Fixed a typo in Documentation/git-cvsimport-script.txt.
[git.git] / git-cvsimport-script
1 #!/usr/bin/perl -w
2
3 # This tool is copyright (c) 2005, Matthias Urlichs.
4 # It is released under the Gnu Public License, version 2.
5 #
6 # The basic idea is to aggregate CVS check-ins into related changes.
7 # Fortunately, "cvsps" does that for us; all we have to do is to parse
8 # its output.
9 #
10 # Checking out the files is done by a single long-running CVS connection
11 # / server process.
12 #
13 # The head revision is on branch "origin" by default.
14 # You can change that with the '-o' option.
15
16 use strict;
17 use warnings;
18 use Getopt::Std;
19 use File::Path qw(mkpath);
20 use File::Basename qw(basename dirname);
21 use Time::Local;
22 use IO::Socket;
23 use IO::Pipe;
24 use POSIX qw(strftime dup2);
25
26 $SIG{'PIPE'}="IGNORE";
27 $ENV{'TZ'}="UTC";
28
29 our($opt_h,$opt_o,$opt_v,$opt_d,$opt_p);
30
31 sub usage() {
32         print STDERR <<END;
33 Usage: ${\basename $0}     # fetch/update GIT from CVS
34            [ -o branch-for-HEAD ] [ -h ] [ -v ] [ -d CVSROOT ]
35        [ -p opts-for-cvsps ]
36        CVS_module [ GIT_repository ]
37 END
38         exit(1);
39 }
40
41 getopts("hqvo:d:p:") or usage();
42 usage if $opt_h;
43
44 @ARGV == 1 or @ARGV == 2 or usage();
45
46 my($cvs_tree, $git_tree) = @ARGV;
47
48 if($opt_d) {
49         $ENV{"CVSROOT"} = $opt_d;
50 } elsif($ENV{"CVSROOT"}) {
51         $opt_d = $ENV{"CVSROOT"};
52 } else {
53         die "CVSROOT needs to be set";
54 }
55 $opt_o ||= "origin";
56 $git_tree ||= ".";
57
58 select(STDERR); $|=1; select(STDOUT);
59
60
61 package CVSconn;
62 # Basic CVS dialog.
63 # We're only interested in connecting and downloading, so ...
64
65 use POSIX qw(strftime dup2);
66
67 sub new {
68         my($what,$repo,$subdir) = @_;
69         $what=ref($what) if ref($what);
70
71         my $self = {};
72         $self->{'buffer'} = "";
73         bless($self,$what);
74
75         $repo =~ s#/+$##;
76         $self->{'fullrep'} = $repo;
77         $self->conn();
78
79         $self->{'subdir'} = $subdir;
80         $self->{'lines'} = undef;
81
82         return $self;
83 }
84
85 sub conn {
86         my $self = shift;
87         my $repo = $self->{'fullrep'};
88         if($repo =~ s/^:pserver:(?:(.*?)(?::(.*?))?@)?([^:\/]*)(?::(\d*))?//) {
89                 my($user,$pass,$serv,$port) = ($1,$2,$3,$4);
90                 $user="anonymous" unless defined $user;
91                 my $rr2 = "-";
92                 unless($port) {
93                         $rr2 = ":pserver:$user\@$serv:$repo";
94                         $port=2401;
95                 }
96                 my $rr = ":pserver:$user\@$serv:$port$repo";
97
98                 unless($pass) {
99                         open(H,$ENV{'HOME'}."/.cvspass") and do {
100                                 # :pserver:cvs@mea.tmt.tele.fi:/cvsroot/zmailer Ah<Z
101                                 while(<H>) {
102                                         chomp;
103                                         s/^\/\d+\s+//;
104                                         my ($w,$p) = split(/\s/,$_,2);
105                                         if($w eq $rr or $w eq $rr2) {
106                                                 $pass = $p;
107                                                 last;
108                                         }
109                                 }
110                         };
111                 }
112                 $pass="A" unless $pass;
113
114                 my $s = IO::Socket::INET->new(PeerHost => $serv, PeerPort => $port);
115                 die "Socket to $serv: $!\n" unless defined $s;
116                 $s->write("BEGIN AUTH REQUEST\n$repo\n$user\n$pass\nEND AUTH REQUEST\n")
117                         or die "Write to $serv: $!\n";
118                 $s->flush();
119
120                 my $rep = <$s>;
121
122                 if($rep ne "I LOVE YOU\n") {
123                         $rep="<unknown>" unless $rep;
124                         die "AuthReply: $rep\n";
125                 }
126                 $self->{'socketo'} = $s;
127                 $self->{'socketi'} = $s;
128         } else { # local: Fork off our own cvs server.
