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($$mid0);
176 my @cur = $hdr->header_raw('Message-Id');
177 $hdr->header_set('Message-Id', "<$$mid0>", @cur);
182 my ($self, $part) = @_;
183 $self->{idx_parts}->[$part];
189 return if $self->{idx_parts};
190 my $ibx = $self->{-inbox};
192 # do not leak read-only FDs to child processes, we only have these
193 # FDs for duplicate detection so they should not be
194 # frequently activated.
195 delete $ibx->{$_} foreach (qw(git mm search));
199 # first time initialization, first we create the skeleton pipe:
200 my $skel = $self->{skel} = PublicInbox::SearchIdxSkeleton->new($self);
202 # need to create all parts before initializing msgmap FD
203 my $max = $self->{partitions} - 1;
204 my $idx = $self->{idx_parts} = [];
205 for my $i (0..$max) {
206 push @$idx, PublicInbox::SearchIdxPart->new($self, $i, $skel);
209 # Now that all subprocesses are up, we can open the FD for SQLite:
210 $skel->_msgmap_init->{dbh}->begin_work;
214 my ($self, $mime, $cmt_msg) = @_;
217 my $im = $self->importer;
218 my $ibx = $self->{-inbox};
219 my $srch = $ibx->search;
220 my $cid = content_id($mime);
221 my $skel = $self->{skel};
222 my $parts = $self->{idx_parts};
223 my $mm = $skel->{mm};
225 my $mids = mids($mime->header_obj);
227 # We avoid introducing new blobs into git since the raw content
228 # can be slightly different, so we do not need the user-supplied
229 # message now that we have the mids and content_id
232 foreach my $mid (@$mids) {
233 $srch->reopen->each_smsg_by_mid($mid, sub {
236 my $msg = $ibx->msg_by_smsg($smsg);
237 if (!defined($msg)) {
238 warn "broken smsg for $mid\n";
242 my $cur = PublicInbox::MIME->new($msg);
243 if (content_id($cur) eq $cid) {
244 $mm->num_delete($smsg->num);
245 # $removed should only be set once assuming
246 # no bugs in our deduplication code:
248 $removed->{mime} = $cur;
249 $im->remove(\$orig, $cmt_msg);
251 $removed->num; # memoize this for callers
253 my $oid = $smsg->{blob};
254 foreach my $idx (@$parts, $skel) {
255 $idx->remote_remove($oid, $mid);
267 my $locked = defined $self->{idx_parts};
268 my $im = delete $self->{im};
269 $im->done if $im; # PublicInbox::Import::done
270 $self->searchidx_checkpoint(0);
271 $self->lock_release if $locked;
276 my $im = $self->{im};
277 $im->checkpoint if $im; # PublicInbox::Import::checkpoint
278 $self->searchidx_checkpoint(1);
281 # issue a write barrier to ensure all data is visible to other processes
282 # and read-only ops. Order of data importance is: git > SQLite > Xapian
286 if (my $im = $self->{im}) {
289 my $skel = $self->{skel};
290 my $parts = $self->{idx_parts};
291 if ($parts && $skel) {
292 my $dbh = $skel->{mm}->{dbh};
293 $dbh->commit; # SQLite data is second in importance
295 # Now deal with Xapian
296 $skel->barrier_init(scalar(@$parts));
297 # each partition needs to issue a barrier command to skel:
298 $_->remote_barrier foreach @$parts;
300 $skel->barrier_wait; # wait for each Xapian partition
304 $self->{transact_bytes} = 0;
307 sub searchidx_checkpoint {
308 my ($self, $more) = @_;
310 # order matters, we can only close {skel} after all partitions
311 # are done because the partitions also write to {skel}
312 if (my $parts = $self->{idx_parts}) {
313 foreach my $idx (@$parts) {
314 $idx->remote_commit; # propagates commit to skel
315 $idx->remote_close unless $more;
317 delete $self->{idx_parts} unless $more;
320 if (my $skel = $self->{skel}) {
321 my $dbh = $skel->{mm}->{dbh};
327 delete $self->{skel};
330 $self->{transact_bytes} = 0;
334 my ($self, $new) = @_;
335 my $pfx = "$self->{-inbox}->{mainrepo}/git";
336 my $git_dir = "$pfx/$new.git";
