M Mbinary E Checked-in Created Updated Merged Removed\n");
+       $self->{'socketo'}->write("Valid-responses ok error Valid-requests Mode M Mbinary E Checked-in Created Updated Merged Removed\n");
 
        $self->{'socketo'}->write("valid-requests\n");
        $self->{'socketo'}->flush();
@@ -165,15 +227,17 @@ sub _file {
        my($self,$fn,$rev) = @_;
        $self->{'socketo'}->write("Argument -N\n") or return undef;
        $self->{'socketo'}->write("Argument -P\n") or return undef;
-       # $self->{'socketo'}->write("Argument -ko\n") or return undef;
-       # -ko: Linus' version doesn't use it
+       # -kk: Linus' version doesn't use it - defaults to off
+       if ($opt_k) {
+           $self->{'socketo'}->write("Argument -kk\n") or return undef;
+       }
        $self->{'socketo'}->write("Argument -r\n") or return undef;
        $self->{'socketo'}->write("Argument $rev\n") or return undef;
        $self->{'socketo'}->write("Argument --\n") or return undef;
        $self->{'socketo'}->write("Argument $self->{'subdir'}/$fn\n") or return undef;
        $self->{'socketo'}->write("Directory .\n") or return undef;
        $self->{'socketo'}->write("$self->{'repo'}\n") or return undef;
-       $self->{'socketo'}->write("Sticky T1.1\n") or return undef;
+       # $self->{'socketo'}->write("Sticky T1.0\n") or return undef;
        $self->{'socketo'}->write("co\n") or return undef;
        $self->{'socketo'}->flush() or return undef;
        $self->{'lines'} = 0;
@@ -182,11 +246,11 @@ sub _file {
 sub _line {
        # Read a line from the server.
        # ... except that 'line' may be an entire file. ;-)
-       my($self) = @_;
+       my($self, $fh) = @_;
        die "Not in lines" unless defined $self->{'lines'};
 
        my $line;
-       my $res="";
+       my $res=0;
        while(defined($line = $self->readline())) {
                # M U gnupg-cvs-rep/AUTHORS
                # Updated gnupg-cvs-rep/
@@ -206,16 +270,18 @@ sub _line {
                        chomp $cnt;
                        die "Duh: Filesize $cnt" if $cnt !~ /^\d+$/;
                        $line="";
-                       $res="";
+                       $res=0;
                        while($cnt) {
                                my $buf;
                                my $num = $self->{'socketi'}->read($buf,$cnt);
                                die "Server: Filesize $cnt: $num: $!\n" if not defined $num or $num<=0;
-                               $res .= $buf;
+                               print $fh $buf;
+                               $res += $num;
                                $cnt -= $num;
                        }
                } elsif($line =~ s/^ //) {
-                       $res .= $line;
+                       print $fh $line;
+                       $res += length($line);
                } elsif($line =~ /^M\b/) {
                        # output, do nothing
                } elsif($line =~ /^Mbinary\b/) {
@@ -228,7 +294,8 @@ sub _line {
                                my $buf;
                                my $num = $self->{'socketi'}->read($buf,$cnt);
                                die "S: Mbinary $cnt: $num: $!\n" if not defined $num or $num<=0;
-                               $res .= $buf;
+                               print $fh $buf;
+                               $res += $num;
                                $cnt -= $num;
                        }
                } else {
@@ -238,6 +305,12 @@ sub _line {
                                return $res;
                        } elsif($line =~ s/^E //) {
                                # print STDERR "S: $line\n";
+                       } elsif($line =~ /^Remove-entry /i) {
+                               $line = $self->readline(); # filename
+                               $line = $self->readline(); # OK
+                               chomp $line;
+                               die "Unknown: $line" if $line ne "ok";
+                               return -1;
                        } else {
                                die "Unknown: $line\n";
                        }
@@ -248,24 +321,28 @@ sub file {
        my($self,$fn,$rev) = @_;
        my $res;
 
-       if ($self->_file($fn,$rev)) {
-               $res = $self->_line();
-               return $res if defined $res;
+       my ($fh, $name) = tempfile('gitcvs.XXXXXX', 
+                   DIR => File::Spec->tmpdir(), UNLINK => 1);
+
+       $self->_file($fn,$rev) and $res = $self->_line($fh);
+
+       if (!defined $res) {
+           # retry
+           $self->conn();
+           $self->_file($fn,$rev)
+                   or die "No file command send\n";
+           $res = $self->_line($fh);
+           die "No input: $fn $rev\n" unless defined $res;
        }
+       close ($fh);
 
-       # retry
-       $self->conn();
-       $self->_file($fn,$rev)
-               or die "No file command send\n";
-       $res = $self->_line();
-       die "No input: $fn $rev\n" unless defined $res;
-       return $res;
+       return ($name, $res);
 }
 
 
 package main;
 
