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::MIME;
12 use PublicInbox::Import;
13 use PublicInbox::MID qw(mids);
14 use PublicInbox::ContentId qw(content_id content_digest);
15 use PublicInbox::Inbox;
16 use PublicInbox::OverIdxFork;
17 use PublicInbox::Msgmap;
18 use PublicInbox::Spawn;
21 # an estimate of the post-packed size to the raw uncompressed size
22 my $PACKING_FACTOR = 0.4;
24 # assume 2 cores if GNU nproc(1) is not available
26 int($ENV{NPROC} || `nproc 2>/dev/null` || 2);
29 sub count_partitions ($) {
32 my $xpfx = $self->{xpfx};
34 # always load existing partitions in case core count changes:
35 # Also, partition count may change while -watch is running
38 foreach my $part (<$xpfx/*>) {
39 -d $part && $part =~ m!/\d+\z! or next;
41 Search::Xapian::Database->new($part)->close;
50 my ($class, $v2ibx, $creat) = @_;
51 my $dir = $v2ibx->{mainrepo} or die "no mainrepo in inbox\n";
55 File::Path::mkpath($dir);
57 die "$dir does not exist\n";
61 $v2ibx = PublicInbox::InboxWritable->new($v2ibx);
63 my $xpfx = "$dir/xap" . PublicInbox::Search::SCHEMA_VERSION;
66 im => undef, # PublicInbox::Import
70 over => PublicInbox::OverIdxFork->new("$xpfx/over.sqlite3"),
71 lock_path => "$dir/inbox.lock",
72 # limit each repo to 1GB or so
73 rotate_bytes => int((1024 * 1024 * 1024) / $PACKING_FACTOR),
76 $self->{partitions} = count_partitions($self) || nproc();
81 my ($self, $parallel) = @_;
82 $self->{parallel} = $parallel;
85 git_dir_latest($self, \$max_git);
86 $self->git_init($max_git >= 0 ? $max_git : 0);
90 # returns undef on duplicate or spam
91 # mimics Import::add and wraps it for v2
93 my ($self, $mime, $check_cb) = @_;
97 $mime = $check_cb->($mime) or return;
100 # All pipes (> $^F) known to Perl 5.6+ have FD_CLOEXEC set,
101 # as does SQLite 3.4.1+ (released in 2007-07-20), and
102 # Xapian 1.3.2+ (released 2015-03-15).
103 # For the most part, we can spawn git-fast-import without
104 # leaking FDs to it...
108 my $num = num_for($self, $mime, \$mid0);
109 defined $num or return; # duplicate
110 defined $mid0 or die "BUG: $mid0 undefined\n";
111 my $im = $self->importer;
112 my $cmt = $im->add($mime);
113 $cmt = $im->get_mark($cmt);
114 my ($oid, $len, $msgref) = @{$im->{last_object}};
116 my $nparts = $self->{partitions};
117 my $part = $num % $nparts;
118 $self->{last_commit}->[$self->{max_git}] = $cmt;
119 my $idx = $self->idx_part($part);
120 $idx->index_raw($len, $msgref, $num, $oid, $mid0, $mime);
121 my $n = $self->{transact_bytes} += $len;
122 if ($n > (PublicInbox::SearchIdx::BATCH_BYTES * $nparts)) {
130 my ($self, $mime, $mid0) = @_;
131 my $mids = mids($mime->header_obj);
133 my $mid = $mids->[0];
134 my $num = $self->{mm}->mid_insert($mid);
135 if (defined $num) { # common case
140 # crap, Message-ID is already known, hope somebody just resent:
142 foreach my $m (@$mids) {
143 # read-only lookup now safe to do after above barrier
144 my $existing = $self->lookup_content($mime, $m);
145 # easy, don't store duplicates
146 # note: do not add more diagnostic info here since
147 # it gets noisy on public-inbox-watch restarts
152 warn "<$mid> reused for mismatched content\n";
154 # try the rest of the mids
155 for(my $i = $#$mids; $i >= 1; $i--) {
157 $num = $self->{mm}->mid_insert($m);
159 warn "alternative <$m> for <$mid> found\n";
165 # none of the existing Message-IDs are good, generate a new one:
166 num_for_harder($self, $mime, $mid0);
170 my ($self, $mime, $mid0) = @_;
172 my $hdr = $mime->header_obj;
173 my $dig = content_digest($mime);
174 $$mid0 = PublicInbox::Import::digest2mid($dig);
