git-branch -D: make it work even when on a yet-to-be-born branch
[gitweb.git] / git-cvsimport.perl
index 60fc86a5beddd7ecf27256b3dee113c80eb8ecb8..4310dea1320efa47eee2861b4f886197ca713a2e 100755 (executable)
 use warnings;
 use Getopt::Std;
 use File::Spec;
-use File::Temp qw(tempfile);
+use File::Temp qw(tempfile tmpnam);
 use File::Path qw(mkpath);
 use File::Basename qw(basename dirname);
 use Time::Local;
 use IO::Socket;
 use IO::Pipe;
-use POSIX qw(strftime dup2 :errno_h);
+use POSIX qw(strftime dup2 ENOENT);
 use IPC::Open2;
 
 $SIG{'PIPE'}="IGNORE";
@@ -161,8 +161,22 @@ sub new {
 sub conn {
        my $self = shift;
        my $repo = $self->{'fullrep'};
-       if($repo =~ s/^:pserver:(?:(.*?)(?::(.*?))?@)?([^:\/]*)(?::(\d*))?//) {
-               my($user,$pass,$serv,$port) = ($1,$2,$3,$4);
+       if($repo =~ s/^:pserver(?:([^:]*)):(?:(.*?)(?::(.*?))?@)?([^:\/]*)(?::(\d*))?//) {
+               my($param,$user,$pass,$serv,$port) = ($1,$2,$3,$4,$5);
+
+               my($proxyhost,$proxyport);
+               if($param && ($param =~ m/proxy=([^;]+)/)) {
+                       $proxyhost = $1;
+                       # Default proxyport, if not specified, is 8080.
+                       $proxyport = 8080;
+                       if($ENV{"CVS_PROXY_PORT"}) {
+                               $proxyport = $ENV{"CVS_PROXY_PORT"};
+                       }
+                       if($param =~ m/proxyport=([^;]+)/){
+                               $proxyport = $1;
+                       }
+               }
+
                $user="anonymous" unless defined $user;
                my $rr2 = "-";
                unless($port) {
@@ -187,13 +201,43 @@ sub conn {
                }
                $pass="A" unless $pass;
 
-               my $s = IO::Socket::INET->new(PeerHost => $serv, PeerPort => $port);
-               die "Socket to $serv: $!\n" unless defined $s;
+               my ($s, $rep);
+               if($proxyhost) {
+
+                       # Use a HTTP Proxy. Only works for HTTP proxies that
+                       # don't require user authentication
+                       #
+                       # See: http://www.ietf.org/rfc/rfc2817.txt
+
+                       $s = IO::Socket::INET->new(PeerHost => $proxyhost, PeerPort => $proxyport);
+                       die "Socket to $proxyhost: $!\n" unless defined $s;
+                       $s->write("CONNECT $serv:$port HTTP/1.1\r\nHost: $serv:$port\r\n\r\n")
+                               or die "Write to $proxyhost: $!\n";
+                       $s->flush();
+
+                       $rep = <$s>;
+
+                       # The answer should look like 'HTTP/1.x 2yy ....'
+                       if(!($rep =~ m#^HTTP/1\.. 2[0-9][0-9]#)) {
+                               die "Proxy connect: $rep\n";
+                       }
+                       # Skip up to the empty line of the proxy server output
+                       # including the response headers.
+                       while ($rep = <$s>) {
+                               last if (!defined $rep ||
+                                        $rep eq "\n" ||
+                                        $rep eq "\r\n");
+                       }
+               } else {
+                       $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")
                        or die "Write to $serv: $!\n";
                $s->flush();
 
-               my $rep = <$s>;
+               $rep = <$s>;
 
                if($rep ne "I LOVE YOU\n") {
                        $rep="<unknown>" unless $rep;
@@ -465,10 +509,9 @@ ($$)
 $ENV{"GIT_DIR"} = $git_dir;
 my $orig_git_index;
 $orig_git_index = $ENV{GIT_INDEX_FILE} if exists $ENV{GIT_INDEX_FILE};
-my ($git_ih, $git_index) = tempfile('gitXXXXXX', SUFFIX => '.idx',
-                                   DIR => File::Spec->tmpdir());
-close ($git_ih);
-$ENV{GIT_INDEX_FILE} = $git_index;
+
+my %index; # holds filenames of one index per branch
+
 unless(-d $git_dir) {
        system("git-init-db");
        die "Cannot init the GIT db at $git_tree: $?\n" if $?;
@@ -495,27 +538,18 @@ ($$)
        $orig_branch = $last_branch;
        $tip_at_start = `git-rev-parse --verify HEAD`;
 