-my $cvs = CVSconn->new(dirname($cvs_tree), basename($cvs_tree));
+my $cvs = CVSconn->new($opt_d, $cvs_tree);
 
 
 sub pdate($) {
@@ -300,25 +377,47 @@ sub pmode($) {
        return $m;
 }
 
-my $tmpcv = "/var/cache/cvs";
-
 sub getwd() {
        my $pwd = `pwd`;
        chomp $pwd;
        return $pwd;
 }
 
+
+sub get_headref($$) {
+    my $name    = shift;
+    my $git_dir = shift; 
+    my $sha;
+    
+    if (open(C,"$git_dir/refs/heads/$name")) {
+       chomp($sha = <C>);
+       close(C);
+       length($sha) == 40
+           or die "Cannot get head id for $name ($sha): $!\n";
+    }
+    return $sha;
+}
+
+
 -d $git_tree
        or mkdir($git_tree,0777)
        or die "Could not create $git_tree: $!";
 chdir($git_tree);
 
 my $last_branch = "";
+my $orig_branch = "";
+my $forward_master = 0;
 my %branch_date;
 
 my $git_dir = $ENV{"GIT_DIR"} || ".git";
 $git_dir = getwd()."/".$git_dir unless $git_dir =~ m#^/#;
 $ENV{"GIT_DIR"} = $git_dir;
+my $orig_git_index;
+$orig_git_index = $ENV{GIT_INDEX_FILE} if exists $ENV{GIT_INDEX_FILE};
+my ($git_ih, $git_index) = tempfile('gitXXXXXX', SUFFIX => '.idx',
+                                   DIR => File::Spec->tmpdir());
+close ($git_ih);
+$ENV{GIT_INDEX_FILE} = $git_index;
 unless(-d $git_dir) {
        system("git-init-db");
        die "Cannot init the GIT db at $git_tree: $?\n" if $?;
@@ -326,9 +425,38 @@ unless(-d $git_dir) {
        die "Cannot init an empty tree: $?\n" if $?;
 
        $last_branch = $opt_o;
+       $orig_branch = "";
 } else {
+       -f "$git_dir/refs/heads/$opt_o"
+               or die "Branch '$opt_o' does not exist.\n".
+                      "Either use the correct '-o branch' option,\n".
+                      "or import to a new repository.\n";
+
        $last_branch = basename(readlink("$git_dir/HEAD"));
-       die "Cannot read the last branch name: $!\n" unless $last_branch;
+       unless($last_branch) {
+               warn "Cannot read the last branch name: $! -- assuming 'master'\n";
+               $last_branch = "master";
+       }
+       $orig_branch = $last_branch;
+       if (-f "$git_dir/CVS2GIT_HEAD") {
+               die <<EOM;
+CVS2GIT_HEAD exists.
+Make sure your working directory corresponds to HEAD and remove CVS2GIT_HEAD.
+You may need to run
+
+    git-read-tree -m -u CVS2GIT_HEAD HEAD
+EOM
+       }
+       system('cp', "$git_dir/HEAD", "$git_dir/CVS2GIT_HEAD");
+
+       $forward_master =
+           $opt_o ne 'master' && -f "$git_dir/refs/heads/master" &&
+           system('cmp', '-s', "$git_dir/refs/heads/master", 
+                               "$git_dir/refs/heads/$opt_o") == 0;
+
+       # populate index
+       system('git-read-tree', $last_branch);
+       die "read-tree failed: $?\n" if $?;
 
        # Get the last import timestamps
        opendir(D,"$git_dir/refs/heads");
@@ -355,7 +483,10 @@ unless(-d $git_dir) {
 my $pid = open(CVS,"-|");
 die "Cannot fork: $!\n" unless defined $pid;
 unless($pid) {
-       exec("cvsps","-A","--cvs-direct",$cvs_tree);
+       my @opt;
+       @opt = split(/,/,$opt_p) if defined $opt_p;
+       unshift @opt, '-z', $opt_z if defined $opt_z;
+       exec("cvsps",@opt,"-u","-A","--cvs-direct",'--root',$opt_d,$cvs_tree);
        die "Could not start cvsps: $!\n";
 }
 
@@ -382,10 +513,29 @@ my($patchset,$date,$author,$branch,$ancestor,$tag,$logmsg);
 my(@old,@new);
 my $commit = sub {
        my $pid;
-       system("git-update-cache","--force-remove","--",@old) if @old;
-       die "Cannot remove files: $?\n" if $?;
-       system("git-update-cache","--add","--",@new) if @new;
-       die "Cannot add files: $?\n" if $?;
+       while(@old) {
+               my @o2;
+               if(@old > 55) {
+                       @o2 = splice(@old,0,50);
+               } else {
+                       @o2 = @old;
+                       @old = ();
+               }
+               system("git-update-cache","--force-remove","--",@o2);
+               die "Cannot remove files: $?\n" if $?;
+       }
+       while(@new) {
+               my @n2;
+               if(@new > 12) {
+                       @n2 = splice(@new,0,10);
+               } else {
+                       @n2 = @new;
+                       @new = ();
+               }
+               system("git-update-cache","--add",
+                       (map { ('--cacheinfo', @$_) } @n2));
+               die "Cannot add files: $?\n" if $?;
+       }
 
