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 PublicInbox::Config;
10 use IO::Uncompress::Gunzip qw(gunzip $GunzipError);
11 use IO::Compress::Gzip qw(gzip $GzipError);
12 use PublicInbox::Spawn qw(popen_rd spawn run_die);
14 use Fcntl qw(SEEK_SET O_CREAT O_EXCL O_WRONLY);
16 sub _wq_done_wait { # dwaitpid callback (via wq_eof)
18 my ($mrr, $lei) = @$arg;
19 my $f = "$mrr->{dst}/mirror.done";
21 $lei->child_error($?);
22 } elsif (!unlink($f)) {
23 $lei->err("unlink($f): $!") unless $!{ENOENT};
25 if ($lei->{cmd} ne 'public-inbox-clone') {
26 $lei->add_external_finish($mrr->{dst});
28 $lei->qerr("# mirrored $mrr->{src} => $mrr->{dst}");
33 # for old installations without manifest.js.gz
36 my $uri = URI->new($self->{src});
37 my $lei = $self->{lei};
38 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
39 my $cmd = $curl->for_uri($lei, $uri, '--compressed');
40 my $opt = { 0 => $lei->{0}, 2 => $lei->{2} };
41 my $fh = popen_rd($cmd, undef, $opt);
42 my $html = do { local $/; <$fh> } // die "read(curl $uri): $!";
43 close($fh) or return $lei->child_error($?, "@$cmd failed");
45 # we grep with URL below, we don't want Subject/From headers
46 # making us clone random URLs
47 my @html = split(/<hr>/, $html);
48 my @urls = ($html[-1] =~ m!\bgit clone --mirror ([a-z\+]+://\S+)!g);
49 my $url = $uri->as_string;
50 chop($url) eq '/' or die "BUG: $uri not canonicalized";
52 # since this is for old instances w/o manifest.js.gz, try v1 first
53 return clone_v1($self) if grep(m!\A\Q$url\E/*\z!, @urls);
54 if (my @v2_urls = grep(m!\A\Q$url\E/[0-9]+\z!, @urls)) {
56 my ($n) = (m!/([0-9]+)\z!);
59 return clone_v2($self, \%v2_epochs);
62 # filter out common URLs served by WWW (e.g /$MSGID/T/)
63 if (@urls && $url =~ s!/+[^/]+\@[^/]+/.*\z!! &&
64 grep(m!\A\Q$url\E/*\z!, @urls)) {
66 E: confused by scraping <$uri>, did you mean <$url>?
70 E: confused by scraping <$uri>, got ambiguous results:
73 die "E: scraping <$uri> revealed nothing\n";
79 $opt->{$_} = $lei->{$_} for (0..2);
80 # we support "-c $key=$val" for arbitrary git config options
81 # e.g.: git -c http.proxy=socks5h://127.0.0.1:9050
82 push(@cmd, '-c', $_) for @{$lei->{opt}->{c} // []};
83 push @cmd, qw(clone --mirror);
84 push @cmd, '-q' if $lei->{opt}->{quiet};
85 push @cmd, '-v' if $lei->{opt}->{verbose};
86 # XXX any other options to support?
87 # --reference is tricky with multiple epochs...