129                 my $pr = IO::Pipe->new();
130                 my $pw = IO::Pipe->new();
131                 my $pid = fork();
132                 die "Fork: $!\n" unless defined $pid;
133                 unless($pid) {
134                         $pr->writer();
135                         $pw->reader();
136                         dup2($pw->fileno(),0);
137                         dup2($pr->fileno(),1);
138                         $pr->close();
139                         $pw->close();
140                         exec("cvs","server");
141                 }
142                 $pw->writer();
143                 $pr->reader();
144                 $self->{'socketo'} = $pw;
145                 $self->{'socketi'} = $pr;
146         }
147         $self->{'socketo'}->write("Root $repo\n");
148
149         # Trial and error says that this probably is the minimum set
150         $self->{'socketo'}->write("Valid-responses ok error Valid-requests Mode M Mbinary E F Checked-in Created Updated Merged Removed\n");
151
152         $self->{'socketo'}->write("valid-requests\n");
153         $self->{'socketo'}->flush();
154
155         chomp(my $rep=$self->readline());
156         if($rep !~ s/^Valid-requests\s*//) {
157                 $rep="<unknown>" unless $rep;
158                 die "Expected Valid-requests from server, but got: $rep\n";
159         }
160         chomp(my $res=$self->readline());
161         die "validReply: $res\n" if $res ne "ok";
162
163         $self->{'socketo'}->write("UseUnchanged\n") if $rep =~ /\bUseUnchanged\b/;
164         $self->{'repo'} = $repo;
165 }
166
167 sub readline {
168         my($self) = @_;
169         return $self->{'socketi'}->getline();
170 }
171
172 sub _file {
173         # Request a file with a given revision.
174         # Trial and error says this is a good way to do it. :-/
175         my($self,$fn,$rev) = @_;
176         $self->{'socketo'}->write("Argument -N\n") or return undef;
177         $self->{'socketo'}->write("Argument -P\n") or return undef;
178         # $self->{'socketo'}->write("Argument -ko\n") or return undef;
179         # -ko: Linus' version doesn't use it
180         $self->{'socketo'}->write("Argument -r\n") or return undef;
181         $self->{'socketo'}->write("Argument $rev\n") or return undef;
182         $self->{'socketo'}->write("Argument --\n") or return undef;
183         $self->{'socketo'}->write("Argument $self->{'subdir'}/$fn\n") or return undef;
184         $self->{'socketo'}->write("Directory .\n") or return undef;
185         $self->{'socketo'}->write("$self->{'repo'}\n") or return undef;
186         # $self->{'socketo'}->write("Sticky T1.0\n") or return undef;
187         $self->{'socketo'}->write("co\n") or return undef;
188         $self->{'socketo'}->flush() or return undef;
189         $self->{'lines'} = 0;
190         return 1;
191 }
192 sub _line {
193         # Read a line from the server.