337 die "$git_dir exists\n" if -e $git_dir;
338 my @cmd = (qw(git init --bare -q), $git_dir);
339 PublicInbox::Import::run_die(\@cmd);
341 my $all = "$self->{-inbox}->{mainrepo}/all.git";
343 @cmd = (qw(git init --bare -q), $all);
344 PublicInbox::Import::run_die(\@cmd);
345 @cmd = (qw/git config/, "--file=$all/config",
346 'repack.writeBitmaps', 'true');
347 PublicInbox::Import::run_die(\@cmd);
350 @cmd = (qw/git config/, "--file=$git_dir/config",
351 'include.path', '../../all.git/config');
352 PublicInbox::Import::run_die(\@cmd);
354 my $alt = "$all/objects/info/alternates";
355 my $new_obj_dir = "../../git/$new.git/objects";
358 open(my $fh, '<', $alt) or die "open < $alt: $!\n";
359 %alts = map { chomp; $_ => 1 } (<$fh>);
361 return $git_dir if $alts{$new_obj_dir};
362 open my $fh, '>>', $alt or die "open >> $alt: $!\n";
363 print $fh "$new_obj_dir\n" or die "print >> $alt: $!\n";
364 close $fh or die "close $alt: $!\n";
369 my ($self, $max) = @_;
371 my $pfx = "$self->{-inbox}->{mainrepo}/git";
372 return unless -d $pfx;
374 opendir my $dh, $pfx or die "opendir $pfx: $!\n";
375 while (defined(my $git_dir = readdir($dh))) {
376 $git_dir =~ m!\A(\d+)\.git\z! or next;
379 $latest = "$pfx/$git_dir";
387 my $im = $self->{im};
389 if ($im->{bytes_added} < $self->{rotate_bytes}) {
394 $self->searchidx_checkpoint(1);
396 my $git_dir = $self->git_init(++$self->{max_git});
397 my $git = PublicInbox::Git->new($git_dir);
398 return $self->import_init($git, 0);
403 my $latest = git_dir_latest($self, \$max);
404 if (defined $latest) {
405 my $git = PublicInbox::Git->new($latest);
406 my $packed_bytes = $git->packed_bytes;
407 if ($packed_bytes >= $self->{rotate_bytes}) {
410 $self->{max_git} = $max;
411 return $self->import_init($git, $packed_bytes);
414 $self->{max_git} = $new;
415 $latest = $self->git_init($new);
416 $self->import_init(PublicInbox::Git->new($latest), 0);
420 my ($self, $git, $packed_bytes) = @_;
421 my $im = PublicInbox::Import->new($git, undef, undef, $self->{-inbox});
422 $im->{bytes_added} = int($packed_bytes / $PACKING_FACTOR);
423 $im->{want_object_info} = 1;
424 $im->{lock_path} = undef;
425 $im->{path_type} = 'v2';
431 my ($mid, $cur, $new) = @_;
432 use File::Temp qw(tempfile);
433 use PublicInbox::Spawn qw(spawn);
435 my ($ah, $an) = tempfile('email-cur-XXXXXXXX');
436 print $ah $cur->as_string or die "print: $!";
437 close $ah or die "close: $!";
438 my ($bh, $bn) = tempfile('email-new-XXXXXXXX');
439 PublicInbox::Import::drop_unwanted_headers($new);
440 print $bh $new->as_string or die "print: $!";
441 close $bh or die "close: $!";
442 my $cmd = [ qw(diff -u), $an, $bn ];
443 print STDERR "# MID conflict <$mid>\n";
444 my $pid = spawn($cmd, undef, { 1 => 2 });
445 defined $pid or die "diff failed to spawn $!";
446 waitpid($pid, 0) == $pid or die "diff did not finish";
451 my ($self, $mime, $mid) = @_;
452 my $ibx = $self->{-inbox};
454 my $srch = $ibx->search->reopen;
455 my $cid = content_id($mime);
457 $srch->each_smsg_by_mid($mid, sub {
460 my $msg = $ibx->msg_by_smsg($smsg);
461 if (!defined($msg)) {
462 warn "broken smsg for $mid\n";
465 my $cur = PublicInbox::MIME->new($msg);
466 if (content_id($cur) eq $cid) {
467 $smsg->{mime} = $cur;
469 return 0; # break out of loop
472 # XXX DEBUG_DIFF is experimental and may be removed
473 diff($mid, $cur, $mime) if $ENV{DEBUG_DIFF};
482 my $fh = delete $self->{reindex_pipe};
484 if (my $parts = $self->{idx_parts}) {
485 $_->atfork_child foreach @$parts;
487 if (my $im = $self->{im}) {
493 my ($self, $D, $git, $oid) = @_;
494 my $msgref = $git->cat_file($oid);
495 my $mime = PublicInbox::MIME->new($$msgref);