175 my $num = $self->{mm}->mid_insert($$mid0);
176 unless (defined $num) {
177 # it's hard to spoof the last Received: header
178 my @recvd = $hdr->header_raw('Received');
179 $dig->add("Received: $_") foreach (@recvd);
180 $$mid0 = PublicInbox::Import::digest2mid($dig);
181 $num = $self->{mm}->mid_insert($$mid0);
183 # fall back to a random Message-ID and give up determinism:
184 until (defined($num)) {
186 $$mid0 = PublicInbox::Import::digest2mid($dig);
187 warn "using random Message-ID <$$mid0> as fallback\n";
188 $num = $self->{mm}->mid_insert($$mid0);
191 PublicInbox::Import::append_mid($hdr, $$mid0);
196 my ($self, $part) = @_;
197 $self->{idx_parts}->[$part];
203 return if $self->{idx_parts};
204 my $ibx = $self->{-inbox};
206 # do not leak read-only FDs to child processes, we only have these
207 # FDs for duplicate detection so they should not be
208 # frequently activated.
209 delete $ibx->{$_} foreach (qw(git mm search));
211 my $over = $self->{over};
213 $ibx->with_umask(sub {
215 $over->create($self);
217 # -compact can change partition count while -watch is idle
218 my $nparts = count_partitions($self);
219 if ($nparts && $nparts != $self->{partitions}) {
220 $self->{partitions} = $nparts;
223 # need to create all parts before initializing msgmap FD
224 my $max = $self->{partitions} - 1;
226 # idx_parts must be visible to all forked processes
227 my $idx = $self->{idx_parts} = [];
228 for my $i (0..$max) {
229 push @$idx, PublicInbox::SearchIdxPart->new($self, $i);
232 # Now that all subprocesses are up, we can open the FDs
234 my $mm = $self->{mm} = PublicInbox::Msgmap->new_file(
235 "$self->{-inbox}->{mainrepo}/msgmap.sqlite3", 1);
236 $mm->{dbh}->begin_work;
241 my ($self, $purge) = @_; # $purge = { $object_id => 1, ... }
243 my $pfx = "$self->{-inbox}->{mainrepo}/git";
245 foreach my $i (0..$self->{max_git}) {
246 my $git = PublicInbox::Git->new("$pfx/$i.git");
247 my $im = $self->import_init($git, 0, 1);
248 $purges->[$i] = $im->purge_oids($purge);
253 sub remove_internal {
254 my ($self, $mime, $cmt_msg, $purge) = @_;
257 my $im = $self->importer unless $purge;
258 my $ibx = $self->{-inbox};
259 my $srch = $ibx->search;
260 my $cid = content_id($mime);
261 my $parts = $self->{idx_parts};
262 my $mm = $self->{mm};
264 my $mids = mids($mime->header_obj);
266 # We avoid introducing new blobs into git since the raw content
267 # can be slightly different, so we do not need the user-supplied
268 # message now that we have the mids and content_id
272 foreach my $mid (@$mids) {
274 $srch->reopen->each_smsg_by_mid($mid, sub {
277 my $msg = $ibx->msg_by_smsg($smsg);
278 if (!defined($msg)) {
279 warn "broken smsg for $mid\n";
283 my $cur = PublicInbox::MIME->new($msg);
284 if (content_id($cur) eq $cid) {
285 $smsg->{mime} = $cur;
286 $gone{$smsg->num} = [ $smsg, \$orig ];
290 my $n = scalar keys %gone;
293 warn "BUG: multiple articles linked to <$mid>\n",
294 join(',', sort keys %gone), "\n";
296 foreach my $num (keys %gone) {
297 my ($smsg, $orig) = @{$gone{$num}};
298 $mm->num_delete($num);
299 # $removed should only be set once assuming
300 # no bugs in our deduplication code:
302 my $oid = $smsg->{blob};
306 ($mark, undef) = $im->remove($orig, $cmt_msg);
309 foreach my $idx (@$parts) {
310 $idx->remote_remove($oid, $mid);
312 $self->{over}->remove_oid($oid, $mid);
318 my $cmt = $im->get_mark($mark);
319 $self->{last_commit}->[$self->{max_git}] = $cmt;
321 if ($purge && scalar keys %$purge) {
322 return purge_oids($self, $purge);
328 my ($self, $mime, $cmt_msg) = @_;
329 remove_internal($self, $mime, $cmt_msg);
