1 # Copyright (C) 2021 all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
4 # "lei add-external --mirror" support (also "public-inbox-clone");
5 package PublicInbox::LeiMirror;
8 use parent qw(PublicInbox::IPC);
9 use IO::Uncompress::Gunzip qw(gunzip $GunzipError);
10 use PublicInbox::Spawn qw(popen_rd spawn run_die);
12 use Fcntl qw(SEEK_SET O_CREAT O_EXCL O_WRONLY);
14 sub _wq_done_wait { # dwaitpid callback (via wq_eof)
16 my ($mrr, $lei) = @$arg;
17 my $f = "$mrr->{dst}/mirror.done";
19 $lei->child_error($?);
20 } elsif (!unlink($f)) {
21 $lei->err("unlink($f): $!") unless $!{ENOENT};
23 if ($lei->{cmd} ne 'public-inbox-clone') {
24 $lei->add_external_finish($mrr->{dst});
26 $lei->qerr("# mirrored $mrr->{src} => $mrr->{dst}");
31 # for old installations without manifest.js.gz
34 my $uri = URI->new($self->{src});
35 my $lei = $self->{lei};
36 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
37 my $cmd = $curl->for_uri($lei, $uri, '--compressed');
38 my $opt = { 0 => $lei->{0}, 2 => $lei->{2} };
39 my $fh = popen_rd($cmd, undef, $opt);
40 my $html = do { local $/; <$fh> } // die "read(curl $uri): $!";
41 close($fh) or return $lei->child_error($?, "@$cmd failed");
43 # we grep with URL below, we don't want Subject/From headers
44 # making us clone random URLs
45 my @urls = ($html =~ m!\bgit clone --mirror ([a-z\+]+://\S+)!g);
46 my $url = $uri->as_string;
47 chop($url) eq '/' or die "BUG: $uri not canonicalized";
49 # since this is for old instances w/o manifest.js.gz, try v1 first
50 return clone_v1($self) if grep(m!\A\Q$url\E/*\z!, @urls);
51 if (my @v2_urls = grep(m!\A\Q$url\E/[0-9]+\z!, @urls)) {
53 my ($n) = (m!/([0-9]+)\z!);
56 return clone_v2($self, \%v2_epochs);
59 # filter out common URLs served by WWW (e.g /$MSGID/T/)
60 if (@urls && $url =~ s!/+[^/]+\@[^/]+/.*\z!! &&
61 grep(m!\A\Q$url\E/*\z!, @urls)) {
63 E: confused by scraping <$uri>, did you mean <$url>?
67 E: confused by scraping <$uri>, got ambiguous results:
70 die "E: scraping <$uri> revealed nothing\n";
76 $opt->{$_} = $lei->{$_} for (0..2);
77 # we support "-c $key=$val" for arbitrary git config options
78 # e.g.: git -c http.proxy=socks5h://127.0.0.1:9050
79 push(@cmd, '-c', $_) for @{$lei->{opt}->{c} // []};
80 push @cmd, qw(clone --mirror);
81 push @cmd, '-q' if $lei->{opt}->{quiet};
82 push @cmd, '-v' if $lei->{opt}->{verbose};
83 # XXX any other options to support?
84 # --reference is tricky with multiple epochs...
88 sub _get_txt { # non-fatal
89 my ($self, $endpoint, $file) = @_;
90 my $uri = URI->new($self->{src});
91 my $lei = $self->{lei};
92 my $path = $uri->path;
93 chop($path) eq '/' or die "BUG: $uri not canonicalized";
94 $uri->path("$path/$endpoint");
95 my $cmd = $self->{curl}->for_uri($lei, $uri, '--compressed');
96 my $ce = "$self->{dst}/$file";
97 my $ft = File::Temp->new(TEMPLATE => "$file-XXXX",
98 UNLINK => 1, DIR => $self->{dst});
99 my $opt = { 0 => $lei->{0}, 1 => $ft, 2 => $lei->{2} };
100 my $cerr = run_reap($lei, $cmd, $opt);
101 return "$uri missing" if ($cerr >> 8) == 22;
102 return "# @$cmd failed (non-fatal)" if $cerr;
103 my $f = $ft->filename;
104 rename($f, $ce) or return "rename($f, $ce): $! (non-fatal)";
105 $ft->unlink_on_destroy(0);
109 # tries the relatively new /$INBOX/_/text/config/raw endpoint
112 my $dst = $self->{dst};
113 if (!-d $dst || !mkdir($dst)) {
115 File::Path::mkpath($dst);
116 -d $dst or die "mkpath($dst): $!\n";
118 my $err = _get_txt($self, qw(_/text/config/raw inbox.config.example));
119 return $self->{lei}->err($err) if $err;
120 my $f = "$self->{dst}/inbox.config.example";
121 my $cfg = PublicInbox::Config->git_config_dump($f, $self->{lei}->{2});
122 my $ibx = $self->{ibx} = {};
123 for my $sec (grep(/\Apublicinbox\./, @{$cfg->{-section_order}})) {
124 for (qw(address newsgroup nntpmirror)) {
125 $ibx->{$_} = $cfg->{"$sec.$_"};
130 sub set_description ($) {