91 sub _get_txt { # non-fatal
92 my ($self, $endpoint, $file) = @_;
93 my $uri = URI->new($self->{src});
94 my $lei = $self->{lei};
95 my $path = $uri->path;
96 chop($path) eq '/' or die "BUG: $uri not canonicalized";
97 $uri->path("$path/$endpoint");
98 my $cmd = $self->{curl}->for_uri($lei, $uri, '--compressed');
99 my $ce = "$self->{dst}/$file";
100 my $ft = File::Temp->new(TEMPLATE => "$file-XXXX",
101 UNLINK => 1, DIR => $self->{dst});
102 my $opt = { 0 => $lei->{0}, 1 => $ft, 2 => $lei->{2} };
103 my $cerr = run_reap($lei, $cmd, $opt);
104 return "$uri missing" if ($cerr >> 8) == 22;
105 return "# @$cmd failed (non-fatal)" if $cerr;
106 my $f = $ft->filename;
107 rename($f, $ce) or return "rename($f, $ce): $! (non-fatal)";
108 $ft->unlink_on_destroy(0);
112 # tries the relatively new /$INBOX/_/text/config/raw endpoint
115 my $dst = $self->{dst};
116 if (!-d $dst || !mkdir($dst)) {
118 File::Path::mkpath($dst);
119 -d $dst or die "mkpath($dst): $!\n";
121 my $err = _get_txt($self, qw(_/text/config/raw inbox.config.example));
122 return $self->{lei}->err($err) if $err;
123 my $f = "$self->{dst}/inbox.config.example";
124 my $cfg = PublicInbox::Config->git_config_dump($f, $self->{lei}->{2});
125 my $ibx = $self->{ibx} = {};
126 for my $sec (grep(/\Apublicinbox\./, @{$cfg->{-section_order}})) {
127 for (qw(address newsgroup nntpmirror)) {
128 $ibx->{$_} = $cfg->{"$sec.$_"};
133 sub set_description ($) {
135 my $f = "$self->{dst}/description";
136 open my $fh, '+>>', $f or die "open($f): $!";
137 seek($fh, 0, SEEK_SET) or die "seek($f): $!";
138 chomp(my $d = do { local $/; <$fh> } // die "read($f): $!");
139 if ($d eq '($INBOX_DIR/description missing)' ||
140 $d =~ /^Unnamed repository/ || $d !~ /\S/) {
141 seek($fh, 0, SEEK_SET) or die "seek($f): $!";
142 truncate($fh, 0) or die "truncate($f): $!";
143 print $fh "mirror of $self->{src}\n" or die "print($f): $!";
144 close $fh or die "close($f): $!";
148 sub index_cloned_inbox {
149 my ($self, $iv) = @_;
150 my $lei = $self->{lei};
151 my $err = _get_txt($self, qw(description description));
152 $lei->err($err) if $err; # non fatal
153 eval { set_description($self) };
156 # n.b. public-inbox-clone works w/o (SQLite || Xapian)
157 # lei is useless without Xapian + SQLite
158 if ($lei->{cmd} ne 'public-inbox-clone') {
159 my $ibx = delete($self->{ibx}) // {
160 address => [ 'lei@example.com' ],
163 $ibx->{inboxdir} = $self->{dst};
164 PublicInbox::Inbox->new($ibx);
165 PublicInbox::InboxWritable->new($ibx);
167 for my $sw ($lei->index_opt) {
168 my ($k) = ($sw =~ /\A([\w-]+)/);
169 $opt->{$k} = $lei->{opt}->{$k};
171 # force synchronous dwaitpid for v2:
172 local $PublicInbox::DS::in_loop = 0;
173 my $cfg = PublicInbox::Config->new(undef, $lei->{2});
174 my $env = PublicInbox::Admin::index_prepare($opt, $cfg);
175 local %ENV = (%ENV, %$env) if $env;
176 PublicInbox::Admin::progress_prepare($opt, $lei->{2});
177 PublicInbox::Admin::index_inbox($ibx, undef, $opt);
179 open my $x, '>', "$self->{dst}/mirror.done"; # for _wq_done_wait
183 my ($lei, $cmd, $opt) = @_;
184 $lei->qerr("# @$cmd");
185 $opt->{pgid} = 0 if $lei->{sock};
186 my $pid = spawn($cmd, undef, $opt);
187 my $reap = PublicInbox::OnDestroy->new($lei->can('sigint_reap'), $pid);
188 waitpid($pid, 0) == $pid or die "waitpid @$cmd: $!";
189 @$reap = (); # cancel reap
191 $? = 0; # don't let it influence normal exit
197 my $lei = $self->{lei};
198 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
199 my $uri = URI->new($self->{src});
200 defined($lei->{opt}->{epoch}) and
201 die "$uri is a v1 inbox, --epoch is not supported\n";
202 my $pfx = $curl->torsocks($lei, $uri) or return;
