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 @html = split(/<hr>/, $html);
46 my @urls = ($html[-1] =~ m!\bgit clone --mirror ([a-z\+]+://\S+)!g);
47 my $url = $uri->as_string;
48 chop($url) eq '/' or die "BUG: $uri not canonicalized";
50 # since this is for old instances w/o manifest.js.gz, try v1 first
51 return clone_v1($self) if grep(m!\A\Q$url\E/*\z!, @urls);
52 if (my @v2_urls = grep(m!\A\Q$url\E/[0-9]+\z!, @urls)) {
54 my ($n) = (m!/([0-9]+)\z!);
57 return clone_v2($self, \%v2_epochs);
60 # filter out common URLs served by WWW (e.g /$MSGID/T/)
61 if (@urls && $url =~ s!/+[^/]+\@[^/]+/.*\z!! &&
62 grep(m!\A\Q$url\E/*\z!, @urls)) {
64 E: confused by scraping <$uri>, did you mean <$url>?
68 E: confused by scraping <$uri>, got ambiguous results:
71 die "E: scraping <$uri> revealed nothing\n";
77 $opt->{$_} = $lei->{$_} for (0..2);
78 # we support "-c $key=$val" for arbitrary git config options
79 # e.g.: git -c http.proxy=socks5h://127.0.0.1:9050
80 push(@cmd, '-c', $_) for @{$lei->{opt}->{c} // []};
81 push @cmd, qw(clone --mirror);
82 push @cmd, '-q' if $lei->{opt}->{quiet};
83 push @cmd, '-v' if $lei->{opt}->{verbose};
84 # XXX any other options to support?
85 # --reference is tricky with multiple epochs...
89 sub _get_txt { # non-fatal
90 my ($self, $endpoint, $file) = @_;
91 my $uri = URI->new($self->{src});
92 my $lei = $self->{lei};
93 my $path = $uri->path;
94 chop($path) eq '/' or die "BUG: $uri not canonicalized";
95 $uri->path("$path/$endpoint");
96 my $cmd = $self->{curl}->for_uri($lei, $uri, '--compressed');
97 my $ce = "$self->{dst}/$file";
98 my $ft = File::Temp->new(TEMPLATE => "$file-XXXX",
99 UNLINK => 1, DIR => $self->{dst});
100 my $opt = { 0 => $lei->{0}, 1 => $ft, 2 => $lei->{2} };
101 my $cerr = run_reap($lei, $cmd, $opt);
102 return "$uri missing" if ($cerr >> 8) == 22;
103 return "# @$cmd failed (non-fatal)" if $cerr;
104 my $f = $ft->filename;
105 rename($f, $ce) or return "rename($f, $ce): $! (non-fatal)";
106 $ft->unlink_on_destroy(0);
110 # tries the relatively new /$INBOX/_/text/config/raw endpoint
113 my $dst = $self->{dst};
114 if (!-d $dst || !mkdir($dst)) {
116 File::Path::mkpath($dst);
117 -d $dst or die "mkpath($dst): $!\n";
119 my $err = _get_txt($self, qw(_/text/config/raw inbox.config.example));
120 return $self->{lei}->err($err) if $err;
121 my $f = "$self->{dst}/inbox.config.example";
122 my $cfg = PublicInbox::Config->git_config_dump($f, $self->{lei}->{2});
123 my $ibx = $self->{ibx} = {};
124 for my $sec (grep(/\Apublicinbox\./, @{$cfg->{-section_order}})) {
125 for (qw(address newsgroup nntpmirror)) {
126 $ibx->{$_} = $cfg->{"$sec.$_"};
131 sub set_description ($) {
133 my $f = "$self->{dst}/description";
134 open my $fh, '+>>', $f or die "open($f): $!";
135 seek($fh, 0, SEEK_SET) or die "seek($f): $!";
136 chomp(my $d = do { local $/; <$fh> } // die "read($f): $!");
137 if ($d eq '($INBOX_DIR/description missing)' ||
138 $d =~ /^Unnamed repository/ || $d !~ /\S/) {
139 seek($fh, 0, SEEK_SET) or die "seek($f): $!";
140 truncate($fh, 0) or die "truncate($f): $!";
141 print $fh "mirror of $self->{src}\n" or die "print($f): $!";
142 close $fh or die "close($f): $!";
146 sub index_cloned_inbox {
147 my ($self, $iv) = @_;
148 my $lei = $self->{lei};
149 my $err = _get_txt($self, qw(description description));
