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($f)) {
23 warn("unlink($f): $!\n") unless $!{ENOENT};
25 if ($lei->{cmd} ne 'public-inbox-clone') {
26 $lei->lazy_cb('add-external', '_finish_'
27 )->($lei, $mrr->{dst});
29 $lei->qerr("# mirrored $mrr->{src} => $mrr->{dst}");
34 # for old installations without manifest.js.gz
37 my $uri = URI->new($self->{src});
38 my $lei = $self->{lei};
39 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
40 my $cmd = $curl->for_uri($lei, $uri, '--compressed');
41 my $opt = { 0 => $lei->{0}, 2 => $lei->{2} };
42 my $fh = popen_rd($cmd, undef, $opt);
43 my $html = do { local $/; <$fh> } // die "read(curl $uri): $!";
44 close($fh) or return $lei->child_error($?, "@$cmd failed");
46 # we grep with URL below, we don't want Subject/From headers
47 # making us clone random URLs
48 my @html = split(/<hr>/, $html);
49 my @urls = ($html[-1] =~ m!\bgit clone --mirror ([a-z\+]+://\S+)!g);
50 my $url = $uri->as_string;
51 chop($url) eq '/' or die "BUG: $uri not canonicalized";
53 # since this is for old instances w/o manifest.js.gz, try v1 first
54 return clone_v1($self) if grep(m!\A\Q$url\E/*\z!, @urls);
55 if (my @v2_urls = grep(m!\A\Q$url\E/[0-9]+\z!, @urls)) {
57 my ($n) = (m!/([0-9]+)\z!);
60 return clone_v2($self, \%v2_epochs);
63 # filter out common URLs served by WWW (e.g /$MSGID/T/)
64 if (@urls && $url =~ s!/+[^/]+\@[^/]+/.*\z!! &&
65 grep(m!\A\Q$url\E/*\z!, @urls)) {
67 E: confused by scraping <$uri>, did you mean <$url>?
71 E: confused by scraping <$uri>, got ambiguous results:
74 die "E: scraping <$uri> revealed nothing\n";
80 $opt->{$_} = $lei->{$_} for (0..2);
81 # we support "-c $key=$val" for arbitrary git config options
82 # e.g.: git -c http.proxy=socks5h://127.0.0.1:9050
83 push(@cmd, '-c', $_) for @{$lei->{opt}->{c} // []};
84 push @cmd, qw(clone --mirror);
85 push @cmd, '-q' if $lei->{opt}->{quiet};
86 push @cmd, '-v' if $lei->{opt}->{verbose};
87 # XXX any other options to support?
88 # --reference is tricky with multiple epochs...
92 sub _get_txt { # non-fatal
93 my ($self, $endpoint, $file) = @_;
94 my $uri = URI->new($self->{src});
95 my $lei = $self->{lei};
96 my $path = $uri->path;
97 chop($path) eq '/' or die "BUG: $uri not canonicalized";
98 $uri->path("$path/$endpoint");
99 my $ft = File::Temp->new(TEMPLATE => "$file-XXXX", DIR => $self->{dst});
100 my $f = $ft->filename;
101 my $opt = { 0 => $lei->{0}, 1 => $lei->{1}, 2 => $lei->{2} };
102 my $cmd = $self->{curl}->for_uri($lei, $uri,
103 qw(--compressed -R -o), $f);
104 my $cerr = run_reap($lei, $cmd, $opt);
105 return "$uri missing" if ($cerr >> 8) == 22;
106 return "# @$cmd failed (non-fatal)" if $cerr;
107 my $ce = "$self->{dst}/$file";
108 rename($f, $ce) or return "rename($f, $ce): $! (non-fatal)";
109 $ft->unlink_on_destroy(0);
113 # tries the relatively new /$INBOX/_/text/config/raw endpoint
116 my $dst = $self->{dst};
117 if (!-d $dst || !mkdir($dst)) {
119 File::Path::mkpath($dst);
120 -d $dst or die "mkpath($dst): $!\n";
122 my $err = _get_txt($self, qw(_/text/config/raw inbox.config.example));
123 return warn($err, "\n") if $err;
124 my $f = "$self->{dst}/inbox.config.example";
125 chmod((stat($f))[2] & 0444, $f) or die "chmod(a-w, $f): $!";
126 my $cfg = PublicInbox::Config->git_config_dump($f, $self->{lei}->{2});
127 my $ibx = $self->{ibx} = {};
128 for my $sec (grep(/\Apublicinbox\./, @{$cfg->{-section_order}})) {
129 for (qw(address newsgroup nntpmirror)) {
130 $ibx->{$_} = $cfg->{"$sec.$_"};
135 sub set_description ($) {
137 my $f = "$self->{dst}/description";
138 open my $fh, '+>>', $f or die "open($f): $!";
139 seek($fh, 0, SEEK_SET) or die "seek($f): $!";
140 chomp(my $d = do { local $/; <$fh> } // die "read($f): $!");