203 my $cmd = [ @$pfx, clone_cmd($lei, my $opt = {}),
204 $uri->as_string, $self->{dst} ];
205 my $cerr = run_reap($lei, $cmd, $opt);
206 return $lei->child_error($cerr, "@$cmd failed") if $cerr;
208 write_makefile($self->{dst}, 1);
209 index_cloned_inbox($self, 1);
212 sub parse_epochs ($$) {
213 my ($opt_epochs, $v2_epochs) = @_; # $epcohs "LOW..HIGH"
214 $opt_epochs // return; # undef => all epochs
215 my ($lo, $dotdot, $hi, @extra) = split(/(\.\.)/, $opt_epochs);
216 undef($lo) if ($lo // '') eq '';
217 my $re = qr/\A~?[0-9]+\z/;
218 if (@extra || (($lo // '0') !~ $re) ||
219 (($hi // '0') !~ $re) ||
220 !(grep(defined, $lo, $hi))) {
222 --epoch=$opt_epochs not in the form of `LOW..HIGH', `LOW..', nor `..HIGH'
225 my @n = sort { $a <=> $b } keys %$v2_epochs;
226 for (grep(defined, $lo, $hi)) {
229 "`$_' exceeds maximum available epoch ($n[-1])\n";
231 "`$_' is lower than minimum available epoch ($n[0])\n";
232 } elsif (/\A~([0-9]+)/) {
234 $n[$off] // die "`$_' is out of range\n";
236 } else { die "`$_' not understood\n" }
238 defined($lo) && defined($hi) && $lo > $hi and die
239 "low value (`$lo') exceeds high (`$hi')\n";
240 $lo //= $n[0] if $dotdot;
241 $hi //= $n[-1] if $dotdot;
245 if (defined $v2_epochs->{$_}) {
249 "# epoch $_ is not available (non-fatal, $lo..$hi)\n";
255 sub init_placeholder ($$) {
256 my ($src, $edst) = @_;
257 PublicInbox::Import::init_bare($edst);
258 my $f = "$edst/config";
259 open my $fh, '>>', $f or die "open($f): $!";
260 print $fh <<EOM or die "print($f): $!";
263 fetch = +refs/*:refs/*
266 ; This git epoch was created read-only and "public-inbox-fetch"
267 ; will not fetch updates for it unless write permission is added.
269 close $fh or die "close:($f): $!";
272 sub clone_v2 ($$;$) {
273 my ($self, $v2_epochs, $m) = @_; # $m => manifest.js.gz hashref
274 my $lei = $self->{lei};
275 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
276 my $pfx = $curl->torsocks($lei, (values %$v2_epochs)[0]) or return;
277 my $dst = $self->{dst};
278 my $want = parse_epochs($lei->{opt}->{epoch}, $v2_epochs);
279 my (@src_edst, @read_only, @skip_nr);
280 for my $nr (sort { $a <=> $b } keys %$v2_epochs) {
281 my $uri = $v2_epochs->{$nr};
282 my $src = $uri->as_string;
284 $src =~ m!/([0-9]+)(?:\.git)?\z! or die <<"";
285 failed to extract epoch number from $src
287 $1 + 0 == $nr or die "BUG: <$uri> miskeyed $1 != $nr";
288 $edst .= "/git/$nr.git";
289 if (!$want || $want->{$nr}) {
290 push @src_edst, $src, $edst;
291 } else { # create a placeholder so users only need to chmod +w
292 init_placeholder($src, $edst);
293 push @read_only, $edst;
297 if (@skip_nr) { # filter out the epochs we skipped
298 my $re = join('|', @skip_nr);
299 my @del = grep(m!/git/$re\.git\z!, keys %$m);
301 $self->{-culled_manifest} = 1;
303 my $lk = bless { lock_path => "$dst/inbox.lock" }, 'PublicInbox::Lock';
305 my $on_destroy = $lk->lock_for_scope($$);
306 my @cmd = clone_cmd($lei, my $opt = {});
307 while (my ($src, $edst) = splice(@src_edst, 0, 2)) {
308 my $cmd = [ @$pfx, @cmd, $src, $edst ];
309 my $cerr = run_reap($lei, $cmd, $opt);
310 return $lei->child_error($cerr, "@$cmd failed") if $cerr;
312 require PublicInbox::MultiGit;
313 my $mg = PublicInbox::MultiGit->new($dst, 'all.git', 'git');
314 $mg->fill_alternates;
315 for my $i ($mg->git_epochs) { $mg->epoch_cfg_set($i) }
316 for my $edst (@read_only) {
317 my @st = stat($edst) or die "stat($edst): $!";
318 chmod($st[2] & 0555, $edst) or die "chmod(a-w, $edst): $!";
320 write_makefile($self->{dst}, 2);
321 undef $on_destroy; # unlock
322 index_cloned_inbox($self, 2);
325 # PSGI mount prefixes and manifest.js.gz prefixes don't always align...