496 my $mids = mids($mime->header_obj);
497 my $cid = content_id($mime);
498 foreach my $mid (@$mids) {
504 my ($self, $mm_tmp, $D, $git, $oid, $regen) = @_;
506 my $msgref = $git->cat_file($oid, \$len);
507 my $mime = PublicInbox::MIME->new($$msgref);
508 my $mids = mids($mime->header_obj);
509 my $cid = content_id($mime);
511 # get the NNTP article number we used before, highest number wins
512 # and gets deleted from mm_tmp;
516 foreach my $mid (@$mids) {
517 $del += (delete $D->{$mid.$cid} || 0);
518 my $n = $mm_tmp->num_for($mid);
519 if (defined $n && $n > $num) {
524 if (!defined($mid0) && $regen && !$del) {
526 die "BUG: ran out of article numbers\n" if $num <= 0;
527 my $mm = $self->{skel}->{mm};
528 foreach my $mid (@$mids) {
529 if ($mm->mid_set($num, $mid) == 1) {
534 if (!defined($mid0)) {
535 my $id = '<' . join('> <', @$mids) . '>';
536 warn "Message-Id $id unusable for $num\n";
540 if (!defined($mid0) || $del) {
541 if (!defined($mid0) && $del) { # expected for deletes
546 my $id = '<' . join('> <', @$mids) . '>';
548 warn "Skipping $id, no article number found\n";
549 if ($del && defined($mid0)) {
550 warn "$id was deleted $del " .
551 "time(s) but mapped to article #$num\n";
556 $mm_tmp->mid_delete($mid0) or
557 die "failed to delete <$mid0> for article #$num\n";
559 my $nparts = $self->{partitions};
560 my $part = $num % $nparts;
561 my $idx = $self->idx_part($part);
562 $idx->index_raw($len, $msgref, $num, $oid, $mid0, $mime);
563 my $n = $self->{transact_bytes} += $len;
564 if ($n > (PublicInbox::SearchIdx::BATCH_BYTES * $nparts)) {
566 $mm_tmp->atfork_prepare;
567 $self->done; # release lock
568 # allow -watch or -mda to write...
569 $self->idx_init; # reacquire lock
570 $mm_tmp->atfork_parent;
575 my ($self, $regen) = @_;
576 my $ibx = $self->{-inbox};
577 my $pfx = "$ibx->{mainrepo}/git";
579 my $latest = git_dir_latest($self, \$max_git);
580 return unless defined $latest;
581 my $head = $ibx->{ref_head} || 'refs/heads/master';
582 $self->idx_init; # acquire lock
583 my $x40 = qr/[a-f0-9]{40}/;
584 my $mm_tmp = $self->{skel}->{mm}->tmp_clone;
586 my (undef, $max) = $mm_tmp->minmax;
587 unless (defined $max) {
590 "empty msgmap.sqlite3, regenerating article numbers\n";
593 my $tip; # latest commit out of all git repos
596 for (my $cur = $max_git; $cur >= 0; $cur--) {
597 die "already reindexing!\n" if $self->{reindex_pipe};
598 my $git = PublicInbox::Git->new("$pfx/$cur.git");
599 chomp($tip = $git->qx('rev-parse', $head)) unless $tip;
600 my $h = $cur == $max_git ? $tip : $head;
601 my @count = ('rev-list', '--count', $h, '--', 'm');
602 $regen_max += $git->qx(@count);
604 die "No messages found in $pfx/*.git, bug?\n" unless $regen_max;
605 $regen = \$regen_max;
608 my @cmd = qw(log --raw -r --pretty=tformat:%h
609 --no-notes --no-color --no-abbrev);
611 # if we are regenerating, we must not use a newer tip commit than what
612 # the regeneration counter used:
615 # work backwards through history
616 for (my $cur = $max_git; $cur >= 0; $cur--) {
617 die "already reindexing!\n" if delete $self->{reindex_pipe};
619 my $git_dir = "$pfx/$cur.git";
620 my $git = PublicInbox::Git->new($git_dir);
621 my $h = $cur == $max_git ? $tip : $head;
622 my $fh = $self->{reindex_pipe} = $git->popen(@cmd, $h);
626 } elsif (/\A:\d{6} 100644 $x40 ($x40) [AM]\tm$/o) {
627 $self->reindex_oid($mm_tmp, $D, $git, $1,
629 } elsif (m!\A:\d{6} 100644 $x40 ($x40) [AM]\t_/D$!o) {
630 $self->mark_deleted($D, $git, $1);
633 delete $self->{reindex_pipe};
635 my ($min, $max) = $mm_tmp->minmax;
636 defined $max and die "leftover article numbers at $min..$max\n";