141 if ($d eq '($INBOX_DIR/description missing)' ||
142 $d =~ /^Unnamed repository/ || $d !~ /\S/) {
143 seek($fh, 0, SEEK_SET) or die "seek($f): $!";
144 truncate($fh, 0) or die "truncate($f): $!";
145 print $fh "mirror of $self->{src}\n" or die "print($f): $!";
146 close $fh or die "close($f): $!";
150 sub index_cloned_inbox {
151 my ($self, $iv) = @_;
152 my $lei = $self->{lei};
153 my $err = _get_txt($self, qw(description description));
154 warn($err, "\n") if $err; # non fatal
155 eval { set_description($self) };
158 # n.b. public-inbox-clone works w/o (SQLite || Xapian)
159 # lei is useless without Xapian + SQLite
160 if ($lei->{cmd} ne 'public-inbox-clone') {
161 my $ibx = delete($self->{ibx}) // {
162 address => [ 'lei@example.com' ],
165 $ibx->{inboxdir} = $self->{dst};
166 PublicInbox::Inbox->new($ibx);
167 PublicInbox::InboxWritable->new($ibx);
169 for my $sw ($lei->index_opt) {
170 my ($k) = ($sw =~ /\A([\w-]+)/);
171 $opt->{$k} = $lei->{opt}->{$k};
173 # force synchronous dwaitpid for v2:
174 local $PublicInbox::DS::in_loop = 0;
175 my $cfg = PublicInbox::Config->new(undef, $lei->{2});
176 my $env = PublicInbox::Admin::index_prepare($opt, $cfg);
177 local %ENV = (%ENV, %$env) if $env;
178 PublicInbox::Admin::progress_prepare($opt, $lei->{2});
179 PublicInbox::Admin::index_inbox($ibx, undef, $opt);
181 open my $x, '>', "$self->{dst}/mirror.done"; # for _wq_done_wait
185 my ($lei, $cmd, $opt) = @_;
186 $lei->qerr("# @$cmd");
187 $opt->{pgid} = 0 if $lei->{sock};
188 my $pid = spawn($cmd, undef, $opt);
189 my $reap = PublicInbox::OnDestroy->new($lei->can('sigint_reap'), $pid);
190 waitpid($pid, 0) == $pid or die "waitpid @$cmd: $!";
191 @$reap = (); # cancel reap
193 $? = 0; # don't let it influence normal exit
199 my $lei = $self->{lei};
200 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
201 my $uri = URI->new($self->{src});
202 defined($lei->{opt}->{epoch}) and
203 die "$uri is a v1 inbox, --epoch is not supported\n";
204 my $pfx = $curl->torsocks($lei, $uri) or return;
205 my $cmd = [ @$pfx, clone_cmd($lei, my $opt = {}),
206 $uri->as_string, $self->{dst} ];
207 my $cerr = run_reap($lei, $cmd, $opt);
208 return $lei->child_error($cerr, "@$cmd failed") if $cerr;
210 write_makefile($self->{dst}, 1);
211 index_cloned_inbox($self, 1);
214 sub parse_epochs ($$) {
215 my ($opt_epochs, $v2_epochs) = @_; # $epcohs "LOW..HIGH"
216 $opt_epochs // return; # undef => all epochs
217 my ($lo, $dotdot, $hi, @extra) = split(/(\.\.)/, $opt_epochs);
218 undef($lo) if ($lo // '') eq '';
219 my $re = qr/\A~?[0-9]+\z/;
220 if (@extra || (($lo // '0') !~ $re) ||
221 (($hi // '0') !~ $re) ||
222 !(grep(defined, $lo, $hi))) {
224 --epoch=$opt_epochs not in the form of `LOW..HIGH', `LOW..', nor `..HIGH'
227 my @n = sort { $a <=> $b } keys %$v2_epochs;
228 for (grep(defined, $lo, $hi)) {
231 "`$_' exceeds maximum available epoch ($n[-1])\n";
233 "`$_' is lower than minimum available epoch ($n[0])\n";
234 } elsif (/\A~([0-9]+)/) {
236 $n[$off] // die "`$_' is out of range\n";
238 } else { die "`$_' not understood\n" }
240 defined($lo) && defined($hi) && $lo > $hi and die
241 "low value (`$lo') exceeds high (`$hi')\n";
242 $lo //= $n[0] if $dotdot;
243 $hi //= $n[-1] if $dotdot;
247 if (defined $v2_epochs->{$_}) {
251 "# epoch $_ is not available (non-fatal, $lo..$hi)\n";
257 sub init_placeholder ($$) {
258 my ($src, $edst) = @_;
259 PublicInbox::Import::init_bare($edst);
260 my $f = "$edst/config";
261 open my $fh, '>>', $f or die "open($f): $!";
262 print $fh <<EOM or die "print($f): $!";
265 fetch = +refs/*:refs/*
268 ; This git epoch was created read-only and "public-inbox-fetch"
269 ; will not fetch updates for it unless write permission is added.
