[RFC] mid: filter out 'y', 'n', and email addresses from references()

[RFC] mid: filter out 'y', 'n', and email addresses from references()

[Eric Wong]

Looking at git@vger history, several emails had broken
References/In-Reply-To pointing to <y>, <n> and email
addresses as Message-IDs in References and In-Reply-To
headers.

This was causing too many unrelated messaes to be linked
together in the same thread.
---
 lib/PublicInbox/MID.pm | 25 +++++++++++++++++++------
 t/mid.t                |  4 ++++
 2 files changed, 23 insertions(+), 6 deletions(-)

diff --git a/lib/PublicInbox/MID.pm b/lib/PublicInbox/MID.pm
index cd56f27..7f1ab15 100644
--- a/lib/PublicInbox/MID.pm
+++ b/lib/PublicInbox/MID.pm
@@ -10,6 +10,7 @@ our @EXPORT_OK = qw/mid_clean id_compress mid2path mid_mime mid_escape MID_ESC
 	mids references/;
 use URI::Escape qw(uri_escape_utf8);
 use Digest::SHA qw/sha1_hex/;
+require PublicInbox::Address;
 use constant {
 	MID_MAX => 40, # SHA-1 hex length # TODO: get rid of this
 	MAX_MID_SIZE => 244, # max term size (Xapian limitation) - length('Q')
@@ -79,22 +80,34 @@ sub references ($) {
 			push(@mids, ($v =~ /<([^>]+)>/sg));
 		}
 	}
-	uniq_mids(\@mids);
+
+	# old versions of git-send-email would prompt users for
+	# In-Reply-To and users' muscle memory would use 'y' or 'n'
+	# as responses:
+	my %addr = ( y => 1, n => 1 );
+
+	foreach my $f (qw(To From Cc)) {
+		my @v = $hdr->header_raw($f);
+		foreach my $v (@v) {
+			$addr{$_} = 1 for (PublicInbox::Address::emails($v));
+		}
+	}
+	uniq_mids(\@mids, \%addr);
 }
 
-sub uniq_mids ($) {
-	my ($mids) = @_;
+sub uniq_mids ($;$) {
+	my ($mids, $seen) = @_;
 	my @ret;
-	my %seen;
+	$seen ||= {};
 	foreach my $mid (@$mids) {
 		$mid =~ tr/\n\t\r//d;
 		if (length($mid) > MAX_MID_SIZE) {
 			warn "Message-ID: <$mid> too long, truncating\n";
 			$mid = substr($mid, 0, MAX_MID_SIZE);
 		}
-		next if $seen{$mid};
+		next if $seen->{$mid};
 		push @ret, $mid;
-		$seen{$mid} = 1;
+		$seen->{$mid} = 1;
 	}
 	\@ret;
 }
diff --git a/t/mid.t b/t/mid.t
index 8c307c8..69a8a70 100644
--- a/t/mid.t
+++ b/t/mid.t
@@ -36,6 +36,10 @@ is(mid_escape('foo%!@(bar)'), 'foo%25!@(bar)');
 	$mime->header_set('Message-ID', "<hello\tworld>");
 	is_deeply(mids($mime->header_obj), ['helloworld'],
 		'drop \t in Message-ID');
+
+	$mime->header_set('To', 'u@example.com');
+	$mime->header_set('References', '<hello> <world> <n> <u@example.com>');
+	is_deeply(references($mime->header_obj), [qw(hello world)]);
 }
 
 done_testing();
-- 
EW

Re: [RFC] mid: filter out 'y', 'n', and email addresses from references()

[Eric Wong]

This seems to work, but not with --reindex for public-inbox-index
because --reindex still reuses "tid" information.

Doing this online without downtime will require more work...