-       # 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");
-       while(defined(my $head = readdir(D))) {
-               next if $head =~ /^\./;
-               open(F,"$git_dir/refs/heads/$head")
-                       or die "Bad head branch: $head: $!\n";
-               chomp(my $ftag = <F>);
-               close(F);
-               open(F,"git-cat-file commit $ftag |");
-               while(<F>) {
-                       next unless /^author\s.*\s(\d+)\s[-+]\d{4}$/;
-                       $branch_date{$head} = $1;
-                       last;
-               }
-               close(F);
+       my $fmt = '($ref, $author) = (%(refname), %(author));';
+       open(H, "git-for-each-ref --perl --format='$fmt' refs/heads |") or
+               die "Cannot run git-for-each-ref: $!\n";
+       while(defined(my $entry = <H>)) {
+               my ($ref, $author);
+               eval($entry) || die "cannot eval refs list: $@";
+               my ($head) = ($ref =~ m|^refs/heads/(.*)|);
+               $author =~ /^.*\s(\d+)\s[-+]\d{4}$/;
+               $branch_date{$head} = $1;
        }
-       closedir(D);
+       close(H);
 }
 
 -d $git_dir
@@ -529,25 +563,39 @@ ($$)
        write_author_info("$git_dir/cvs-authors");
 }
 
-my $pid = open(CVS,"-|");
-die "Cannot fork: $!\n" unless defined $pid;
-unless($pid) {
-       my @opt;
-       @opt = split(/,/,$opt_p) if defined $opt_p;
-       unshift @opt, '-z', $opt_z if defined $opt_z;
-       unshift @opt, '-q'         unless defined $opt_v;
-       unless (defined($opt_p) && $opt_p =~ m/--no-cvs-direct/) {
-               push @opt, '--cvs-direct';
+
+#
+# run cvsps into a file unless we are getting
+# it passed as a file via $opt_P
+#
+unless ($opt_P) {
+       print "Running cvsps...\n" if $opt_v;
+       my $pid = open(CVSPS,"-|");
+       die "Cannot fork: $!\n" unless defined $pid;
+       unless($pid) {
+               my @opt;
+               @opt = split(/,/,$opt_p) if defined $opt_p;
+               unshift @opt, '-z', $opt_z if defined $opt_z;
+               unshift @opt, '-q'         unless defined $opt_v;
+               unless (defined($opt_p) && $opt_p =~ m/--no-cvs-direct/) {
+                       push @opt, '--cvs-direct';
+               }
+               exec("cvsps","--norc",@opt,"-u","-A",'--root',$opt_d,$cvs_tree);
+               die "Could not start cvsps: $!\n";
        }
-       if ($opt_P) {
-           exec("cat", $opt_P);
-       } else {
-           exec("cvsps","--norc",@opt,"-u","-A",'--root',$opt_d,$cvs_tree);
-           die "Could not start cvsps: $!\n";
+       my ($cvspsfh, $cvspsfile) = tempfile('gitXXXXXX', SUFFIX => '.cvsps',
+                                            DIR => File::Spec->tmpdir());
+       while (<CVSPS>) {
+           print $cvspsfh $_;
        }
+       close CVSPS;
+       close $cvspsfh;
+       $opt_P = $cvspsfile;
 }
 
