1 # Copyright (C) 2018 all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
4 # This interface wraps and mimics PublicInbox::Import
5 package PublicInbox::V2Writable;
8 use base qw(PublicInbox::Lock);
9 use PublicInbox::SearchIdxPart;
10 use PublicInbox::SearchIdxSkeleton;
11 use PublicInbox::MIME;
13 use PublicInbox::Import;
14 use PublicInbox::MID qw(mids);
15 use PublicInbox::ContentId qw(content_id content_digest);
16 use PublicInbox::Inbox;
18 # an estimate of the post-packed size to the raw uncompressed size
19 my $PACKING_FACTOR = 0.4;
21 # assume 2 cores if GNU nproc(1) is not available
23 int($ENV{NPROC} || `nproc 2>/dev/null` || 2);
27 my ($class, $v2ibx, $creat) = @_;
28 my $dir = $v2ibx->{mainrepo} or die "no mainrepo in inbox\n";
32 File::Path::mkpath($dir);
34 die "$dir does not exist\n";
39 my $xpfx = "$dir/xap" . PublicInbox::Search::SCHEMA_VERSION;
41 # always load existing partitions in case core count changes:
43 foreach my $part (<$xpfx/*>) {
44 -d $part && $part =~ m!/\d+\z! or next;
46 Search::Xapian::Database->new($part)->close;
51 $nparts = nproc() if ($nparts == 0);
55 im => undef, # PublicInbox::Import
56 xap_rw => undef, # PublicInbox::V2SearchIdx
58 partitions => $nparts,
61 lock_path => "$dir/inbox.lock",
62 # limit each repo to 1GB or so
63 rotate_bytes => int((1024 * 1024 * 1024) / $PACKING_FACTOR),
69 my ($self, $parallel) = @_;
70 $self->{parallel} = $parallel;
76 # returns undef on duplicate or spam
77 # mimics Import::add and wraps it for v2
79 my ($self, $mime, $check_cb) = @_;
83 $mime = $check_cb->($mime) or return;
86 # All pipes (> $^F) known to Perl 5.6+ have FD_CLOEXEC set,
87 # as does SQLite 3.4.1+ (released in 2007-07-20), and
88 # Xapian 1.3.2+ (released 2015-03-15).
89 # For the most part, we can spawn git-fast-import without
90 # leaking FDs to it...
94 my $num = num_for($self, $mime, \$mid0);
95 defined $num or return; # duplicate
96 defined $mid0 or die "BUG: $mid0 undefined\n";
97 my $im = $self->importer;
98 my $cmt = $im->add($mime);
99 $cmt = $im->get_mark($cmt);
100 my ($oid, $len, $msgref) = @{$im->{last_object}};
102 my $nparts = $self->{partitions};
103 my $part = $num % $nparts;
104 my $idx = $self->idx_part($part);
105 $idx->index_raw($len, $msgref, $num, $oid, $mid0, $mime);
106 my $n = $self->{transact_bytes} += $len;
107 if ($n > (PublicInbox::SearchIdx::BATCH_BYTES * $nparts)) {
115 my ($self, $mime, $mid0) = @_;
116 my $mids = mids($mime->header_obj);
118 my $mid = $mids->[0];
119 my $num = $self->{skel}->{mm}->mid_insert($mid);
120 if (defined $num) { # common case
125 # crap, Message-ID is already known, hope somebody just resent:
127 foreach my $m (@$mids) {
128 # read-only lookup now safe to do after above barrier
129 my $existing = $self->lookup_content($mime, $m);
130 # easy, don't store duplicates
131 # note: do not add more diagnostic info here since
132 # it gets noisy on public-inbox-watch restarts
137 warn "<$mid> reused for mismatched content\n";
139 # try the rest of the mids
140 foreach my $i (1..$#$mids) {
142 $num = $self->{skel}->{mm}->mid_insert($m);
144 warn "alternative <$m> for <$mid> found\n";
150 # none of the existing Message-IDs are good, generate a new one:
151 num_for_harder($self, $mime, $mid0);
155 my ($self, $mime, $mid0) = @_;
157 my $hdr = $mime->header_obj;
158 my $dig = content_digest($mime);
159 $$mid0 = PublicInbox::Import::digest2mid($dig);
160 my $num = $self->{skel}->{mm}->mid_insert($$mid0);
161 unless (defined $num) {
162 # it's hard to spoof the last Received: header
163 my @recvd = $hdr->header_raw('Received');
164 $dig->add("Received: $_") foreach (@recvd);
165 $$mid0 = PublicInbox::Import::digest2mid($dig);
166 $num = $self->{skel}->{mm}->mid_insert($$mid0);
168 # fall back to a random Message-ID and give up determinism:
169 until (defined($num)) {
171 $$mid0 = PublicInbox::Import::digest2mid($dig);
172 warn "using random Message-ID <$$mid0> as fallback\n";
173 $num = $self->{skel}->{mm}->mid_insert($$mid0);
176 PublicInbox::Import::prepend_mid($hdr, $$mid0);
181 my ($self, $part) = @_;
182 $self->{idx_parts}->[$part];
188 return if $self->{idx_parts};
189 my $ibx = $self->{-inbox};
191 # do not leak read-only FDs to child processes, we only have these
192 # FDs for duplicate detection so they should not be
193 # frequently activated.