333 my ($self, $mime) = @_;
334 my $purges = remove_internal($self, $mime, undef, {});
335 $self->idx_init if @$purges; # ->done is called on purges
336 for my $i (0..$#$purges) {
337 defined(my $cmt = $purges->[$i]) or next;
338 $self->{last_commit}->[$i] = $cmt;
343 sub set_last_commits ($) {
345 defined(my $max_git = $self->{max_git}) or return;
346 my $mm = $self->{mm};
347 my $last_commit = $self->{last_commit};
348 foreach my $i (0..$max_git) {
349 defined(my $cmt = $last_commit->[$i]) or next;
350 $last_commit->[$i] = undef;
351 $mm->last_commit_n($i, $cmt);
357 my $im = delete $self->{im};
358 $im->done if $im; # PublicInbox::Import::done
360 my $mm = $self->{mm};
361 $mm->{dbh}->commit if $mm;
363 # order matters, we can only close {over} after all partitions
364 # are done because the partitions also write to {over}
365 my $parts = delete $self->{idx_parts};
367 $_->remote_commit for @$parts;
368 $_->remote_close for @$parts;
371 my $over = $self->{over};
372 $over->remote_commit;
376 $mm->{dbh}->begin_work;
377 set_last_commits($self);
382 $self->{transact_bytes} = 0;
383 $self->lock_release if $parts;
388 my $im = $self->{im};
389 $im->checkpoint if $im; # PublicInbox::Import::checkpoint
393 # issue a write barrier to ensure all data is visible to other processes
394 # and read-only ops. Order of data importance is: git > SQLite > Xapian
396 my ($self, $fsync) = @_;
398 if (my $im = $self->{im}) {
401 my $parts = $self->{idx_parts};
403 my $dbh = $self->{mm}->{dbh};
404 $dbh->commit; # SQLite msgmap data is second in importance
406 my $over = $self->{over};
408 # Now deal with Xapian and overview DB
409 $over->barrier_init(scalar(@$parts));
411 # each partition needs to issue a barrier command to over
412 $_->remote_barrier foreach @$parts;
414 $over->barrier_wait; # wait for each Xapian partition
415 $over->commit_fsync if $fsync;
417 # last_commit is special, don't commit these until
418 # remote partitions are done:
420 set_last_commits($self);
425 $self->{transact_bytes} = 0;
429 my ($self, $new) = @_;
430 my $pfx = "$self->{-inbox}->{mainrepo}/git";
431 my $git_dir = "$pfx/$new.git";
432 my @cmd = (qw(git init --bare -q), $git_dir);
433 PublicInbox::Import::run_die(\@cmd);
435 my $all = "$self->{-inbox}->{mainrepo}/all.git";
437 @cmd = (qw(git init --bare -q), $all);
438 PublicInbox::Import::run_die(\@cmd);
439 @cmd = (qw/git config/, "--file=$all/config",
440 'repack.writeBitmaps', 'true');
441 PublicInbox::Import::run_die(\@cmd);
444 @cmd = (qw/git config/, "--file=$git_dir/config",
445 'include.path', '../../all.git/config');
446 PublicInbox::Import::run_die(\@cmd);
448 my $alt = "$all/objects/info/alternates";
449 my $new_obj_dir = "../../git/$new.git/objects";
452 open(my $fh, '<', $alt) or die "open < $alt: $!\n";
453 %alts = map { chomp; $_ => 1 } (<$fh>);
455 return $git_dir if $alts{$new_obj_dir};
456 open my $fh, '>>', $alt or die "open >> $alt: $!\n";
457 print $fh "$new_obj_dir\n" or die "print >> $alt: $!\n";
458 close $fh or die "close $alt: $!\n";
463 my ($self, $max) = @_;
465 my $pfx = "$self->{-inbox}->{mainrepo}/git";
466 return unless -d $pfx;
468 opendir my $dh, $pfx or die "opendir $pfx: $!\n";
469 while (defined(my $git_dir = readdir($dh))) {
470 $git_dir =~ m!\A(\d+)\.git\z! or next;
473 $latest = "$pfx/$git_dir";
481 my $im = $self->{im};
483 if ($im->{bytes_added} < $self->{rotate_bytes}) {
490 my $git_dir = $self->git_init(++$self->{max_git});
491 my $git = PublicInbox::Git->new($git_dir);
492 return $self->import_init($git, 0);