132 my $f = "$self->{dst}/description";
133 open my $fh, '+>>', $f or die "open($f): $!";
134 seek($fh, 0, SEEK_SET) or die "seek($f): $!";
135 chomp(my $d = do { local $/; <$fh> } // die "read($f): $!");
136 if ($d eq '($INBOX_DIR/description missing)' ||
137 $d =~ /^Unnamed repository/ || $d !~ /\S/) {
138 seek($fh, 0, SEEK_SET) or die "seek($f): $!";
139 truncate($fh, 0) or die "truncate($f): $!";
140 print $fh "mirror of $self->{src}\n" or die "print($f): $!";
141 close $fh or die "close($f): $!";
145 sub index_cloned_inbox {
146 my ($self, $iv) = @_;
147 my $lei = $self->{lei};
148 my $err = _get_txt($self, qw(description description));
149 $lei->err($err) if $err; # non fatal
150 eval { set_description($self) };
153 # n.b. public-inbox-clone works w/o (SQLite || Xapian)
154 # lei is useless without Xapian + SQLite
155 if ($lei->{cmd} ne 'public-inbox-clone') {
156 my $ibx = delete($self->{ibx}) // {
157 address => [ 'lei@example.com' ],
160 $ibx->{inboxdir} = $self->{dst};
161 PublicInbox::Inbox->new($ibx);
162 PublicInbox::InboxWritable->new($ibx);
164 for my $sw ($lei->index_opt) {
165 my ($k) = ($sw =~ /\A([\w-]+)/);
166 $opt->{$k} = $lei->{opt}->{$k};
168 # force synchronous dwaitpid for v2:
169 local $PublicInbox::DS::in_loop = 0;
170 my $cfg = PublicInbox::Config->new(undef, $lei->{2});
171 my $env = PublicInbox::Admin::index_prepare($opt, $cfg);
172 local %ENV = (%ENV, %$env) if $env;
173 PublicInbox::Admin::progress_prepare($opt, $lei->{2});
174 PublicInbox::Admin::index_inbox($ibx, undef, $opt);
176 open my $x, '>', "$self->{dst}/mirror.done"; # for _wq_done_wait
180 my ($lei, $cmd, $opt) = @_;
181 $lei->qerr("# @$cmd");
182 $opt->{pgid} = 0 if $lei->{sock};
183 my $pid = spawn($cmd, undef, $opt);
184 my $reap = PublicInbox::OnDestroy->new($lei->can('sigint_reap'), $pid);
185 waitpid($pid, 0) == $pid or die "waitpid @$cmd: $!";
186 @$reap = (); # cancel reap
192 my $lei = $self->{lei};
193 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
194 my $uri = URI->new($self->{src});
195 defined($lei->{opt}->{epoch}) and
196 die "$uri is a v1 inbox, --epoch is not supported\n";
197 my $pfx = $curl->torsocks($lei, $uri) or return;
198 my $cmd = [ @$pfx, clone_cmd($lei, my $opt = {}),
199 $uri->as_string, $self->{dst} ];
200 my $cerr = run_reap($lei, $cmd, $opt);
201 return $lei->child_error($cerr, "@$cmd failed") if $cerr;
203 write_makefile($self->{dst}, 1);
204 index_cloned_inbox($self, 1);
207 sub parse_epochs ($$) {
208 my ($opt_epochs, $v2_epochs) = @_; # $epcohs "LOW..HIGH"
209 $opt_epochs // return; # undef => all epochs
210 my ($lo, $dotdot, $hi, @extra) = split(/(\.\.)/, $opt_epochs);
211 undef($lo) if ($lo // '') eq '';
212 my $re = qr/\A~?[0-9]+\z/;
213 if (@extra || (($lo // '0') !~ $re) ||
214 (($hi // '0') !~ $re) ||
215 !(grep(defined, $lo, $hi))) {
217 --epoch=$opt_epochs not in the form of `LOW..HIGH', `LOW..', nor `..HIGH'
220 my @n = sort { $a <=> $b } keys %$v2_epochs;
221 for (grep(defined, $lo, $hi)) {
224 "`$_' exceeds maximum available epoch ($n[-1])\n";
226 "`$_' is lower than minimum available epoch ($n[0])\n";
227 } elsif (/\A~([0-9]+)/) {
229 $n[$off] // die "`$_' is out of range\n";
231 } else { die "`$_' not understood\n" }
233 defined($lo) && defined($hi) && $lo > $hi and die
234 "low value (`$lo') exceeds high (`$hi')\n";
235 $lo //= $n[0] if $dotdot;
236 $hi //= $n[-1] if $dotdot;
240 if (defined $v2_epochs->{$_}) {
244 "# epoch $_ is not available (non-fatal, $lo..$hi)\n";
250 sub init_placeholder ($$) {
251 my ($src, $edst) = @_;
252 PublicInbox::Import::init_bare($edst);
253 my $f = "$edst/config";
254 open my $fh, '>>', $f or die "open($f): $!";
255 print $fh <<EOM or die "print($f): $!";
258 fetch = +refs/*:refs/*
261 ; This git epoch was created read-only and "public-inbox-fetch"
262 ; will not fetch updates for it unless write permission is added.