194 delete $ibx->{$_} foreach (qw(git mm search));
198 # first time initialization, first we create the skeleton pipe:
199 my $skel = $self->{skel} = PublicInbox::SearchIdxSkeleton->new($self);
201 # need to create all parts before initializing msgmap FD
202 my $max = $self->{partitions} - 1;
203 my $idx = $self->{idx_parts} = [];
204 for my $i (0..$max) {
205 push @$idx, PublicInbox::SearchIdxPart->new($self, $i, $skel);
208 # Now that all subprocesses are up, we can open the FD for SQLite:
209 $skel->_msgmap_init->{dbh}->begin_work;
213 my ($self, $mime, $cmt_msg) = @_;
216 my $im = $self->importer;
217 my $ibx = $self->{-inbox};
218 my $srch = $ibx->search;
219 my $cid = content_id($mime);
220 my $skel = $self->{skel};
221 my $parts = $self->{idx_parts};
222 my $mm = $skel->{mm};
224 my $mids = mids($mime->header_obj);
226 # We avoid introducing new blobs into git since the raw content
227 # can be slightly different, so we do not need the user-supplied
228 # message now that we have the mids and content_id
231 foreach my $mid (@$mids) {
232 $srch->reopen->each_smsg_by_mid($mid, sub {
235 my $msg = $ibx->msg_by_smsg($smsg);
236 if (!defined($msg)) {
237 warn "broken smsg for $mid\n";
241 my $cur = PublicInbox::MIME->new($msg);
242 if (content_id($cur) eq $cid) {
243 $mm->num_delete($smsg->num);
244 # $removed should only be set once assuming
245 # no bugs in our deduplication code:
247 $removed->{mime} = $cur;
248 $im->remove(\$orig, $cmt_msg);
250 $removed->num; # memoize this for callers
252 my $oid = $smsg->{blob};
253 foreach my $idx (@$parts, $skel) {
254 $idx->remote_remove($oid, $mid);
266 my $locked = defined $self->{idx_parts};
267 my $im = delete $self->{im};
268 $im->done if $im; # PublicInbox::Import::done
269 $self->searchidx_checkpoint(0);
270 $self->lock_release if $locked;
275 my $im = $self->{im};
276 $im->checkpoint if $im; # PublicInbox::Import::checkpoint
277 $self->searchidx_checkpoint(1);
280 # issue a write barrier to ensure all data is visible to other processes
281 # and read-only ops. Order of data importance is: git > SQLite > Xapian
285 if (my $im = $self->{im}) {
288 my $skel = $self->{skel};
289 my $parts = $self->{idx_parts};
290 if ($parts && $skel) {
291 my $dbh = $skel->{mm}->{dbh};
292 $dbh->commit; # SQLite data is second in importance
294 # Now deal with Xapian
295 $skel->barrier_init(scalar(@$parts));
296 # each partition needs to issue a barrier command to skel:
297 $_->remote_barrier foreach @$parts;
299 $skel->barrier_wait; # wait for each Xapian partition
303 $self->{transact_bytes} = 0;
306 sub searchidx_checkpoint {
307 my ($self, $more) = @_;
309 # order matters, we can only close {skel} after all partitions
310 # are done because the partitions also write to {skel}
311 if (my $parts = $self->{idx_parts}) {
312 foreach my $idx (@$parts) {
313 $idx->remote_commit; # propagates commit to skel
314 $idx->remote_close unless $more;
316 delete $self->{idx_parts} unless $more;
319 if (my $skel = $self->{skel}) {
320 my $dbh = $skel->{mm}->{dbh};
326 delete $self->{skel};
329 $self->{transact_bytes} = 0;
333 my ($self, $new) = @_;
334 my $pfx = "$self->{-inbox}->{mainrepo}/git";
335 my $git_dir = "$pfx/$new.git";
336 die "$git_dir exists\n" if -e $git_dir;
337 my @cmd = (qw(git init --bare -q), $git_dir);
338 PublicInbox::Import::run_die(\@cmd);
340 my $all = "$self->{-inbox}->{mainrepo}/all.git";
342 @cmd = (qw(git init --bare -q), $all);
343 PublicInbox::Import::run_die(\@cmd);