497 my $latest = git_dir_latest($self, \$max);
498 if (defined $latest) {
499 my $git = PublicInbox::Git->new($latest);
500 my $packed_bytes = $git->packed_bytes;
501 if ($packed_bytes >= $self->{rotate_bytes}) {
504 $self->{max_git} = $max;
505 return $self->import_init($git, $packed_bytes);
508 $self->{max_git} = $new;
509 $latest = $self->git_init($new);
510 $self->import_init(PublicInbox::Git->new($latest), 0);
514 my ($self, $git, $packed_bytes, $tmp) = @_;
515 my $im = PublicInbox::Import->new($git, undef, undef, $self->{-inbox});
516 $im->{bytes_added} = int($packed_bytes / $PACKING_FACTOR);
517 $im->{want_object_info} = 1;
518 $im->{lock_path} = undef;
519 $im->{path_type} = 'v2';
520 $self->{im} = $im unless $tmp;
526 my ($mid, $cur, $new) = @_;
527 use File::Temp qw(tempfile);
528 use PublicInbox::Spawn qw(spawn);
530 my ($ah, $an) = tempfile('email-cur-XXXXXXXX', TMPDIR => 1);
531 print $ah $cur->as_string or die "print: $!";
532 close $ah or die "close: $!";
533 my ($bh, $bn) = tempfile('email-new-XXXXXXXX', TMPDIR => 1);
534 PublicInbox::Import::drop_unwanted_headers($new);
535 print $bh $new->as_string or die "print: $!";
536 close $bh or die "close: $!";
537 my $cmd = [ qw(diff -u), $an, $bn ];
538 print STDERR "# MID conflict <$mid>\n";
539 my $pid = spawn($cmd, undef, { 1 => 2 });
540 defined $pid or die "diff failed to spawn $!";
541 waitpid($pid, 0) == $pid or die "diff did not finish";
546 my ($self, $mime, $mid) = @_;
547 my $ibx = $self->{-inbox};
549 my $srch = $ibx->search->reopen;
550 my $cid = content_id($mime);
552 $srch->each_smsg_by_mid($mid, sub {
555 my $msg = $ibx->msg_by_smsg($smsg);
556 if (!defined($msg)) {
557 warn "broken smsg for $mid\n";
560 my $cur = PublicInbox::MIME->new($msg);
561 if (content_id($cur) eq $cid) {
562 $smsg->{mime} = $cur;
564 return 0; # break out of loop
567 # XXX DEBUG_DIFF is experimental and may be removed
568 diff($mid, $cur, $mime) if $ENV{DEBUG_DIFF};
577 my $fh = delete $self->{reindex_pipe};
579 if (my $parts = $self->{idx_parts}) {
580 $_->atfork_child foreach @$parts;
582 if (my $im = $self->{im}) {
585 die "unexpected mm" if $self->{mm};
589 my ($self, $D, $git, $oid) = @_;
590 my $msgref = $git->cat_file($oid);
591 my $mime = PublicInbox::MIME->new($$msgref);
592 my $mids = mids($mime->header_obj);
593 my $cid = content_id($mime);
594 foreach my $mid (@$mids) {
595 $D->{"$mid\0$cid"} = 1;
600 my ($self, $mm_tmp, $D, $git, $oid, $regen) = @_;
602 my $msgref = $git->cat_file($oid, \$len);
603 my $mime = PublicInbox::MIME->new($$msgref);
604 my $mids = mids($mime->header_obj);
605 my $cid = content_id($mime);
607 # get the NNTP article number we used before, highest number wins
608 # and gets deleted from mm_tmp;
612 foreach my $mid (@$mids) {
613 $del += (delete $D->{"$mid\0$cid"} || 0);
614 my $n = $mm_tmp->num_for($mid);
615 if (defined $n && $n > $num) {
620 if (!defined($mid0) && $regen && !$del) {
622 die "BUG: ran out of article numbers\n" if $num <= 0;
623 my $mm = $self->{mm};
624 foreach my $mid (reverse @$mids) {
625 if ($mm->mid_set($num, $mid) == 1) {
630 if (!defined($mid0)) {
631 my $id = '<' . join('> <', @$mids) . '>';
632 warn "Message-ID $id unusable for $num\n";
633 foreach my $mid (@$mids) {
634 defined(my $n = $mm->num_for($mid)) or next;
635 warn "#$n previously mapped for <$mid>\n";
640 if (!defined($mid0) || $del) {
641 if (!defined($mid0) && $del) { # expected for deletes
646 my $id = '<' . join('> <', @$mids) . '>';
648 warn "Skipping $id, no article number found\n";
649 if ($del && defined($mid0)) {
650 warn "$id was deleted $del " .