264 close $fh or die "close:($f): $!";
268 my ($self, $v2_epochs) = @_;
269 my $lei = $self->{lei};
270 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
271 my $pfx = $curl->torsocks($lei, (values %$v2_epochs)[0]) or return;
272 my $dst = $self->{dst};
273 my $want = parse_epochs($lei->{opt}->{epoch}, $v2_epochs);
274 my (@src_edst, @read_only);
275 for my $nr (sort { $a <=> $b } keys %$v2_epochs) {
276 my $uri = $v2_epochs->{$nr};
277 my $src = $uri->as_string;
279 $src =~ m!/([0-9]+)(?:\.git)?\z! or die <<"";
280 failed to extract epoch number from $src
282 $1 + 0 == $nr or die "BUG: <$uri> miskeyed $1 != $nr";
283 $edst .= "/git/$nr.git";
284 if (!$want || $want->{$nr}) {
285 push @src_edst, $src, $edst;
286 } else { # create a placeholder so users only need to chmod +w
287 init_placeholder($src, $edst);
288 push @read_only, $edst;
291 my $lk = bless { lock_path => "$dst/inbox.lock" }, 'PublicInbox::Lock';
293 my $on_destroy = $lk->lock_for_scope($$);
294 my @cmd = clone_cmd($lei, my $opt = {});
295 while (my ($src, $edst) = splice(@src_edst, 0, 2)) {
296 my $cmd = [ @$pfx, @cmd, $src, $edst ];
297 my $cerr = run_reap($lei, $cmd, $opt);
298 return $lei->child_error($cerr, "@$cmd failed") if $cerr;
300 require PublicInbox::MultiGit;
301 my $mg = PublicInbox::MultiGit->new($dst, 'all.git', 'git');
302 $mg->fill_alternates;
303 for my $i ($mg->git_epochs) { $mg->epoch_cfg_set($i) }
304 for my $edst (@read_only) {
305 my @st = stat($edst) or die "stat($edst): $!";
306 chmod($st[2] & 0555, $edst) or die "chmod(a-w, $edst): $!";
308 write_makefile($self->{dst}, 2);
309 undef $on_destroy; # unlock
310 index_cloned_inbox($self, 2);
313 # PSGI mount prefixes and manifest.js.gz prefixes don't always align...
314 sub deduce_epochs ($$) {
316 my ($v1_ent, @v2_epochs);
320 $v1_ent = $m->{$path};
321 @v2_epochs = grep(m!\A\Q$path\E/git/[0-9]+\.git\z!, keys %$m);
322 } while (!defined($v1_ent) && !@v2_epochs &&
323 $path =~ s!\A(/[^/]+)/!/! and $path_pfx .= $1);
324 ($path_pfx, $v1_ent ? $path : undef, @v2_epochs);
327 sub decode_manifest ($$$) {
328 my ($fh, $fn, $uri) = @_;
330 my $gz = do { local $/; <$fh> } // die "slurp($fn): $!";
331 gunzip(\$gz => \$js, MultiStream => 1) or
332 die "gunzip($uri): $GunzipError\n";
333 my $m = eval { PublicInbox::Config->json->decode($js) };
334 die "$uri: error decoding `$js': $@\n" if $@;
335 ref($m) eq 'HASH' or die "$uri unknown type: ".ref($m);
341 my $uri = URI->new($self->{src});
342 my $lei = $self->{lei};
343 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
344 my $path = $uri->path;
345 chop($path) eq '/' or die "BUG: $uri not canonicalized";
346 $uri->path($path . '/manifest.js.gz');
347 my $pdir = $lei->rel2abs($self->{dst});
348 $pdir =~ s!/[^/]+/?\z!!;
349 my $ft = File::Temp->new(TEMPLATE => 'm-XXXX',
350 UNLINK => 1, DIR => $pdir, SUFFIX => '.tmp');
351 my $fn = $ft->filename;
352 my ($bn) = ($fn =~ m!/([^/]+)\z!);
353 my $cmd = $curl->for_uri($lei, $uri, '-R', '-o', $bn);
354 my $opt = { -C => $pdir };
355 $opt->{$_} = $lei->{$_} for (0..2);
356 my $cerr = run_reap($lei, $cmd, $opt);
358 return try_scrape($self) if ($cerr >> 8) == 22; # 404 missing
359 return $lei->child_error($cerr, "@$cmd failed");
361 my $m = decode_manifest($ft, $fn, $uri);
362 my ($path_pfx, $v1_path, @v2_epochs) = deduce_epochs($m, $path);