326 sub deduce_epochs ($$) {
328 my ($v1_ent, @v2_epochs);
332 $v1_ent = $m->{$path};
333 @v2_epochs = grep(m!\A\Q$path\E/git/[0-9]+\.git\z!, keys %$m);
334 } while (!defined($v1_ent) && !@v2_epochs &&
335 $path =~ s!\A(/[^/]+)/!/! and $path_pfx .= $1);
336 ($path_pfx, $v1_ent ? $path : undef, @v2_epochs);
339 sub decode_manifest ($$$) {
340 my ($fh, $fn, $uri) = @_;
342 my $gz = do { local $/; <$fh> } // die "slurp($fn): $!";
343 gunzip(\$gz => \$js, MultiStream => 1) or
344 die "gunzip($uri): $GunzipError\n";
345 my $m = eval { PublicInbox::Config->json->decode($js) };
346 die "$uri: error decoding `$js': $@\n" if $@;
347 ref($m) eq 'HASH' or die "$uri unknown type: ".ref($m);
353 my $uri = URI->new($self->{src});
354 my $lei = $self->{lei};
355 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
356 my $path = $uri->path;
357 chop($path) eq '/' or die "BUG: $uri not canonicalized";
358 $uri->path($path . '/manifest.js.gz');
359 my $pdir = $lei->rel2abs($self->{dst});
360 $pdir =~ s!/[^/]+/?\z!!;
361 my $ft = File::Temp->new(TEMPLATE => 'm-XXXX',
362 UNLINK => 1, DIR => $pdir, SUFFIX => '.tmp');
363 my $fn = $ft->filename;
364 my ($bn) = ($fn =~ m!/([^/]+)\z!);
365 my $cmd = $curl->for_uri($lei, $uri, '-R', '-o', $bn);
366 my $opt = { -C => $pdir };
367 $opt->{$_} = $lei->{$_} for (0..2);
368 my $cerr = run_reap($lei, $cmd, $opt);
370 return try_scrape($self) if ($cerr >> 8) == 22; # 404 missing
371 return $lei->child_error($cerr, "@$cmd failed");
373 my $m = eval { decode_manifest($ft, $fn, $uri) };
376 return try_scrape($self);
378 my ($path_pfx, $v1_path, @v2_epochs) = deduce_epochs($m, $path);
380 # It may be possible to have v1 + v2 in parallel someday:
381 $lei->err(<<EOM) if defined $v1_path;
382 # `$v1_path' appears to be a v1 inbox while v2 epochs exist:
384 # ignoring $v1_path (use --inbox-version=1 to force v1 instead)
386 my %v2_epochs = map {
387 $uri->path($path_pfx.$_);
388 my ($n) = ("$uri" =~ m!/([0-9]+)\.git\z!);
391 clone_v2($self, \%v2_epochs, $m);
392 } elsif (defined $v1_path) {
395 die "E: confused by <$uri>, possible matches:\n\t",
396 join(', ', sort keys %$m), "\n";
398 if (delete $self->{-culled_manifest}) { # set by clone_v2
399 # write the smaller manifest if epochs were skipped so
400 # users won't have to delete manifest if they +w an
401 # epoch they no longer want to skip
402 my $json = PublicInbox::Config->json->encode($m);
403 gzip(\$json => $fn) or die "gzip: $GzipError";
405 my $fin = "$self->{dst}/manifest.js.gz";
406 rename($fn, $fin) or die "E: rename($fn, $fin): $!";
407 $ft->unlink_on_destroy(0);
410 sub start_clone_url {
412 return try_manifest($self) if $self->{src} =~ m!\Ahttps?://!;
