From: Eric Wong <e@80x24.org>
To: meta@public-inbox.org
Subject: [PATCH 3/3] v2writable: reindex handles 3-headered monsters
Date: Mon, 21 Oct 2019 11:02:21 +0000 [thread overview]
Message-ID: <20191021110221.23753-4-e@80x24.org> (raw)
In-Reply-To: <20191021110221.23753-1-e@80x24.org>
And maybe 8-headered ones, too...
I noticed --reindex failing on the linux-renesas-soc mirror due
one 3-headed monster of a message having 3 sets of headers;
while another normal message had a Message-ID that matched one
of the 3 IDs of the 3-headed monster.
We still try to do the majority of indexing backwards, but we
defer indexing multi-Message-ID'd messages until the end to
ensure we get all the "good" messages in before we process the
multi-headered ones.
Link: https://public-inbox.org/meta/20191016211415.GA6084@dcvr/
---
TODO | 3 +
lib/PublicInbox/OverIdx.pm | 14 +++
lib/PublicInbox/V2Writable.pm | 213 ++++++++++++++++++++++++----------
t/v2reindex.t | 66 +++++++++++
4 files changed, 236 insertions(+), 60 deletions(-)
diff --git a/TODO b/TODO
index a327ca06..61c44a84 100644
--- a/TODO
+++ b/TODO
@@ -133,3 +133,6 @@ all need to be considered for everything we introduce)
for coderepos
* configurable diff output for solver-generated blobs
+
+* fix search for messages with multiple Subject:/To:/From:/Date:
+ headers (some wacky examples out there...)
diff --git a/lib/PublicInbox/OverIdx.pm b/lib/PublicInbox/OverIdx.pm
index 7fd1905d..64277342 100644
--- a/lib/PublicInbox/OverIdx.pm
+++ b/lib/PublicInbox/OverIdx.pm
@@ -343,6 +343,20 @@ sub remove_oid {
$nr;
}
+sub num_mid0_for_oid {
+ my ($self, $oid, $mid) = @_;
+ my ($num, $mid0);
+ $self->begin_lazy;
+ each_by_mid($self, $mid, ['ddd'], sub {
+ my ($smsg) = @_;
+ my $blob = $smsg->{blob};
+ return 1 if (!defined($blob) || $blob ne $oid); # continue;
+ ($num, $mid0) = ($smsg->{num}, $smsg->{mid});
+ 0; # done
+ });
+ ($num, $mid0);
+}
+
sub create_tables {
my ($dbh) = @_;
diff --git a/lib/PublicInbox/V2Writable.pm b/lib/PublicInbox/V2Writable.pm
index 9d507828..7ece6b01 100644
--- a/lib/PublicInbox/V2Writable.pm
+++ b/lib/PublicInbox/V2Writable.pm
@@ -841,81 +841,157 @@ sub mark_deleted ($$$$) {
}
}
-sub reindex_oid ($$$$) {
+sub reindex_checkpoint ($$$) {
+ my ($self, $sync, $git) = @_;
+
+ $git->cleanup;
+ $sync->{mm_tmp}->atfork_prepare;
+ $self->done; # release lock
+
+ if (my $pr = $sync->{-opt}->{-progress}) {
+ my ($bn) = (split('/', $git->{git_dir}))[-1];
+ $pr->("$bn ".sprintf($sync->{-regen_fmt}, $sync->{nr}));
+ }
+
+ # allow -watch or -mda to write...
+ $self->idx_init; # reacquire lock
+ $sync->{mm_tmp}->atfork_parent;
+}
+
+# only for a few odd messages with multiple Message-IDs
+sub reindex_oid_m ($$$$) {
my ($self, $sync, $git, $oid) = @_;
- my $len;
+ my ($num, $mid0, $len);
my $msgref = $git->cat_file($oid, \$len);
my $mime = PublicInbox::MIME->new($$msgref);
my $mids = mids($mime->header_obj);
my $cid = content_id($mime);
+ die "BUG: reindex_oid_m called for <=1 mids" if scalar(@$mids) <= 1;
- # get the NNTP article number we used before, highest number wins
- # and gets deleted from sync->{mm_tmp};
- my $mid0;
- my $num = -1;
- my $del = 0;
- foreach my $mid (@$mids) {
- $del += delete($sync->{D}->{"$mid\0$cid"}) ? 1 : 0;
- my $n = $sync->{mm_tmp}->num_for($mid);
- if (defined $n && $n > $num) {
+ for my $mid (reverse @$mids) {
+ delete($sync->{D}->{"$mid\0$cid"}) and
+ die "BUG: reindex_oid should handle <$mid> delete";
+ }
+ for my $mid (reverse @$mids) {
+ ($num, $mid0) = $self->{over}->num_mid0_for_oid($oid, $mid);
+ last if defined $num;
+ }
+ unless (defined($num)) {
+ for my $mid (reverse @$mids) {
+ # is this a number we got before?