364 # It may be possible to have v1 + v2 in parallel someday:
365 $lei->err(<<EOM) if defined $v1_path;
366 # `$v1_path' appears to be a v1 inbox while v2 epochs exist:
368 # ignoring $v1_path (use --inbox-version=1 to force v1 instead)
370 my %v2_epochs = map {
371 $uri->path($path_pfx.$_);
372 my ($n) = ("$uri" =~ m!/([0-9]+)\.git\z!);
375 clone_v2($self, \%v2_epochs);
376 } elsif (defined $v1_path) {
379 die "E: confused by <$uri>, possible matches:\n\t",
380 join(', ', sort keys %$m), "\n";
382 my $fin = "$self->{dst}/manifest.js.gz";
383 rename($fn, $fin) or die "E: rename($fn, $fin): $!";
384 $ft->unlink_on_destroy(0);
387 sub start_clone_url {
389 return try_manifest($self) if $self->{src} =~ m!\Ahttps?://!;
390 die "TODO: non-HTTP/HTTPS clone of $self->{src} not supported, yet";
393 sub do_mirror { # via wq_io_do
395 my $lei = $self->{lei};
397 my $iv = $lei->{opt}->{'inbox-version'};
399 return clone_v1($self) if $iv == 1;
400 return try_scrape($self) if $iv == 2;
401 die "bad --inbox-version=$iv\n";
403 return start_clone_url($self) if $self->{src} =~ m!://!;
404 die "TODO: cloning local directories not supported, yet";
406 $lei->fail($@) if $@;
410 my ($cls, $lei, $src, $dst) = @_;
411 my $self = bless { src => $src, dst => $dst }, $cls;
412 if ($src =~ m!https?://!) {
414 require PublicInbox::LeiCurl;
416 require PublicInbox::Lock;
417 require PublicInbox::Inbox;
418 require PublicInbox::Admin;
419 require PublicInbox::InboxWritable;
420 my ($op_c, $ops) = $lei->workers_start($self, 1);
422 $self->wq_io_do('do_mirror', []);
424 $lei->wait_wq_events($op_c, $ops);
427 sub ipc_atfork_child {
429 $self->{lei}->_lei_atfork_child;
430 $SIG{TERM} = sub { exit(128 + 15) }; # trigger OnDestroy $reap
431 $self->SUPER::ipc_atfork_child;
435 my ($dir, $ibx_ver) = @_;
436 my $f = "$dir/Makefile";
437 if (sysopen my $fh, $f, O_CREAT|O_EXCL|O_WRONLY) {
438 print $fh <<EOM or die "print($f) $!";
439 # This is a v$ibx_ver public-inbox, see the public-inbox-v$ibx_ver-format(5)
440 # manpage for more information on the format. This Makefile is
441 # intended as a familiar wrapper for users unfamiliar with
442 # public-inbox-* commands.
444 # See the respective manpages for public-inbox-fetch(1),
445 # public-inbox-index(1), etc for more information on
446 # some of the commands used by this Makefile.
448 # This Makefile will not be modified nor read by public-inbox,
449 # so you may edit it freely with your own convenience targets
450 # and notes. public-inbox-fetch will recreate it if removed.
452 print $fh <<'EOM' or die "print($f): $!";
453 # the default target:
455 @echo Common targets:
456 @echo ' make fetch - fetch from remote git repostorie(s)'
457 @echo ' make update - fetch and update index '
459 @echo Rarely needed targets:
460 @echo ' make reindex - may be needed for new features/bugfixes'
461 @echo ' make compact - rewrite Xapian storage to save space'
466 @if ! public-inbox-fetch --exit-code; \
469 test $$c -eq 127 && exit 0; \
471 elif test -f msgmap.sqlite3 || test -f public-inbox/msgmap.sqlite3; \
473 public-inbox-index; \
475 echo 'public-inbox index not initialized'; \
476 echo 'see public-inbox-index(1) man page'; \
479 public-inbox-index --reindex
483 .PHONY : help fetch update reindex compact
485 close $fh or die "close($f): $!";
487 die "open($f): $!" unless $!{EEXIST};