194         # ... except that 'line' may be an entire file. ;-)
195         my($self) = @_;
196         die "Not in lines" unless defined $self->{'lines'};
197
198         my $line;
199         my $res="";
200         while(defined($line = $self->readline())) {
201                 # M U gnupg-cvs-rep/AUTHORS
202                 # Updated gnupg-cvs-rep/
203                 # /daten/src/rsync/gnupg-cvs-rep/AUTHORS
204                 # /AUTHORS/1.1///T1.1
205                 # u=rw,g=rw,o=rw
206                 # 0
207                 # ok
208
209                 if($line =~ s/^(?:Created|Updated) //) {
210                         $line = $self->readline(); # path
211                         $line = $self->readline(); # Entries line
212                         my $mode = $self->readline(); chomp $mode;
213                         $self->{'mode'} = $mode;
214                         defined (my $cnt = $self->readline())
215                                 or die "EOF from server after 'Changed'\n";
216                         chomp $cnt;
217                         die "Duh: Filesize $cnt" if $cnt !~ /^\d+$/;
218                         $line="";
219                         $res="";
220                         while($cnt) {
221                                 my $buf;
222                                 my $num = $self->{'socketi'}->read($buf,$cnt);
223                                 die "Server: Filesize $cnt: $num: $!\n" if not defined $num or $num<=0;
224                                 $res .= $buf;
225                                 $cnt -= $num;
226                         }
227                 } elsif($line =~ s/^ //) {
228                         $res .= $line;
229                 } elsif($line =~ /^M\b/) {
230                         # output, do nothing
231                 } elsif($line =~ /^Mbinary\b/) {
232                         my $cnt;
233                         die "EOF from server after 'Mbinary'" unless defined ($cnt = $self->readline());
234                         chomp $cnt;
235                         die "Duh: Mbinary $cnt" if $cnt !~ /^\d+$/ or $cnt<1;
236                         $line="";
237                         while($cnt) {
238                                 my $buf;
239                                 my $num = $self->{'socketi'}->read($buf,$cnt);
240                                 die "S: Mbinary $cnt: $num: $!\n" if not defined $num or $num<=0;
241                                 $res .= $buf;
242                                 $cnt -= $num;
243                         }
244                 } else {
245                         chomp $line;
246                         if($line eq "ok") {
247                                 # print STDERR "S: ok (".length($res).")\n";
248                                 return $res;
249                         } elsif($line =~ s/^E //) {
250                                 # print STDERR "S: $line\n";
251                         } else {
252                                 die "Unknown: $line\n";
253                         }
254                 }
255         }
256 }
257 sub file {
258         my($self,$fn,$rev) = @_;
259         my $res;
260
261         if ($self->_file($fn,$rev)) {
262                 $res = $self->_line();
263                 return $res if defined $res;
264         }
265
266         # retry
267         $self->conn();
268         $self->_file($fn,$rev)
269                 or die "No file command send\n";
270         $res = $self->_line();
271         die "No input: $fn $rev\n" unless defined $res;
272         return $res;
273 }
274
275
276 package main;
277
278 my $cvs = CVSconn->new($opt_d, $cvs_tree);
279
280
281 sub pdate($) {
282         my($d) = @_;
283         m#(\d{2,4})/(\d\d)/(\d\d)\s(\d\d):(\d\d)(?::(\d\d))?#
284                 or die "Unparseable date: $d\n";
285         my $y=$1; $y-=1900 if $y>1900;
286         return timegm($6||0,$5,$4,$3,$2-1,$y);
287 }
288
289 sub pmode($) {
290         my($mode) = @_;
291         my $m = 0;
292         my $mm = 0;
293         my $um = 0;
294         for my $x(split(//,$mode)) {
295                 if($x eq ",") {
296                         $m |= $mm&$um;
297                         $mm = 0;
298                         $um = 0;
299                 } elsif($x eq "u") { $um |= 0700;
300                 } elsif($x eq "g") { $um |= 0070;
301                 } elsif($x eq "o") { $um |= 0007;
302                 } elsif($x eq "r") { $mm |= 0444;
303                 } elsif($x eq "w") { $mm |= 0222;
304                 } elsif($x eq "x") { $mm |= 0111;
305                 } elsif($x eq "=") { # do nothing
306                 } else { die "Unknown mode: $mode\n";
307                 }
308         }
309         $m |= $mm&$um;
310         return $m;
311 }
312
313 my $tmpcv = "/var/cache/cvs";
314
315 sub getwd() {
316         my $pwd = `pwd`;
317         chomp $pwd;
318         return $pwd;
319 }
320
321 -d $git_tree
322         or mkdir($git_tree,0777)
323         or die "Could not create $git_tree: $!";
324 chdir($git_tree);
325
326 my $last_branch = "";
327 my $orig_branch = "";
328 my %branch_date;
329
330 my $git_dir = $ENV{"GIT_DIR"} || ".git";
331 $git_dir = getwd()."/".$git_dir unless $git_dir =~ m#^/#;
332 $ENV{"GIT_DIR"} = $git_dir;
333 unless(-d $git_dir) {
334         system("git-init-db");
335         die "Cannot init the GIT db at $git_tree: $?\n" if $?;
336         system("git-read-tree");
337         die "Cannot init an empty tree: $?\n" if $?;
338
339         $last_branch = $opt_o;
340         $orig_branch = "";
341 } else {
342         -f "$git_dir/refs/head/$opt_o"
343                 or die "Branch '$opt_o' does not exist.\n".