        $pid = open(C,"-|");
        die "Cannot fork: $!" unless defined $pid;
@@ -409,15 +559,36 @@ my $commit = sub {
                print "Parent ID $parent\n" if $opt_v;
        }
 
-       pipe(I,O)
-               or die "No pipe: $!\n";
-       $pid = open(C,"|-");
-       die "Cannot fork: $!" unless defined $pid;
+       my $pr = IO::Pipe->new() or die "Cannot open pipe: $!\n";
+       my $pw = IO::Pipe->new() or die "Cannot open pipe: $!\n";
+       $pid = fork();
+       die "Fork: $!\n" unless defined $pid;
        unless($pid) {
-               close(I);
-               open(STDOUT,">&O");
+               $pr->writer();
+               $pw->reader();
+               dup2($pw->fileno(),0);
+               dup2($pr->fileno(),1);
+               $pr->close();
+               $pw->close();
+
                my @par = ();
                @par = ("-p",$parent) if $parent;
+
+               # loose detection of merges
+               # based on the commit msg
+               foreach my $rx (@mergerx) {
+                       if ($logmsg =~ $rx) {
+                               my $mparent = $1;
+                               if ($mparent eq 'HEAD') { $mparent = $opt_o };
+                               if ( -e "$git_dir/refs/heads/$mparent") {
+                                       $mparent = get_headref($mparent, $git_dir);
+                                       push @par, '-p', $mparent;
+                                       # printing here breaks import # 
+                                       # # print "Merge parent branch: $mparent\n" if $opt_v;
+                               }
+                       } 
+               }
+
                exec("env",
                        "GIT_AUTHOR_NAME=$author",
                        "GIT_AUTHOR_EMAIL=$author",
@@ -428,17 +599,26 @@ my $commit = sub {
                        "git-commit-tree", $tree,@par);
                die "Cannot exec git-commit-tree: $!\n";
        }
-       close(O);
-       print C $logmsg
+       $pw->writer();
+       $pr->reader();
+
+       # compatibility with git2cvs
+       substr($logmsg,32767) = "" if length($logmsg) > 32767;
+       $logmsg =~ s/[\s\n]+\z//;
+
+       print $pw "$logmsg\n"
                or die "Error writing to git-commit-tree: $!\n";
-       close(C)
-               or die "Error running git-commit-tree: $?\n";
-       print "Committed patch $patchset ($branch)\n" if $opt_v;
-       chomp(my $cid = <I>);
+       $pw->close();
+
+       print "Committed patch $patchset ($branch ".strftime("%Y-%m-%d %H:%M:%S",gmtime($date)).")\n" if $opt_v;
+       chomp(my $cid = <$pr>);
        length($cid) == 40
                or die "Cannot get commit id ($cid): $!\n";
        print "Commit ID $cid\n" if $opt_v;
-       close(I);
+       $pr->close();
+
+       waitpid($pid,0);
+       die "Error running git-commit-tree: $?\n" if $?;
 
        open(C,">$git_dir/refs/heads/$branch")
                or die "Cannot open branch $branch for update: $!\n";
@@ -456,9 +636,6 @@ my $commit = sub {
                        or die "Cannot write tag $branch: $!\n";
                print "Created tag '$tag' on '$branch'\n" if $opt_v;
        }
-
-       @old = ();
-       @new = ();
 };
 