150 $lei->err($err) if $err; # non fatal
151 eval { set_description($self) };
154 # n.b. public-inbox-clone works w/o (SQLite || Xapian)
155 # lei is useless without Xapian + SQLite
156 if ($lei->{cmd} ne 'public-inbox-clone') {
157 my $ibx = delete($self->{ibx}) // {
158 address => [ 'lei@example.com' ],
161 $ibx->{inboxdir} = $self->{dst};
162 PublicInbox::Inbox->new($ibx);
163 PublicInbox::InboxWritable->new($ibx);
165 for my $sw ($lei->index_opt) {
166 my ($k) = ($sw =~ /\A([\w-]+)/);
167 $opt->{$k} = $lei->{opt}->{$k};
169 # force synchronous dwaitpid for v2:
170 local $PublicInbox::DS::in_loop = 0;
171 my $cfg = PublicInbox::Config->new(undef, $lei->{2});
172 my $env = PublicInbox::Admin::index_prepare($opt, $cfg);
173 local %ENV = (%ENV, %$env) if $env;
174 PublicInbox::Admin::progress_prepare($opt, $lei->{2});
175 PublicInbox::Admin::index_inbox($ibx, undef, $opt);
177 open my $x, '>', "$self->{dst}/mirror.done"; # for _wq_done_wait
181 my ($lei, $cmd, $opt) = @_;
182 $lei->qerr("# @$cmd");
183 $opt->{pgid} = 0 if $lei->{sock};
184 my $pid = spawn($cmd, undef, $opt);
185 my $reap = PublicInbox::OnDestroy->new($lei->can('sigint_reap'), $pid);
186 waitpid($pid, 0) == $pid or die "waitpid @$cmd: $!";
187 @$reap = (); # cancel reap
189 $? = 0; # don't let it influence normal exit
195 my $lei = $self->{lei};
196 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
197 my $uri = URI->new($self->{src});
198 defined($lei->{opt}->{epoch}) and
199 die "$uri is a v1 inbox, --epoch is not supported\n";
200 my $pfx = $curl->torsocks($lei, $uri) or return;
201 my $cmd = [ @$pfx, clone_cmd($lei, my $opt = {}),
202 $uri->as_string, $self->{dst} ];
203 my $cerr = run_reap($lei, $cmd, $opt);
204 return $lei->child_error($cerr, "@$cmd failed") if $cerr;
206 write_makefile($self->{dst}, 1);
207 index_cloned_inbox($self, 1);
210 sub parse_epochs ($$) {
211 my ($opt_epochs, $v2_epochs) = @_; # $epcohs "LOW..HIGH"
212 $opt_epochs // return; # undef => all epochs
213 my ($lo, $dotdot, $hi, @extra) = split(/(\.\.)/, $opt_epochs);
214 undef($lo) if ($lo // '') eq '';
215 my $re = qr/\A~?[0-9]+\z/;
216 if (@extra || (($lo // '0') !~ $re) ||
217 (($hi // '0') !~ $re) ||
218 !(grep(defined, $lo, $hi))) {
220 --epoch=$opt_epochs not in the form of `LOW..HIGH', `LOW..', nor `..HIGH'
223 my @n = sort { $a <=> $b } keys %$v2_epochs;
224 for (grep(defined, $lo, $hi)) {
227 "`$_' exceeds maximum available epoch ($n[-1])\n";
229 "`$_' is lower than minimum available epoch ($n[0])\n";
230 } elsif (/\A~([0-9]+)/) {
232 $n[$off] // die "`$_' is out of range\n";
234 } else { die "`$_' not understood\n" }
236 defined($lo) && defined($hi) && $lo > $hi and die
237 "low value (`$lo') exceeds high (`$hi')\n";
238 $lo //= $n[0] if $dotdot;
239 $hi //= $n[-1] if $dotdot;
243 if (defined $v2_epochs->{$_}) {
247 "# epoch $_ is not available (non-fatal, $lo..$hi)\n";
253 sub init_placeholder ($$) {
254 my ($src, $edst) = @_;
255 PublicInbox::Import::init_bare($edst);
256 my $f = "$edst/config";
257 open my $fh, '>>', $f or die "open($f): $!";
258 print $fh <<EOM or die "print($f): $!";
261 fetch = +refs/*:refs/*
264 ; This git epoch was created read-only and "public-inbox-fetch"
265 ; will not fetch updates for it unless write permission is added.