651 "time(s) but mapped to article #$num\n";
656 $mm_tmp->mid_delete($mid0) or
657 die "failed to delete <$mid0> for article #$num\n";
659 my $nparts = $self->{partitions};
660 my $part = $num % $nparts;
661 my $idx = $self->idx_part($part);
662 $idx->index_raw($len, $msgref, $num, $oid, $mid0, $mime);
663 my $n = $self->{transact_bytes} += $len;
664 if ($n > (PublicInbox::SearchIdx::BATCH_BYTES * $nparts)) {
666 $mm_tmp->atfork_prepare;
667 $self->done; # release lock
668 # allow -watch or -mda to write...
669 $self->idx_init; # reacquire lock
670 $mm_tmp->atfork_parent;
674 # only update last_commit for $i on reindex iff newer than current
675 sub update_last_commit {
676 my ($self, $git, $i, $cmt) = @_;
677 my $last = $self->{mm}->last_commit_n($i);
678 if (defined $last && is_ancestor($git, $last, $cmt)) {
679 my @cmd = (qw(rev-list --count), "$last..$cmt");
680 chomp(my $n = $git->qx(@cmd));
681 return if $n ne '' && $n == 0;
683 $self->{mm}->last_commit_n($i, $cmt);
686 sub git_dir_n ($$) { "$_[0]->{-inbox}->{mainrepo}/git/$_[1].git" }
689 my ($self, $max_git) = @_;
691 for (my $i = $max_git; $i >= 0; $i--) {
692 $heads->[$i] = $self->{mm}->last_commit_n($i);
697 sub is_ancestor ($$$) {
698 my ($git, $cur, $tip) = @_;
699 return 0 unless $git->check($cur);
700 my $cmd = [ 'git', "--git-dir=$git->{git_dir}",
701 qw(merge-base --is-ancestor), $cur, $tip ];
702 my $pid = spawn($cmd);
703 defined $pid or die "spawning ".join(' ', @$cmd)." failed: $!";
704 waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
709 my ($self, $opts, $max_git, $ranges) = @_;
711 my $head = $self->{-inbox}->{ref_head} || 'refs/heads/master';
712 for (my $i = $max_git; $i >= 0; $i--) {
713 die "already indexing!\n" if $self->{index_pipe};
714 my $git_dir = git_dir_n($self, $i);
715 -d $git_dir or next; # missing parts are fine
716 my $git = PublicInbox::Git->new($git_dir);
717 chomp(my $tip = $git->qx('rev-parse', $head));
719 if (defined(my $cur = $ranges->[$i])) {
720 $range = "$cur..$tip";
721 if (is_ancestor($git, $cur, $tip)) { # common case
722 my $n = $git->qx(qw(rev-list --count), $range);
725 $ranges->[$i] = undef;
730 discontiguous range: $range
731 Rewritten history? (in $git_dir)
733 my $base = $git->qx('merge-base', $tip, $cur);
736 $range = "$base..$tip";
737 warn "found merge-base: $base\n"
741 discarding history at $cur
745 reindexing $git_dir starting at
748 $self->{"unindex-range.$i"} = "$base..$cur";
751 $range = $tip; # all of it
753 $ranges->[$i] = $range;
755 # can't use 'rev-list --count' if we use --diff-filter
756 my $fh = $git->popen(qw(log --pretty=tformat:%h
757 --no-notes --no-color --no-renames
758 --diff-filter=AM), $range, '--', 'm');
759 ++$regen_max while <$fh>;
765 my ($self, $git, $oid) = @_;
766 my $msgref = $git->cat_file($oid);