+ $num = $sync->{mm_tmp}->num_for($mid);
+ next unless defined $num;
$mid0 = $mid;
- $num = $n;
- $self->{mm}->mid_set($num, $mid0);
+ last;
}
}
- if (!defined($mid0) && !$del) {
+ if (defined($num)) {
+ $sync->{mm_tmp}->num_delete($num);
+ } else {
$num = $sync->{regen}--;
- die "BUG: ran out of article numbers\n" if $num <= 0;
- my $mm = $self->{mm};
- foreach my $mid (reverse @$mids) {
- if ($mm->mid_set($num, $mid) == 1) {
+ if ($num <= 0) {
+ # fixup bugs in old mirrors on reindex
+ for my $mid (reverse @$mids) {
+ $num = $self->{mm}->mid_insert($mid);
+ next unless defined $num;
$mid0 = $mid;
last;
}
- }
- if (!defined($mid0)) {
- my $id = '<' . join('> <', @$mids) . '>';
- warn "Message-ID $id unusable for $num\n";
- foreach my $mid (@$mids) {
- defined(my $n = $mm->num_for($mid)) or next;
- warn "#$n previously mapped for <$mid>\n";
+ if (defined $mid0) {
+ if ($sync->{reindex}) {
+ warn "reindex added #$num <$mid0>\n";
+ }
+ } else {
+ warn "E: cannot find article #\n";
+ return;
+ }
+ } else { # $num > 0, use the new article number
+ for my $mid (reverse @$mids) {
+ $self->{mm}->mid_set($num, $mid) == 1 or next;
+ $mid0 = $mid;
+ last;
+ }
+ unless (defined $mid0) {
+ warn "E: cannot regen #$num\n";
+ return;
}
}
}
+ $sync->{nr}++;
+ if (do_idx($self, $msgref, $mime, $len, $num, $oid, $mid0)) {
+ reindex_checkpoint($self, $sync, $git);
+ }
+}
- if (!defined($mid0) || $del) {
- if (!defined($mid0) && $del) { # expected for deletes
- $num = $sync->{regen}--;
- $self->{mm}->num_highwater($num) if !$sync->{reindex};
- return
+sub check_unindexed ($$$) {
+ my ($self, $num, $mid0) = @_;
+ my $unindexed = $self->{unindexed} // {};
+ my $n = delete($unindexed->{$mid0});
+ defined $n or return;
+ if ($n != $num) {
+ die "BUG: unindexed $n != $num <$mid0>\n";
+ } else {
+ $self->{mm}->mid_set($num, $mid0);
+ }
+}
+
+sub reindex_oid ($$$$) {
+ my ($self, $sync, $git, $oid) = @_;
+ my ($num, $mid0, $len);
+ my $msgref = $git->cat_file($oid, \$len);
+ return if $len == 0; # purged
+ my $mime = PublicInbox::MIME->new($$msgref);
+ my $mids = mids($mime->header_obj);
+ my $cid = content_id($mime);
+
+ if (scalar(@$mids) == 0) {
+ warn "E: $oid has no Message-ID, skipping\n";
+ return;
+ } elsif (scalar(@$mids) == 1) {
+ my $mid = $mids->[0];
+
+ # was the file previously marked as deleted?, skip if so
+ if (delete($sync->{D}->{"$mid\0$cid"})) {
+ if (!$sync->{reindex}) {
+ $num = $sync->{regen}--;
+ $self->{mm}->num_highwater($num);
+ }
+ return;
}
- my $id = '<' . join('> <', @$mids) . '>';
- defined($mid0) or
- warn "Skipping $id, no article number found\n";
- if ($del && defined($mid0)) {
- warn "$id was deleted $del " .
- "time(s) but mapped to article #$num\n";
+ # is this a number we got before?
+ $num = $sync->{mm_tmp}->num_for($mid);
+ if (defined $num) {
+ $mid0 = $mid;
+ check_unindexed($self, $num, $mid0);
+ } else {
+ $num = $sync->{regen}--;
+ die "BUG: ran out of article numbers\n" if $num <= 0;
+ if ($self->{mm}->mid_set($num, $mid) != 1) {
+ warn "E: unable to assign $num => <$mid>\n";
+ return;
+ }
+ $mid0 = $mid;
+ }
+ } else { # multiple MIDs are a weird case:
+ my $del = 0;
+ for (@$mids) {
+ $del += delete($sync->{D}->{"$_\0$cid"}) // 0;
+ }
+ if ($del) {
+ unindex_oid_remote($self, $oid, $_) for @$mids;
+ # do not delete from {mm_tmp}, since another
+ # single-MID message may use it.