413 die "TODO: non-HTTP/HTTPS clone of $self->{src} not supported, yet";
416 sub do_mirror { # via wq_io_do
418 my $lei = $self->{lei};
420 my $iv = $lei->{opt}->{'inbox-version'};
422 return clone_v1($self) if $iv == 1;
423 return try_scrape($self) if $iv == 2;
424 die "bad --inbox-version=$iv\n";
426 return start_clone_url($self) if $self->{src} =~ m!://!;
427 die "TODO: cloning local directories not supported, yet";
429 $lei->fail($@) if $@;
433 my ($cls, $lei, $src, $dst) = @_;
434 my $self = bless { src => $src, dst => $dst }, $cls;
435 if ($src =~ m!https?://!) {
437 require PublicInbox::LeiCurl;
439 require PublicInbox::Lock;
440 require PublicInbox::Inbox;
441 require PublicInbox::Admin;
442 require PublicInbox::InboxWritable;
443 my ($op_c, $ops) = $lei->workers_start($self, 1);
445 $self->wq_io_do('do_mirror', []);
447 $lei->wait_wq_events($op_c, $ops);
450 sub ipc_atfork_child {
452 $self->{lei}->_lei_atfork_child;
453 $SIG{TERM} = sub { exit(128 + 15) }; # trigger OnDestroy $reap
454 $self->SUPER::ipc_atfork_child;
458 my ($dir, $ibx_ver) = @_;
459 my $f = "$dir/Makefile";
460 if (sysopen my $fh, $f, O_CREAT|O_EXCL|O_WRONLY) {
461 print $fh <<EOM or die "print($f) $!";
462 # This is a v$ibx_ver public-inbox, see the public-inbox-v$ibx_ver-format(5)
463 # manpage for more information on the format. This Makefile is
464 # intended as a familiar wrapper for users unfamiliar with
465 # public-inbox-* commands.
467 # See the respective manpages for public-inbox-fetch(1),
468 # public-inbox-index(1), etc for more information on
469 # some of the commands used by this Makefile.
471 # This Makefile will not be modified nor read by public-inbox,
472 # so you may edit it freely with your own convenience targets
473 # and notes. public-inbox-fetch will recreate it if removed.
475 print $fh <<'EOM' or die "print($f): $!";
476 # the default target:
478 @echo Common targets:
479 @echo ' make fetch - fetch from remote git repostorie(s)'
480 @echo ' make update - fetch and update index '
482 @echo Rarely needed targets:
483 @echo ' make reindex - may be needed for new features/bugfixes'
484 @echo ' make compact - rewrite Xapian storage to save space'
489 @if ! public-inbox-fetch --exit-code; \
492 test $$c -eq 127 && exit 0; \
494 elif test -f msgmap.sqlite3 || test -f public-inbox/msgmap.sqlite3; \
496 public-inbox-index; \
498 echo 'public-inbox index not initialized'; \
499 echo 'see public-inbox-index(1) man page'; \
502 public-inbox-index --reindex
506 .PHONY : help fetch update reindex compact
508 close $fh or die "close($f): $!";
510 die "open($f): $!" unless $!{EEXIST};