 
+open(CVS, "<$opt_P") or die $!;
+
 ## cvsps output:
 #---------------------
 #PatchSet 314
@@ -595,8 +643,33 @@ ()
 }
 
 my($patchset,$date,$author_name,$author_email,$branch,$ancestor,$tag,$logmsg);
-my(@old,@new,@skipped);
+my(@old,@new,@skipped,%ignorebranch);
+
+# commits that cvsps cannot place anywhere...
+$ignorebranch{'#CVSPS_NO_BRANCH'} = 1;
+
 sub commit {
+       if ($branch eq $opt_o && !$index{branch} && !get_headref($branch, $git_dir)) {
+           # looks like an initial commit
+           # use the index primed by git-init-db
+           $ENV{GIT_INDEX_FILE} = '.git/index';
+           $index{$branch} = '.git/index';
+       } else {
+           # use an index per branch to speed up
+           # imports of projects with many branches
+           unless ($index{$branch}) {
+               $index{$branch} = tmpnam();
+               $ENV{GIT_INDEX_FILE} = $index{$branch};
+               if ($ancestor) {
+                   system("git-read-tree", $ancestor);
+               } else {
+                   system("git-read-tree", $branch);
+               }
+               die "read-tree failed: $?\n" if $?;
+           }
+       }
+        $ENV{GIT_INDEX_FILE} = $index{$branch};
+
        update_index(@old, @new);
        @old = @new = ();
        my $tree = write_tree();
@@ -618,14 +691,13 @@ sub commit {
        }
 
        my $commit_date = strftime("+0000 %Y-%m-%d %H:%M:%S",gmtime($date));
+       $ENV{GIT_AUTHOR_NAME} = $author_name;
+       $ENV{GIT_AUTHOR_EMAIL} = $author_email;
+       $ENV{GIT_AUTHOR_DATE} = $commit_date;
+       $ENV{GIT_COMMITTER_NAME} = $author_name;
+       $ENV{GIT_COMMITTER_EMAIL} = $author_email;
+       $ENV{GIT_COMMITTER_DATE} = $commit_date;
        my $pid = open2(my $commit_read, my $commit_write,
-               'env',
-               "GIT_AUTHOR_NAME=$author_name",
-               "GIT_AUTHOR_EMAIL=$author_email",
-               "GIT_AUTHOR_DATE=$commit_date",
-               "GIT_COMMITTER_NAME=$author_name",
-               "GIT_COMMITTER_EMAIL=$author_email",
-               "GIT_COMMITTER_DATE=$commit_date",
                'git-commit-tree', $tree, @commit_args);
 
        # compatibility with git2cvs
@@ -752,7 +824,16 @@ sub commit {
                        $state = 11;
                        next;
                }
+               if (exists $ignorebranch{$branch}) {
+                       print STDERR "Skipping $branch\n";
+                       $state = 11;
+                       next;
+               }
                if($ancestor) {
+                       if($ancestor eq $branch) {
+                               print STDERR "Branch $branch erroneously stems from itself -- changed ancestor to $opt_o\n";
+                               $ancestor = $opt_o;
+                       }
                        if(-f "$git_dir/refs/heads/$branch") {
                                print STDERR "Branch $branch already exists!\n";
                                $state=11;
@@ -760,6 +841,7 @@ sub commit {
                        }
                        unless(open(H,"$git_dir/refs/heads/$ancestor")) {
                                print STDERR "Branch $ancestor does not exist!\n";
+                               $ignorebranch{$branch} = 1;
                                $state=11;
                                next;
                        }
@@ -767,6 +849,7 @@ sub commit {
                        close(H);
                        unless(open(H,"> $git_dir/refs/heads/$branch")) {
                                print STDERR "Could not create branch $branch: $!\n";
+                               $ignorebranch{$branch} = 1;
                                $state=11;
                                next;
                        }
@@ -775,11 +858,6 @@ sub commit {
                        close(H)
                                or die "Could not write branch $branch: $!";
                }
-               if(($ancestor || $branch) ne $last_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) {
@@ -842,7 +920,21 @@ sub commit {
 }
 commit() if $branch and $state != 11;
 
-unlink($git_index);
+# The heuristic of repacking every 1024 commits can leave a
+# lot of unpacked data.  If there is more than 1MB worth of
+# not-packed objects, repack once more.
+my $line = `git-count-objects`;
+if ($line =~ /^(\d+) objects, (\d+) kilobytes$/) {
+  my ($n_objects, $kb) = ($1, $2);
+  1024 < $kb
+    and system("git repack -a -d");
+}
+
+foreach my $git_index (values %index) {
+    if ($git_index ne '.git/index') {
+       unlink($git_index);
+    }
+}
 
 if (defined $orig_git_index) {
        $ENV{GIT_INDEX_FILE} = $orig_git_index;