267 close $fh or die "close:($f): $!";
271 my ($self, $v2_epochs) = @_;
272 my $lei = $self->{lei};
273 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
274 my $pfx = $curl->torsocks($lei, (values %$v2_epochs)[0]) or return;
275 my $dst = $self->{dst};
276 my $want = parse_epochs($lei->{opt}->{epoch}, $v2_epochs);
277 my (@src_edst, @read_only);
278 for my $nr (sort { $a <=> $b } keys %$v2_epochs) {
279 my $uri = $v2_epochs->{$nr};
280 my $src = $uri->as_string;
282 $src =~ m!/([0-9]+)(?:\.git)?\z! or die <<"";
283 failed to extract epoch number from $src
285 $1 + 0 == $nr or die "BUG: <$uri> miskeyed $1 != $nr";
286 $edst .= "/git/$nr.git";
287 if (!$want || $want->{$nr}) {
288 push @src_edst, $src, $edst;
289 } else { # create a placeholder so users only need to chmod +w
290 init_placeholder($src, $edst);
291 push @read_only, $edst;
294 my $lk = bless { lock_path => "$dst/inbox.lock" }, 'PublicInbox::Lock';
296 my $on_destroy = $lk->lock_for_scope($$);
297 my @cmd = clone_cmd($lei, my $opt = {});
298 while (my ($src, $edst) = splice(@src_edst, 0, 2)) {
299 my $cmd = [ @$pfx, @cmd, $src, $edst ];
300 my $cerr = run_reap($lei, $cmd, $opt);
301 return $lei->child_error($cerr, "@$cmd failed") if $cerr;
303 require PublicInbox::MultiGit;
304 my $mg = PublicInbox::MultiGit->new($dst, 'all.git', 'git');
305 $mg->fill_alternates;
306 for my $i ($mg->git_epochs) { $mg->epoch_cfg_set($i) }
307 for my $edst (@read_only) {
308 my @st = stat($edst) or die "stat($edst): $!";
309 chmod($st[2] & 0555, $edst) or die "chmod(a-w, $edst): $!";
311 write_makefile($self->{dst}, 2);
312 undef $on_destroy; # unlock
313 index_cloned_inbox($self, 2);
316 # PSGI mount prefixes and manifest.js.gz prefixes don't always align...
317 sub deduce_epochs ($$) {
319 my ($v1_ent, @v2_epochs);
323 $v1_ent = $m->{$path};
324 @v2_epochs = grep(m!\A\Q$path\E/git/[0-9]+\.git\z!, keys %$m);
325 } while (!defined($v1_ent) && !@v2_epochs &&
326 $path =~ s!\A(/[^/]+)/!/! and $path_pfx .= $1);
327 ($path_pfx, $v1_ent ? $path : undef, @v2_epochs);
330 sub decode_manifest ($$$) {
331 my ($fh, $fn, $uri) = @_;
333 my $gz = do { local $/; <$fh> } // die "slurp($fn): $!";
334 gunzip(\$gz => \$js, MultiStream => 1) or
335 die "gunzip($uri): $GunzipError\n";
336 my $m = eval { PublicInbox::Config->json->decode($js) };
337 die "$uri: error decoding `$js': $@\n" if $@;
338 ref($m) eq 'HASH' or die "$uri unknown type: ".ref($m);
344 my $uri = URI->new($self->{src});
345 my $lei = $self->{lei};
346 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
347 my $path = $uri->path;
348 chop($path) eq '/' or die "BUG: $uri not canonicalized";
349 $uri->path($path . '/manifest.js.gz');
350 my $pdir = $lei->rel2abs($self->{dst});
351 $pdir =~ s!/[^/]+/?\z!!;
352 my $ft = File::Temp->new(TEMPLATE => 'm-XXXX',
353 UNLINK => 1, DIR => $pdir, SUFFIX => '.tmp');
354 my $fn = $ft->filename;
355 my ($bn) = ($fn =~ m!/([^/]+)\z!);
356 my $cmd = $curl->for_uri($lei, $uri, '-R', '-o', $bn);
357 my $opt = { -C => $pdir };
358 $opt->{$_} = $lei->{$_} for (0..2);
359 my $cerr = run_reap($lei, $cmd, $opt);
361 return try_scrape($self) if ($cerr >> 8) == 22; # 404 missing
362 return $lei->child_error($cerr, "@$cmd failed");
364 my $m = eval { decode_manifest($ft, $fn, $uri) };
367 return try_scrape($self);
369 my ($path_pfx, $v1_path, @v2_epochs) = deduce_epochs($m, $path);