271 close $fh or die "close:($f): $!";
274 sub clone_v2 ($$;$) {
275 my ($self, $v2_epochs, $m) = @_; # $m => manifest.js.gz hashref
276 my $lei = $self->{lei};
277 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
278 my $pfx = $curl->torsocks($lei, (values %$v2_epochs)[0]) or return;
279 my $dst = $self->{dst};
280 my $want = parse_epochs($lei->{opt}->{epoch}, $v2_epochs);
281 my (@src_edst, @read_only, @skip_nr);
282 for my $nr (sort { $a <=> $b } keys %$v2_epochs) {
283 my $uri = $v2_epochs->{$nr};
284 my $src = $uri->as_string;
286 $src =~ m!/([0-9]+)(?:\.git)?\z! or die <<"";
287 failed to extract epoch number from $src
289 $1 + 0 == $nr or die "BUG: <$uri> miskeyed $1 != $nr";
290 $edst .= "/git/$nr.git";
291 if (!$want || $want->{$nr}) {
292 push @src_edst, $src, $edst;
293 } else { # create a placeholder so users only need to chmod +w
294 init_placeholder($src, $edst);
295 push @read_only, $edst;
299 if (@skip_nr) { # filter out the epochs we skipped
300 my $re = join('|', @skip_nr);
301 my @del = grep(m!/git/$re\.git\z!, keys %$m);
303 $self->{-culled_manifest} = 1;
305 my $lk = bless { lock_path => "$dst/inbox.lock" }, 'PublicInbox::Lock';
307 my $on_destroy = $lk->lock_for_scope($$);
308 my @cmd = clone_cmd($lei, my $opt = {});
309 while (my ($src, $edst) = splice(@src_edst, 0, 2)) {
310 my $cmd = [ @$pfx, @cmd, $src, $edst ];
311 my $cerr = run_reap($lei, $cmd, $opt);
312 return $lei->child_error($cerr, "@$cmd failed") if $cerr;
314 require PublicInbox::MultiGit;
315 my $mg = PublicInbox::MultiGit->new($dst, 'all.git', 'git');
316 $mg->fill_alternates;
317 for my $i ($mg->git_epochs) { $mg->epoch_cfg_set($i) }
318 for my $edst (@read_only) {
319 my @st = stat($edst) or die "stat($edst): $!";
320 chmod($st[2] & 0555, $edst) or die "chmod(a-w, $edst): $!";
322 write_makefile($self->{dst}, 2);
323 undef $on_destroy; # unlock
324 index_cloned_inbox($self, 2);
327 # PSGI mount prefixes and manifest.js.gz prefixes don't always align...
328 sub deduce_epochs ($$) {
330 my ($v1_ent, @v2_epochs);
334 $v1_ent = $m->{$path};
335 @v2_epochs = grep(m!\A\Q$path\E/git/[0-9]+\.git\z!, keys %$m);
336 } while (!defined($v1_ent) && !@v2_epochs &&
337 $path =~ s!\A(/[^/]+)/!/! and $path_pfx .= $1);
338 ($path_pfx, $v1_ent ? $path : undef, @v2_epochs);
341 sub decode_manifest ($$$) {
342 my ($fh, $fn, $uri) = @_;
344 my $gz = do { local $/; <$fh> } // die "slurp($fn): $!";
345 gunzip(\$gz => \$js, MultiStream => 1) or
346 die "gunzip($uri): $GunzipError\n";
347 my $m = eval { PublicInbox::Config->json->decode($js) };
348 die "$uri: error decoding `$js': $@\n" if $@;
349 ref($m) eq 'HASH' or die "$uri unknown type: ".ref($m);
355 my $uri = URI->new($self->{src});
356 my $lei = $self->{lei};
357 my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
358 my $path = $uri->path;
359 chop($path) eq '/' or die "BUG: $uri not canonicalized";
360 $uri->path($path . '/manifest.js.gz');
361 my $pdir = $lei->rel2abs($self->{dst});
362 $pdir =~ s!/[^/]+/?\z!!;
363 my $ft = File::Temp->new(TEMPLATE => 'm-XXXX',
364 UNLINK => 1, DIR => $pdir, SUFFIX => '.tmp');
365 my $fn = $ft->filename;
366 my ($bn) = ($fn =~ m!/([^/]+)\z!);
367 my $cmd = $curl->for_uri($lei, $uri, '-R', '-o', $bn);
368 my $opt = { -C => $pdir };
369 $opt->{$_} = $lei->{$_} for (0..2);
370 my $cerr = run_reap($lei, $cmd, $opt);
372 return try_scrape($self) if ($cerr >> 8) == 22; # 404 missing
373 return $lei->child_error($cerr, "@$cmd failed");
375 my $m = eval { decode_manifest($ft, $fn, $uri) };
378 return try_scrape($self);
380 my ($path_pfx, $v1_path, @v2_epochs) = deduce_epochs($m, $path);