344                        "Either use the correct '-o branch' option,\n".
345                        "or import to a new repository.\n";
346
347         $last_branch = basename(readlink("$git_dir/HEAD"));
348         unless($last_branch) {
349                 warn "Cannot read the last branch name: $! -- assuming 'master'\n";
350                 $last_branch = "master";
351         }
352         $orig_branch = $last_branch;
353
354         # Get the last import timestamps
355         opendir(D,"$git_dir/refs/heads");
356         while(defined(my $head = readdir(D))) {
357                 next if $head =~ /^\./;
358                 open(F,"$git_dir/refs/heads/$head")
359                         or die "Bad head branch: $head: $!\n";
360                 chomp(my $ftag = <F>);
361                 close(F);
362                 open(F,"git-cat-file commit $ftag |");
363                 while(<F>) {
364                         next unless /^author\s.*\s(\d+)\s[-+]\d{4}$/;
365                         $branch_date{$head} = $1;
366                         last;
367                 }
368                 close(F);
369         }
370         closedir(D);
371 }
372
373 -d $git_dir
374         or die "Could not create git subdir ($git_dir).\n";
375
376 my $pid = open(CVS,"-|");
377 die "Cannot fork: $!\n" unless defined $pid;
378 unless($pid) {
379         my @opt;
380         @opt = split(/,/,$opt_p) if defined $opt_p;
381         exec("cvsps",@opt,"-x","-A","--cvs-direct",$cvs_tree);
382         die "Could not start cvsps: $!\n";
383 }
384
385
386 ## cvsps output:
387 #---------------------
388 #PatchSet 314
389 #Date: 1999/09/18 13:03:59
390 #Author: wkoch
391 #Branch: STABLE-BRANCH-1-0
392 #Ancestor branch: HEAD
393 #Tag: (none)
394 #Log:
395 #    See ChangeLog: Sat Sep 18 13:03:28 CEST 1999  Werner Koch
396 #Members:
397 #       README:1.57->1.57.2.1
398 #       VERSION:1.96->1.96.2.1
399 #
400 #---------------------
401
402 my $state = 0;
403
404 my($patchset,$date,$author,$branch,$ancestor,$tag,$logmsg);
405 my(@old,@new);
406 my $commit = sub {
407         my $pid;
408         while(@old) {
409                 my @o2;
410                 if(@old > 55) {
411                         @o2 = splice(@old,0,50);
412                 } else {
413                         @o2 = @old;
414                         @old = ();
415                 }
416                 system("git-update-cache","--force-remove","--",@o2);
417                 die "Cannot remove files: $?\n" if $?;
418         }
419         while(@new) {
420                 my @n2;
421                 if(@new > 55) {
422                         @n2 = splice(@new,0,50);
423                 } else {
424                         @n2 = @new;
425                         @new = ();
426                 }
427                 system("git-update-cache","--add","--",@n2);
428                 die "Cannot add files: $?\n" if $?;
429         }
430
431         $pid = open(C,"-|");
432         die "Cannot fork: $!" unless defined $pid;
433         unless($pid) {
434                 exec("git-write-tree");
435                 die "Cannot exec git-write-tree: $!\n";
436         }
437         chomp(my $tree = <C>);
438         length($tree) == 40
439                 or die "Cannot get tree id ($tree): $!\n";
440         close(C)
441                 or die "Error running git-write-tree: $?\n";
442         print "Tree ID $tree\n" if $opt_v;
443
444         my $parent = "";
445         if(open(C,"$git_dir/refs/heads/$last_branch")) {
446                 chomp($parent = <C>);
447                 close(C);
448                 length($parent) == 40
449                         or die "Cannot get parent id ($parent): $!\n";
450                 print "Parent ID $parent\n" if $opt_v;
451         }
452
453         my $pr = IO::Pipe->new();
454         my $pw = IO::Pipe->new();
455         $pid = fork();
456         die "Fork: $!\n" unless defined $pid;
457         unless($pid) {
458                 $pr->writer();
459                 $pw->reader();
460                 dup2($pw->fileno(),0);
461                 dup2($pr->fileno(),1);
462                 $pr->close();
463                 $pw->close();
464
465                 my @par = ();