767 my $mime = PublicInbox::MIME->new($msgref);
768 my $mids = mids($mime->header_obj);
769 $mime = $msgref = undef;
771 foreach my $mid (@$mids) {
773 $self->{-inbox}->search->reopen->each_smsg_by_mid($mid, sub {
776 $gone{$smsg->num} = 1 if $oid eq $smsg->{blob};
779 my $n = scalar keys %gone;
782 warn "BUG: multiple articles linked to $oid\n",
783 join(',',sort keys %gone), "\n";
785 $self->{unindexed}->{$_}++ foreach keys %gone;
786 $_->remote_remove($oid, $mid) foreach @{$self->{idx_parts}};
787 $self->{over}->remove_oid($oid, $mid);
792 my $x40 = qr/[a-f0-9]{40}/;
794 my ($self, $opts, $git, $unindex_range) = @_;
795 my $un = $self->{unindexed} ||= {}; # num => removal count
797 my $before = scalar keys %$un;
798 my @cmd = qw(log --raw -r
799 --no-notes --no-color --no-abbrev --no-renames);
800 my $fh = $self->{reindex_pipe} = $git->popen(@cmd, $unindex_range);
802 /\A:\d{6} 100644 $x40 ($x40) [AM]\tm$/o or next;
803 $self->unindex_oid($git, $1);
805 delete $self->{reindex_pipe};
808 return unless $opts->{prune};
809 my $after = scalar keys %$un;
810 return if $before == $after;
812 # ensure any blob can not longer be accessed via dumb HTTP
813 PublicInbox::Import::run_die(['git', "--git-dir=$git->{git_dir}",
814 qw(-c gc.reflogExpire=now gc --prune=all)]);
818 my ($self, $opts) = @_;
820 my $ibx = $self->{-inbox};
822 my $latest = git_dir_latest($self, \$max_git);
823 return unless defined $latest;
824 $self->idx_init; # acquire lock
825 my $mm_tmp = $self->{mm}->tmp_clone;
826 my $ranges = $opts->{reindex} ? [] : $self->last_commits($max_git);
828 my ($min, $max) = $mm_tmp->minmax;
829 my $regen = $self->index_prepare($opts, $max_git, $ranges);
830 $$regen += $max if $max;
832 my @cmd = qw(log --raw -r --pretty=tformat:%h
833 --no-notes --no-color --no-abbrev --no-renames);
835 # work backwards through history
836 my $last_commit = [];
837 for (my $i = $max_git; $i >= 0; $i--) {
838 my $git_dir = git_dir_n($self, $i);
839 die "already reindexing!\n" if delete $self->{reindex_pipe};
840 -d $git_dir or next; # missing parts are fine
841 my $git = PublicInbox::Git->new($git_dir);
842 my $unindex = delete $self->{"unindex-range.$i"};
843 $self->unindex($opts, $git, $unindex) if $unindex;
844 defined(my $range = $ranges->[$i]) or next;
845 my $fh = $self->{reindex_pipe} = $git->popen(@cmd, $range);
848 if (/\A$x40$/o && !defined($cmt)) {
850 } elsif (/\A:\d{6} 100644 $x40 ($x40) [AM]\tm$/o) {
851 $self->reindex_oid($mm_tmp, $D, $git, $1,
853 } elsif (/\A:\d{6} 100644 $x40 ($x40) [AM]\td$/o) {
854 $self->mark_deleted($D, $git, $1);
858 delete $self->{reindex_pipe};
859 $self->update_last_commit($git, $i, $cmt) if defined $cmt;
861 my @d = sort keys %$D;
863 warn "BUG: ", scalar(@d)," unseen deleted messages marked\n";
865 my ($mid, undef) = split(/\0/, $_, 2);