382 # It may be possible to have v1 + v2 in parallel someday:
383 warn(<<EOM) if defined $v1_path;
384 # `$v1_path' appears to be a v1 inbox while v2 epochs exist:
386 # ignoring $v1_path (use --inbox-version=1 to force v1 instead)
388 my %v2_epochs = map {
389 $uri->path($path_pfx.$_);
390 my ($n) = ("$uri" =~ m!/([0-9]+)\.git\z!);
393 clone_v2($self, \%v2_epochs, $m);
394 } elsif (defined $v1_path) {
397 die "E: confused by <$uri>, possible matches:\n\t",
398 join(', ', sort keys %$m), "\n";
400 if (delete $self->{-culled_manifest}) { # set by clone_v2
401 # write the smaller manifest if epochs were skipped so
402 # users won't have to delete manifest if they +w an
403 # epoch they no longer want to skip
404 my $json = PublicInbox::Config->json->encode($m);
405 gzip(\$json => $fn) or die "gzip: $GzipError";
407 my $fin = "$self->{dst}/manifest.js.gz";
408 rename($fn, $fin) or die "E: rename($fn, $fin): $!";
409 $ft->unlink_on_destroy(0);
412 sub start_clone_url {
414 return try_manifest($self) if $self->{src} =~ m!\Ahttps?://!;
415 die "TODO: non-HTTP/HTTPS clone of $self->{src} not supported, yet";
418 sub do_mirror { # via wq_io_do
420 my $lei = $self->{lei};
422 my $iv = $lei->{opt}->{'inbox-version'};
424 return clone_v1($self) if $iv == 1;
425 return try_scrape($self) if $iv == 2;
426 die "bad --inbox-version=$iv\n";
428 return start_clone_url($self) if $self->{src} =~ m!://!;
429 die "TODO: cloning local directories not supported, yet";
431 $lei->fail($@) if $@;
435 my ($cls, $lei, $src, $dst) = @_;
436 my $self = bless { src => $src, dst => $dst }, $cls;
437 if ($src =~ m!https?://!) {
439 require PublicInbox::LeiCurl;
441 require PublicInbox::Lock;
442 require PublicInbox::Inbox;
443 require PublicInbox::Admin;
444 require PublicInbox::InboxWritable;
445 my ($op_c, $ops) = $lei->workers_start($self, 1);
447 $self->wq_io_do('do_mirror', []);
449 $lei->wait_wq_events($op_c, $ops);
452 sub ipc_atfork_child {
454 $self->{lei}->_lei_atfork_child;
455 $SIG{TERM} = sub { exit(128 + 15) }; # trigger OnDestroy $reap
456 $self->SUPER::ipc_atfork_child;
460 my ($dir, $ibx_ver) = @_;
461 my $f = "$dir/Makefile";
462 if (sysopen my $fh, $f, O_CREAT|O_EXCL|O_WRONLY) {
463 print $fh <<EOM or die "print($f) $!";
464 # This is a v$ibx_ver public-inbox, see the public-inbox-v$ibx_ver-format(5)
465 # manpage for more information on the format. This Makefile is
466 # intended as a familiar wrapper for users unfamiliar with
467 # public-inbox-* commands.
469 # See the respective manpages for public-inbox-fetch(1),
470 # public-inbox-index(1), etc for more information on
471 # some of the commands used by this Makefile.
473 # This Makefile will not be modified nor read by public-inbox,
474 # so you may edit it freely with your own convenience targets
475 # and notes. public-inbox-fetch will recreate it if removed.
477 print $fh <<'EOM' or die "print($f): $!";
478 # the default target:
480 @echo Common targets:
481 @echo ' make fetch - fetch from remote git repostorie(s)'
482 @echo ' make update - fetch and update index '
484 @echo Rarely needed targets:
485 @echo ' make reindex - may be needed for new features/bugfixes'
486 @echo ' make compact - rewrite Xapian storage to save space'
491 @if ! public-inbox-fetch --exit-code; \
494 test $$c -eq 127 && exit 0; \
496 elif test -f msgmap.sqlite3 || test -f public-inbox/msgmap.sqlite3; \
498 public-inbox-index; \
500 echo 'public-inbox index not initialized'; \
501 echo 'see public-inbox-index(1) man page'; \
504 public-inbox-index --reindex
508 .PHONY : help fetch update reindex compact
510 close $fh or die "close($f): $!";
512 die "open($f): $!" unless $!{EEXIST};