344 @cmd = (qw/git config/, "--file=$all/config",
345 'repack.writeBitmaps', 'true');
346 PublicInbox::Import::run_die(\@cmd);
349 @cmd = (qw/git config/, "--file=$git_dir/config",
350 'include.path', '../../all.git/config');
351 PublicInbox::Import::run_die(\@cmd);
353 my $alt = "$all/objects/info/alternates";
354 my $new_obj_dir = "../../git/$new.git/objects";
357 open(my $fh, '<', $alt) or die "open < $alt: $!\n";
358 %alts = map { chomp; $_ => 1 } (<$fh>);
360 return $git_dir if $alts{$new_obj_dir};
361 open my $fh, '>>', $alt or die "open >> $alt: $!\n";
362 print $fh "$new_obj_dir\n" or die "print >> $alt: $!\n";
363 close $fh or die "close $alt: $!\n";
368 my ($self, $max) = @_;
370 my $pfx = "$self->{-inbox}->{mainrepo}/git";
371 return unless -d $pfx;
373 opendir my $dh, $pfx or die "opendir $pfx: $!\n";
374 while (defined(my $git_dir = readdir($dh))) {
375 $git_dir =~ m!\A(\d+)\.git\z! or next;
378 $latest = "$pfx/$git_dir";
386 my $im = $self->{im};
388 if ($im->{bytes_added} < $self->{rotate_bytes}) {
393 $self->searchidx_checkpoint(1);
395 my $git_dir = $self->git_init(++$self->{max_git});
396 my $git = PublicInbox::Git->new($git_dir);
397 return $self->import_init($git, 0);
402 my $latest = git_dir_latest($self, \$max);
403 if (defined $latest) {
404 my $git = PublicInbox::Git->new($latest);
405 my $packed_bytes = $git->packed_bytes;
406 if ($packed_bytes >= $self->{rotate_bytes}) {
409 $self->{max_git} = $max;
410 return $self->import_init($git, $packed_bytes);
413 $self->{max_git} = $new;
414 $latest = $self->git_init($new);
415 $self->import_init(PublicInbox::Git->new($latest), 0);
419 my ($self, $git, $packed_bytes) = @_;
420 my $im = PublicInbox::Import->new($git, undef, undef, $self->{-inbox});
421 $im->{bytes_added} = int($packed_bytes / $PACKING_FACTOR);
422 $im->{want_object_info} = 1;
423 $im->{lock_path} = undef;
424 $im->{path_type} = 'v2';
430 my ($mid, $cur, $new) = @_;
431 use File::Temp qw(tempfile);
432 use PublicInbox::Spawn qw(spawn);
434 my ($ah, $an) = tempfile('email-cur-XXXXXXXX', TMPDIR => 1);
435 print $ah $cur->as_string or die "print: $!";
436 close $ah or die "close: $!";
437 my ($bh, $bn) = tempfile('email-new-XXXXXXXX', TMPDIR => 1);
438 PublicInbox::Import::drop_unwanted_headers($new);
439 print $bh $new->as_string or die "print: $!";
440 close $bh or die "close: $!";
441 my $cmd = [ qw(diff -u), $an, $bn ];
442 print STDERR "# MID conflict <$mid>\n";
443 my $pid = spawn($cmd, undef, { 1 => 2 });
444 defined $pid or die "diff failed to spawn $!";
445 waitpid($pid, 0) == $pid or die "diff did not finish";
450 my ($self, $mime, $mid) = @_;
451 my $ibx = $self->{-inbox};
453 my $srch = $ibx->search->reopen;
454 my $cid = content_id($mime);
456 $srch->each_smsg_by_mid($mid, sub {
459 my $msg = $ibx->msg_by_smsg($smsg);
460 if (!defined($msg)) {
461 warn "broken smsg for $mid\n";
464 my $cur = PublicInbox::MIME->new($msg);
465 if (content_id($cur) eq $cid) {
466 $smsg->{mime} = $cur;
468 return 0; # break out of loop
471 # XXX DEBUG_DIFF is experimental and may be removed
472 diff($mid, $cur, $mime) if $ENV{DEBUG_DIFF};
481 my $fh = delete $self->{reindex_pipe};
483 if (my $parts = $self->{idx_parts}) {
484 $_->atfork_child foreach @$parts;
486 if (my $im = $self->{im}) {
492 my ($self, $D, $git, $oid) = @_;
493 my $msgref = $git->cat_file($oid);
494 my $mime = PublicInbox::MIME->new($$msgref);
495 my $mids = mids($mime->header_obj);
496 my $cid = content_id($mime);
497 foreach my $mid (@$mids) {