371 # It may be possible to have v1 + v2 in parallel someday:
372 $lei->err(<<EOM) if defined $v1_path;
373 # `$v1_path' appears to be a v1 inbox while v2 epochs exist:
375 # ignoring $v1_path (use --inbox-version=1 to force v1 instead)
377 my %v2_epochs = map {
378 $uri->path($path_pfx.$_);
379 my ($n) = ("$uri" =~ m!/([0-9]+)\.git\z!);
382 clone_v2($self, \%v2_epochs);
383 } elsif (defined $v1_path) {
386 die "E: confused by <$uri>, possible matches:\n\t",
387 join(', ', sort keys %$m), "\n";
389 my $fin = "$self->{dst}/manifest.js.gz";
390 rename($fn, $fin) or die "E: rename($fn, $fin): $!";
391 $ft->unlink_on_destroy(0);
394 sub start_clone_url {
396 return try_manifest($self) if $self->{src} =~ m!\Ahttps?://!;
397 die "TODO: non-HTTP/HTTPS clone of $self->{src} not supported, yet";
400 sub do_mirror { # via wq_io_do
402 my $lei = $self->{lei};
404 my $iv = $lei->{opt}->{'inbox-version'};
406 return clone_v1($self) if $iv == 1;
407 return try_scrape($self) if $iv == 2;
408 die "bad --inbox-version=$iv\n";
410 return start_clone_url($self) if $self->{src} =~ m!://!;
411 die "TODO: cloning local directories not supported, yet";
413 $lei->fail($@) if $@;
417 my ($cls, $lei, $src, $dst) = @_;
418 my $self = bless { src => $src, dst => $dst }, $cls;
419 if ($src =~ m!https?://!) {
421 require PublicInbox::LeiCurl;
423 require PublicInbox::Lock;
424 require PublicInbox::Inbox;
425 require PublicInbox::Admin;
426 require PublicInbox::InboxWritable;
427 my ($op_c, $ops) = $lei->workers_start($self, 1);
429 $self->wq_io_do('do_mirror', []);
431 $lei->wait_wq_events($op_c, $ops);
434 sub ipc_atfork_child {
436 $self->{lei}->_lei_atfork_child;
437 $SIG{TERM} = sub { exit(128 + 15) }; # trigger OnDestroy $reap
438 $self->SUPER::ipc_atfork_child;
442 my ($dir, $ibx_ver) = @_;
443 my $f = "$dir/Makefile";
444 if (sysopen my $fh, $f, O_CREAT|O_EXCL|O_WRONLY) {
445 print $fh <<EOM or die "print($f) $!";
446 # This is a v$ibx_ver public-inbox, see the public-inbox-v$ibx_ver-format(5)
447 # manpage for more information on the format. This Makefile is
448 # intended as a familiar wrapper for users unfamiliar with
449 # public-inbox-* commands.
451 # See the respective manpages for public-inbox-fetch(1),
452 # public-inbox-index(1), etc for more information on
453 # some of the commands used by this Makefile.
455 # This Makefile will not be modified nor read by public-inbox,
456 # so you may edit it freely with your own convenience targets
457 # and notes. public-inbox-fetch will recreate it if removed.
459 print $fh <<'EOM' or die "print($f): $!";
460 # the default target:
462 @echo Common targets:
463 @echo ' make fetch - fetch from remote git repostorie(s)'
464 @echo ' make update - fetch and update index '
466 @echo Rarely needed targets:
467 @echo ' make reindex - may be needed for new features/bugfixes'
468 @echo ' make compact - rewrite Xapian storage to save space'
473 @if ! public-inbox-fetch --exit-code; \
476 test $$c -eq 127 && exit 0; \
478 elif test -f msgmap.sqlite3 || test -f public-inbox/msgmap.sqlite3; \
480 public-inbox-index; \
482 echo 'public-inbox index not initialized'; \
483 echo 'see public-inbox-index(1) man page'; \
486 public-inbox-index --reindex
490 .PHONY : help fetch update reindex compact
492 close $fh or die "close($f): $!";
494 die "open($f): $!" unless $!{EEXIST};