466                 @par = ("-p",$parent) if $parent;
467                 exec("env",
468                         "GIT_AUTHOR_NAME=$author",
469                         "GIT_AUTHOR_EMAIL=$author",
470                         "GIT_AUTHOR_DATE=".strftime("+0000 %Y-%m-%d %H:%M:%S",gmtime($date)),
471                         "GIT_COMMITTER_NAME=$author",
472                         "GIT_COMMITTER_EMAIL=$author",
473                         "GIT_COMMITTER_DATE=".strftime("+0000 %Y-%m-%d %H:%M:%S",gmtime($date)),
474                         "git-commit-tree", $tree,@par);
475                 die "Cannot exec git-commit-tree: $!\n";
476         }
477         $pw->writer();
478         $pr->reader();
479
480         # compatibility with git2cvs
481         substr($logmsg,32767) = "" if length($logmsg) > 32767;
482         $logmsg =~ s/[\s\n]+\z//;
483
484         print $pw "$logmsg\n"
485                 or die "Error writing to git-commit-tree: $!\n";
486         $pw->close();
487
488         print "Committed patch $patchset ($branch)\n" if $opt_v;
489         chomp(my $cid = <$pr>);
490         length($cid) == 40
491                 or die "Cannot get commit id ($cid): $!\n";
492         print "Commit ID $cid\n" if $opt_v;
493         $pr->close();
494
495         waitpid($pid,0);
496         die "Error running git-commit-tree: $?\n" if $?;
497
498         open(C,">$git_dir/refs/heads/$branch")
499                 or die "Cannot open branch $branch for update: $!\n";
500         print C "$cid\n"
501                 or die "Cannot write branch $branch for update: $!\n";
502         close(C)
503                 or die "Cannot write branch $branch for update: $!\n";
504
505         if($tag) {
506                 open(C,">$git_dir/refs/tags/$tag")
507                         or die "Cannot create tag $tag: $!\n";
508                 print C "$cid\n"
509                         or die "Cannot write tag $branch: $!\n";
510                 close(C)
511                         or die "Cannot write tag $branch: $!\n";
512                 print "Created tag '$tag' on '$branch'\n" if $opt_v;
513         }
514 };
515
516 while(<CVS>) {
517         chomp;
518         if($state == 0 and /^-+$/) {
519                 $state = 1;
520         } elsif($state == 0) {
521                 $state = 1;
522                 redo;
523         } elsif(($state==0 or $state==1) and s/^PatchSet\s+//) {
524                 $patchset = 0+$_;
525                 $state=2;
526         } elsif($state == 2 and s/^Date:\s+//) {
527                 $date = pdate($_);
528                 unless($date) {
529                         print STDERR "Could not parse date: $_\n";
530                         $state=0;
531                         next;
532                 }
533                 $state=3;
534         } elsif($state == 3 and s/^Author:\s+//) {
535                 s/\s+$//;
536                 $author = $_;
537                 $state = 4;
538         } elsif($state == 4 and s/^Branch:\s+//) {
539                 s/\s+$//;
540                 $branch = $_;
541                 $state = 5;
542         } elsif($state == 5 and s/^Ancestor branch:\s+//) {
543                 s/\s+$//;
544                 $ancestor = $_;
545                 $ancestor = $opt_o if $ancestor eq "HEAD";
546                 $state = 6;
547         } elsif($state == 5) {
548                 $ancestor = undef;
549                 $state = 6;
550                 redo;
551         } elsif($state == 6 and s/^Tag:\s+//) {
552                 s/\s+$//;
553                 if($_ eq "(none)") {
554                         $tag = undef;
555                 } else {
556                         $tag = $_;
557                 }
558                 $state = 7;
559         } elsif($state == 7 and /^Log:/) {
560                 $logmsg = "";
561                 $state = 8;
562         } elsif($state == 8 and /^Members:/) {
563                 $branch = $opt_o if $branch eq "HEAD";
564                 if(defined $branch_date{$branch} and $branch_date{$branch} >= $date) {
565                         # skip
566                         print "skip patchset $patchset: $date before $branch_date{$branch}\n";
567                         $state = 11;
568                         next;