503 my ($self, $mm_tmp, $D, $git, $oid, $regen) = @_;
505 my $msgref = $git->cat_file($oid, \$len);
506 my $mime = PublicInbox::MIME->new($$msgref);
507 my $mids = mids($mime->header_obj);
508 my $cid = content_id($mime);
510 # get the NNTP article number we used before, highest number wins
511 # and gets deleted from mm_tmp;
515 foreach my $mid (@$mids) {
516 $del += (delete $D->{$mid.$cid} || 0);
517 my $n = $mm_tmp->num_for($mid);
518 if (defined $n && $n > $num) {
523 if (!defined($mid0) && $regen && !$del) {
525 die "BUG: ran out of article numbers\n" if $num <= 0;
526 my $mm = $self->{skel}->{mm};
527 foreach my $mid (@$mids) {
528 if ($mm->mid_set($num, $mid) == 1) {
533 if (!defined($mid0)) {
534 my $id = '<' . join('> <', @$mids) . '>';
535 warn "Message-Id $id unusable for $num\n";
539 if (!defined($mid0) || $del) {
540 if (!defined($mid0) && $del) { # expected for deletes
545 my $id = '<' . join('> <', @$mids) . '>';
547 warn "Skipping $id, no article number found\n";
548 if ($del && defined($mid0)) {
549 warn "$id was deleted $del " .
550 "time(s) but mapped to article #$num\n";
555 $mm_tmp->mid_delete($mid0) or
556 die "failed to delete <$mid0> for article #$num\n";
558 my $nparts = $self->{partitions};
559 my $part = $num % $nparts;
560 my $idx = $self->idx_part($part);
561 $idx->index_raw($len, $msgref, $num, $oid, $mid0, $mime);
562 my $n = $self->{transact_bytes} += $len;
563 if ($n > (PublicInbox::SearchIdx::BATCH_BYTES * $nparts)) {
565 $mm_tmp->atfork_prepare;
566 $self->done; # release lock
567 # allow -watch or -mda to write...
568 $self->idx_init; # reacquire lock
569 $mm_tmp->atfork_parent;
574 my ($self, $regen) = @_;
575 my $ibx = $self->{-inbox};
576 my $pfx = "$ibx->{mainrepo}/git";
578 my $latest = git_dir_latest($self, \$max_git);
579 return unless defined $latest;
580 my $head = $ibx->{ref_head} || 'refs/heads/master';
581 $self->idx_init; # acquire lock
582 my $x40 = qr/[a-f0-9]{40}/;
583 my $mm_tmp = $self->{skel}->{mm}->tmp_clone;
585 my (undef, $max) = $mm_tmp->minmax;
586 unless (defined $max) {
589 "empty msgmap.sqlite3, regenerating article numbers\n";
592 my $tip; # latest commit out of all git repos
595 for (my $cur = $max_git; $cur >= 0; $cur--) {
596 die "already reindexing!\n" if $self->{reindex_pipe};
597 my $git = PublicInbox::Git->new("$pfx/$cur.git");
598 chomp($tip = $git->qx('rev-parse', $head)) unless $tip;
599 my $h = $cur == $max_git ? $tip : $head;
600 my @count = ('rev-list', '--count', $h, '--', 'm');
601 $regen_max += $git->qx(@count);
603 die "No messages found in $pfx/*.git, bug?\n" unless $regen_max;
604 $regen = \$regen_max;
607 my @cmd = qw(log --raw -r --pretty=tformat:%h
608 --no-notes --no-color --no-abbrev);
610 # if we are regenerating, we must not use a newer tip commit than what
611 # the regeneration counter used:
614 # work backwards through history
615 for (my $cur = $max_git; $cur >= 0; $cur--) {
616 die "already reindexing!\n" if delete $self->{reindex_pipe};
618 my $git_dir = "$pfx/$cur.git";
619 my $git = PublicInbox::Git->new($git_dir);
620 my $h = $cur == $max_git ? $tip : $head;
621 my $fh = $self->{reindex_pipe} = $git->popen(@cmd, $h);
625 } elsif (/\A:\d{6} 100644 $x40 ($x40) [AM]\tm$/o) {
626 $self->reindex_oid($mm_tmp, $D, $git, $1,
628 } elsif (m!\A:\d{6} 100644 $x40 ($x40) [AM]\t_/D$!o) {
629 $self->mark_deleted($D, $git, $1);
632 delete $self->{reindex_pipe};
634 my ($min, $max) = $mm_tmp->minmax;
635 defined $max and die "leftover article numbers at $min..$max\n";