+ } else { # handle them at the end:
+ push @{$sync->{multi_mid} //= []}, $oid;
}
return;
-
}
$sync->{mm_tmp}->mid_delete($mid0) or
die "failed to delete <$mid0> for article #$num\n";
$sync->{nr}++;
if (do_idx($self, $msgref, $mime, $len, $num, $oid, $mid0)) {
- $git->cleanup;
- $sync->{mm_tmp}->atfork_prepare;
- $self->done; # release lock
-
- if (my $pr = $sync->{-opt}->{-progress}) {
- my ($bn) = (split('/', $git->{git_dir}))[-1];
- $pr->("$bn ".sprintf($sync->{-regen_fmt}, $sync->{nr}));
- }
-
- # allow -watch or -mda to write...
- $self->idx_init; # reacquire lock
- $sync->{mm_tmp}->atfork_parent;
+ reindex_checkpoint($self, $sync, $git);
}
}
@@ -1047,8 +1123,9 @@ sub unindex_oid_remote ($$$) {
$self->{over}->remove_oid($oid, $mid);
}
-sub unindex_oid ($$$) {
- my ($self, $git, $oid) = @_;
+sub unindex_oid ($$$;$) {
+ my ($self, $git, $oid, $unindexed) = @_;
+ my $mm = $self->{mm};
my $msgref = $git->cat_file($oid);
my $mime = PublicInbox::MIME->new($msgref);
my $mids = mids($mime->header_obj);
@@ -1068,8 +1145,11 @@ sub unindex_oid ($$$) {
join(',',sort keys %gone), "\n";
}
foreach my $num (keys %gone) {
- $self->{unindexed}->{$num}++;
- $self->{mm}->num_delete($num);
+ if ($unindexed) {
+ my $mid0 = $mm->mid_for($num);
+ $unindexed->{$mid0} = $num;
+ }
+ $mm->num_delete($num);
}
unindex_oid_remote($self, $oid, $mid);
}
@@ -1078,20 +1158,21 @@ sub unindex_oid ($$$) {
my $x40 = qr/[a-f0-9]{40}/;
sub unindex ($$$$) {
my ($self, $sync, $git, $unindex_range) = @_;
- my $un = $self->{unindexed} ||= {}; # num => removal count
- my $before = scalar keys %$un;
+ my $unindexed = $self->{unindexed} ||= {}; # $mid0 => $num
+ my $before = scalar keys %$unindexed;
+ # order does not matter, here:
my @cmd = qw(log --raw -r
--no-notes --no-color --no-abbrev --no-renames);
my $fh = $self->{reindex_pipe} = $git->popen(@cmd, $unindex_range);
while (<$fh>) {
/\A:\d{6} 100644 $x40 ($x40) [AM]\tm$/o or next;
- unindex_oid($self, $git, $1);
+ unindex_oid($self, $git, $1, $unindexed);
}
delete $self->{reindex_pipe};
$fh = undef;
return unless $sync->{-opt}->{prune};
- my $after = scalar keys %$un;
+ my $after = scalar keys %$unindexed;
return if $before == $after;
# ensure any blob can not longer be accessed via dumb HTTP
@@ -1184,10 +1265,22 @@ sub index_sync {
# unindex is required for leftovers if "deletes" affect messages
# in a previous fetch+index window:
+ my $git;
if (my @leftovers = values %{delete $sync->{D}}) {
- my $git = $self->{-inbox}->git;
- unindex_oid($self, $git, $_) for @leftovers;
- $git->cleanup;
+ $git = $self->{-inbox}->git;
+ for my $oid (@leftovers) {
+ $self->{current_info} = "leftover $oid";
+ unindex_oid($self, $git, $oid);
+ }
+ }
+ if (my $multi_mid = delete $sync->{multi_mid}) {
+ $git //= $self->{-inbox}->git;
+
+ while (defined(my $oid = pop(@$multi_mid))) {
+ $self->{current_info} = "multi_mid $oid";
+ reindex_oid_m($self, $sync, $git, $oid);
+ }
+ $git->cleanup if $git;
}
$self->done;
diff --git a/t/v2reindex.t b/t/v2reindex.t
index 7c5a6b07..52711f8f 100644
--- a/t/v2reindex.t
+++ b/t/v2reindex.t
@@ -431,4 +431,70 @@ ok(!-d $xap, 'Xapian directories removed again');
], 'msgmap as expected' );
}
+# A real example from linux-renesas-soc on lore where a 3-headed monster
+# of a message has 3 sets of common headers. Another normal message
+# previously existed with a single Message-ID that conflicts with one
+# of the Message-IDs in the 3-headed monster.