569                 }
570                 if($ancestor) {
571                         if(-f "$git_dir/refs/heads/$branch") {
572                                 print STDERR "Branch $branch already exists!\n";
573                                 $state=11;
574                                 next;
575                         }
576                         unless(open(H,"$git_dir/refs/heads/$ancestor")) {
577                                 print STDERR "Branch $ancestor does not exist!\n";
578                                 $state=11;
579                                 next;
580                         }
581                         chomp(my $id = <H>);
582                         close(H);
583                         unless(open(H,"> $git_dir/refs/heads/$branch")) {
584                                 print STDERR "Could not create branch $branch: $!\n";
585                                 $state=11;
586                                 next;
587                         }
588                         print H "$id\n"
589                                 or die "Could not write branch $branch: $!";
590                         close(H)
591                                 or die "Could not write branch $branch: $!";
592                 }
593                 if(($ancestor || $branch) ne $last_branch) {
594                         print "Switching from $last_branch to $branch\n" if $opt_v;
595                         system("git-read-tree","-m","-u","$last_branch","$branch");
596                         die "read-tree failed: $?\n" if $?;
597                 }
598                 if($branch ne $last_branch) {
599                         unlink("$git_dir/HEAD");
600                         symlink("refs/heads/$branch","$git_dir/HEAD");
601                         $last_branch = $branch;
602                 }
603                 $state = 9;
604         } elsif($state == 8) {
605                 $logmsg .= "$_\n";
606         } elsif($state == 9 and /^\s+(\S+):(INITIAL|\d(?:\.\d+)+)->(\d(?:\.\d+)+)\s*$/) {
607 #       VERSION:1.96->1.96.2.1
608                 my $init = ($2 eq "INITIAL");
609                 my $fn = $1;
610                 my $rev = $3;
611                 $fn =~ s#^/+##;
612                 my $data = $cvs->file($fn,$rev);
613                 print "".($init ? "New" : "Update")." $fn: ".length($data)." bytes.\n";
614                 mkpath(dirname($fn),$opt_v);
615                 open(F,"> ./$fn")
616                         or die "Cannot create '$fn': $!\n";
617                 print F $data
618                         or die "Cannot write to '$fn': $!\n";
619                 close(F)
620                         or die "Cannot write to '$fn': $!\n";
621                 chmod(pmode($cvs->{'mode'}), $fn);
622                 push(@new,$fn); # may be resurrected!
623         } elsif($state == 9 and /^\s+(\S+):\d(?:\.\d+)+->(\d(?:\.\d+)+)\(DEAD\)\s*$/) {
624                 my $fn = $1;
625                 $fn =~ s#^/+##;
626                 push(@old,$fn);
627         } elsif($state == 9 and /^\s*$/) {
628                 $state = 10;
629         } elsif(($state == 9 or $state == 10) and /^-+$/) {
630                 &$commit();
631                 $state = 1;
632         } elsif($state == 11 and /^-+$/) {
633                 $state = 1;
634         } elsif(/^-+$/) { # end of unknown-line processing
635                 $state = 1;
636         } elsif($state != 11) { # ignore stuff when skipping
637                 print "* UNKNOWN LINE * $_\n";
638         }
639 }
640 &$commit() if $branch and $state != 11;
641
642 # Now switch back to the branch we were in before all of this happened
643 if($orig_branch) {
644         print "DONE; switching back to $orig_branch\n" if $opt_v;
645 } else {
646         $orig_branch = "master";
647         print "DONE; creating $orig_branch branch\n" if $opt_v;
648         system("cp","$git_dir/refs/heads/$opt_o","$git_dir/refs/heads/master")
649                 unless -f "$git_dir/refs/heads/master";
650 }
651
652 system("git-read-tree","-m","-u","$last_branch","$orig_branch");
653 die "read-tree failed: $?\n" if $?;
654
655 unlink("$git_dir/HEAD");
656 symlink("refs/heads/$orig_branch","$git_dir/HEAD");
657