 while(<CVS>) {
@@ -485,12 +662,13 @@ while(<CVS>) {
                $state = 4;
        } elsif($state == 4 and s/^Branch:\s+//) {
                s/\s+$//;
+               s/[\/]/$opt_s/g;
                $branch = $_;
                $state = 5;
        } elsif($state == 5 and s/^Ancestor branch:\s+//) {
                s/\s+$//;
                $ancestor = $_;
-               $ancestor = $opt_o if $ancestor == "HEAD";
+               $ancestor = $opt_o if $ancestor eq "HEAD";
                $state = 6;
        } elsif($state == 5) {
                $ancestor = undef;
@@ -511,7 +689,7 @@ while(<CVS>) {
                $branch = $opt_o if $branch eq "HEAD";
                if(defined $branch_date{$branch} and $branch_date{$branch} >= $date) {
                        # skip
-                       print "skip patchset $patchset: $date before $branch_date{$branch}\n";
+                       print "skip patchset $patchset: $date before $branch_date{$branch}\n" if $opt_v;
                        $state = 11;
                        next;
                }
@@ -539,33 +717,40 @@ while(<CVS>) {
                                or die "Could not write branch $branch: $!";
                }
                if(($ancestor || $branch) ne $last_branch) {
-                       system("git-read-tree -m -u $last_branch $branch");
-                       die "read-tree $branch failed at $branch: $?\n" if $?;
-               }
-               if($branch ne $last_branch) {
-                       unlink("$git_dir/HEAD");
-                       symlink("refs/heads/$branch","$git_dir/HEAD");
-                       $last_branch = $branch;
+                       print "Switching from $last_branch to $branch\n" if $opt_v;
+                       system("git-read-tree", $branch);
+                       die "read-tree failed: $?\n" if $?;
                }
+               $last_branch = $branch if $branch ne $last_branch;
                $state = 9;
        } elsif($state == 8) {
                $logmsg .= "$_\n";
-       } elsif($state == 9 and /^\s+(\S+):(?:INITIAL|\d(?:\.\d+)+)->(\d(?:\.\d+)+)\s*$/) {
+       } elsif($state == 9 and /^\s+(.+?):(INITIAL|\d+(?:\.\d+)+)->(\d+(?:\.\d+)+)\s*$/) {
 #      VERSION:1.96->1.96.2.1
+               my $init = ($2 eq "INITIAL");
                my $fn = $1;
-               my $data = $cvs->file($fn,$2);
-               print "Update $fn: ".length($data)." bytes.\n";
-               mkpath(dirname($fn),$opt_v);
-               open(F,"> ./$fn")
-                       or die "Cannot create '$fn': $!\n";
-               print F $data
-                       or die "Cannot write to '$fn': $!\n";
-               close(F)
-                       or die "Cannot write to '$fn': $!\n";
-               chmod(pmode($cvs->{'mode'}), $fn);
-               push(@new,$fn); # may be resurrected!
-       } elsif($state == 9 and /^\s+(\S+):\d(?:\.\d+)+->(\d(?:\.\d+)+)\(DEAD\)\s*$/) {
-               push(@old,$1);
+               my $rev = $3;
+               $fn =~ s#^/+##;
+               my ($tmpname, $size) = $cvs->file($fn,$rev);
+               if($size == -1) {
+                       push(@old,$fn);
+                       print "Drop $fn\n" if $opt_v;
+               } else {
+                       print "".($init ? "New" : "Update")." $fn: $size bytes\n" if $opt_v;
+                       open my $F, '-|', "git-hash-object -w $tmpname"
+                               or die "Cannot create object: $!\n";
+                       my $sha = <$F>;
+                       chomp $sha;
+                       close $F;
+                       my $mode = pmode($cvs->{'mode'});
+                       push(@new,[$mode, $sha, $fn]); # may be resurrected!
+               }
+               unlink($tmpname);
+       } elsif($state == 9 and /^\s+(.+?):\d+(?:\.\d+)+->(\d+(?:\.\d+)+)\(DEAD\)\s*$/) {
+               my $fn = $1;
+               $fn =~ s#^/+##;
+               push(@old,$fn);
+               print "Delete $fn\n" if $opt_v;
        } elsif($state == 9 and /^\s*$/) {
                $state = 10;
        } elsif(($state == 9 or $state == 10) and /^-+$/) {
@@ -581,4 +766,33 @@ while(<CVS>) {
 }
 &$commit() if $branch and $state != 11;
 
-print "DONE\n" if $opt_v;
+unlink($git_index);
+
+if (defined $orig_git_index) {
+       $ENV{GIT_INDEX_FILE} = $orig_git_index;
+} else {
+       delete $ENV{GIT_INDEX_FILE};
+}
+
+# Now switch back to the branch we were in before all of this happened
+if($orig_branch) {
+       print "DONE\n" if $opt_v;
+       system("cp","$git_dir/refs/heads/$opt_o","$git_dir/refs/heads/master")
+               if $forward_master;
+       unless ($opt_i) {
+               system('git-read-tree', '-m', '-u', 'CVS2GIT_HEAD', 'HEAD');
+               die "read-tree failed: $?\n" if $?;
+       }
+} else {
+       $orig_branch = "master";
+       print "DONE; creating $orig_branch branch\n" if $opt_v;
+       system("cp","$git_dir/refs/heads/$opt_o","$git_dir/refs/heads/master")
+               unless -f "$git_dir/refs/heads/master";
+       unlink("$git_dir/HEAD");
+       symlink("refs/heads/$orig_branch","$git_dir/HEAD");
+       unless ($opt_i) {
+               system('git checkout');
+               die "checkout failed: $?\n" if $?;
+       }
+}
+unlink("$git_dir/CVS2GIT_HEAD");