+{
+ my @warn;
+ local $SIG{__WARN__} = sub { push @warn, @_ };
+ my %config = %$ibx_config;
+ $config{indexlevel} = 'basic';
+ my $ibx = PublicInbox::Inbox->new(\%config);
+ my $im = PublicInbox::V2Writable->new($ibx);
+ my $m3 = PublicInbox::MIME->new(<<'EOF');
+Date: Tue, 24 May 2016 14:34:22 -0700 (PDT)
+Message-Id: <20160524.143422.552507610109476444.d@example.com>
+To: t@example.com
+Cc: c@example.com
+Subject: Re: [PATCH v2 2/2]
+From: <f@example.com>
+In-Reply-To: <1463825855-7363-2-git-send-email-y@example.com>
+References: <1463825855-7363-1-git-send-email-y@example.com>
+ <1463825855-7363-2-git-send-email-y@example.com>
+Date: Wed, 25 May 2016 10:01:51 +0900
+From: h@example.com
+To: g@example.com
+Cc: m@example.com
+Subject: Re: [PATCH]
+Message-ID: <20160525010150.GD7292@example.com>
+References: <1463498133-23918-1-git-send-email-g+r@example.com>
+In-Reply-To: <1463498133-23918-1-git-send-email-g+r@example.com>
+From: s@example.com
+To: h@example.com
+Cc: m@example.com
+Subject: [PATCH 12/13]
+Date: Wed, 01 Jun 2016 01:32:35 +0300
+Message-ID: <1923946.Jvi0TDUXFC@wasted.example.com>
+In-Reply-To: <13205049.n7pM8utpHF@wasted.example.com>
+References: <13205049.n7pM8utpHF@wasted.example.com>
+
+Somehow we got a message with 3 sets of headers into one
+message, could've been something broken on the archiver side.
+EOF
+
+ my $m1 = PublicInbox::MIME->new(<<'EOF');
+From: a@example.com
+To: t@example.com
+Subject: [PATCH 12/13]
+Date: Wed, 01 Jun 2016 01:32:35 +0300
+Message-ID: <1923946.Jvi0TDUXFC@wasted.example.com>
+In-Reply-To: <13205049.n7pM8utpHF@wasted.example.com>
+References: <13205049.n7pM8utpHF@wasted.example.com>
+
+This is probably one of the original messages
+
+EOF
+ $im->add($m1);
+ $im->add($m3);
+ $im->done;
+ remove_tree($xap);
+ eval { $im->index_sync() };
+ is($@, '', 'no error from initial indexing');
+ is_deeply(\@warn, [], 'no warnings from initial index');
+ eval { $im->index_sync({reindex=>1}) };
+ is($@, '', 'no error from reindexing after reused Message-ID (x3)');
+ is_deeply(\@warn, [], 'no warnings on reindex');
+}
+
done_testing();
next prev parent reply other threads:[~2019-10-21 11:02 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-16 21:14 BUG: --reindex broken on multiple Message-IDs reuse Eric Wong
2019-10-17 11:22 ` [RFC] v2writable: reindex handles 3-headed monsters [was: BUG: --reindex broken on multiple Message-IDs reuse] Eric Wong
2019-10-22 8:09 ` [RFC/HELP] search: multiple From/To/Cc/Subject (what about Date?) Eric Wong
2019-10-21 11:02 ` [PATCH 0/3] fix reindex on multiple + overlapping Message-IDs Eric Wong
2019-10-21 11:02 ` [PATCH 1/3] v2writable: set unindexed article number Eric Wong
2019-10-21 11:02 ` [PATCH 2/3] v2writable: improve "num_for" API and disambiguate Eric Wong
2019-10-21 11:02 ` Eric Wong [this message]
2019-10-21 11:34 ` [PATCH 3/3] v2writable: reindex handles 3-headered monsters Eric Wong
2019-10-22 1:28 ` [PATCH 4/3] v2writable: move git->cleanup to the correct place Eric Wong
2019-10-22 1:29 ` [PATCH 5/3] v2writable: use msgmap as multi_mid queue Eric Wong
2019-10-23 18:19 ` [PUSHED] fix reindex on multiple + overlapping Message-IDs Eric Wong
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: https://public-inbox.org/README
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20191021110221.23753-4-e@80x24.org \
--to=e@80x24.org \
--cc=meta@public-inbox.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this public inbox
https://80x24.org/public-inbox.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).