X-Git-Url: http://www.git.stargrave.org/?p=public-inbox.git;a=blobdiff_plain;f=lib%2FPublicInbox%2FFilter.pm;h=8b78a44163743dcd4a235a1bc661471e29f091e6;hp=a1587682a3f3a8d4dd866db535b15a0623382252;hb=c1abb946e53e4179666ebb290e31c2d9ddc40711;hpb=e022d3377fd2c50fd9931bf96394728958a90bf3

diff --git a/lib/PublicInbox/Filter.pm b/lib/PublicInbox/Filter.pm
index a1587682..8b78a441 100644
--- a/lib/PublicInbox/Filter.pm
+++ b/lib/PublicInbox/Filter.pm
@@ -1,6 +1,7 @@
-# Copyright (C) 2013, Eric Wong <normalperson@yhbt.net> and all contributors
+# Copyright (C) 2013-2015 all contributors <meta@public-inbox.org>
 # License: AGPLv3 or later (https://www.gnu.org/licenses/agpl-3.0.txt)
 #
+# Used to filter incoming mail for -mda and importers
 # This only exposes one function: run
 # Note: the settings here are highly opinionated.  Obviously, this is
 # Free Software (AGPLv3), so you may change it if you host yourself.
@@ -12,32 +13,25 @@ use Email::MIME::ContentType qw/parse_content_type/;
 use Email::Filter;
 use IPC::Run;
 our $VERSION = '0.0.1';
+use constant NO_HTML => '*** We only accept plain-text email, no HTML ***';
+use constant TEXT_ONLY => '*** We only accept plain-text email ***';
 
 # start with the same defaults as mailman
-our $BAD_EXT = qr/\.(?:exe|bat|cmd|com|pif|scr|vbs|cpl)\z/i;
-our $MIME_HTML = qr!\btext/html\b!i;
+our $BAD_EXT = qr/\.(exe|bat|cmd|com|pif|scr|vbs|cpl|zip)\s*\z/i;
+our $MIME_HTML = qr!\btext/x?html\b!i;
 our $MIME_TEXT_ANY = qr!\btext/[a-z0-9\+\._-]+\b!i;
 
 # this is highly opinionated delivery
 # returns 0 only if there is nothing to deliver
 sub run {
-	my ($class, $mime) = @_;
+	my ($class, $mime, $filter) = @_;
 
 	my $content_type = $mime->header('Content-Type') || 'text/plain';
 
-	# kill potentially bad/confusing headers
-	# Note: ssoma already does this, but since we mangle the message,
-	# we should do this before it gets to ssoma.
-	# We also kill Mail-{Followup,Reply}-To and Reply-To headers due to
-	# the nature of public-inbox having no real subscribers.
-	foreach my $d (qw(status lines content-length
-			mail-followup-to mail-reply-to reply-to)) {
-		$mime->header_set($d);
-	}
-
 	if ($content_type =~ m!\btext/plain\b!i) {
 		return 1; # yay, nothing to do
 	} elsif ($content_type =~ $MIME_HTML) {
+		$filter->reject(NO_HTML) if $filter;
 		# HTML-only, non-multipart
 		my $body = $mime->body;
 		my $ct_parsed = parse_content_type($content_type);
@@ -45,8 +39,9 @@ sub run {
 		replace_body($mime, $body);
 		return 1;
 	} elsif ($content_type =~ m!\bmultipart/!i) {
-		return strip_multipart($mime, $content_type);
+		return strip_multipart($mime, $content_type, $filter);
 	} else {
+		$filter->reject(TEXT_ONLY) if $filter;
 		replace_body($mime, "$content_type message scrubbed");
 		return 0;
 	}
@@ -97,6 +92,7 @@ sub dump_html {
 		push @cmd, "-assume_charset=$charset";
 	}
 	if (IPC::Run::run(\@cmd, $body, \$out, \$err)) {
+		$out =~ s/\r\n/\n/sg;
 		$$body = $out;
 	} else {
 		# give them an ugly version:
@@ -105,12 +101,9 @@ sub dump_html {
 	}
 }
 
-# this is to correct user errors and not expected to cover all corner cases
-# if users don't want to hit this, they should be sending text/plain messages
-# unfortunately, too many people send HTML mail and we'll attempt to convert
-# it to something safer, smaller and harder-to-track.
+# this is to correct old archives during import.
 sub strip_multipart {
-	my ($mime, $content_type) = @_;
+	my ($mime, $content_type, $filter) = @_;
 
 	my (@html, @keep);
 	my $rejected = 0;
@@ -124,14 +117,16 @@ sub strip_multipart {
 		# some extensions are just bad, reject them outright
 		my $fn = $part->filename;
 		if (defined($fn) && $fn =~ $BAD_EXT) {
+			$filter->reject("Bad file type: $1") if $filter;
 			$rejected++;
 			return;
 		}
 
-		my $part_type = $part->content_type;
+		my $part_type = $part->content_type || '';
 		if ($part_type =~ m!\btext/plain\b!i) {
 			push @keep, $part;
 		} elsif ($part_type =~ $MIME_HTML) {
+			$filter->reject(NO_HTML) if $filter;
 			push @html, $part;
 		} elsif ($part_type =~ $MIME_TEXT_ANY) {
 			# Give other text attachments the benefit of the doubt,
@@ -139,22 +134,29 @@ sub strip_multipart {
 			# help with.
 
 			push @keep, $part;
-		} elsif ($part_type =~ m!\Aapplication/octet-stream\z!i) {
+		} elsif ($part_type eq '' ||
+		         $part_type =~ m!\bapplication/octet-stream\b!i) {
 			# unfortunately, some mailers don't set correct types,
 			# let messages of unknown type through but do not
 			# change the sender-specified type
 			if (recheck_type_ok($part)) {
 				push @keep, $part;
+			} elsif ($filter) {
+				$filter->reject("Bad attachment: $part_type ".
+						TEXT_ONLY);
 			} else {
 				$rejected++;
 			}
-		} elsif ($part_type =~ m!\Aapplication/pgp-signature\z!i) {
+		} elsif ($part_type =~ m!\bapplication/pgp-signature\b!i) {
 			# PGP signatures are not huge, we may keep them.
 			# They can only be valid if it's the last element,
 			# so we keep them iff the message is unmodified:
 			if ($rejected == 0 && !@html) {
 				push @keep, $part;
 			}
+		} elsif ($filter) {
+			$filter->reject("unacceptable mime-type: $part_type ".
+					TEXT_ONLY);
 		} else {
 			# reject everything else, including non-PGP signatures
 			$rejected++;
@@ -201,6 +203,10 @@ sub collapse {
 	my ($mime, $part) = @_;
 	$mime->header_set('Content-Type', $part->content_type);
 	$mime->body_set($part->body_raw);
+	my $cte = $part->header('Content-Transfer-Encoding');
+	if (defined($cte) && $cte ne '') {
+		$mime->header_set('Content-Transfer-Encoding', $cte);
+	}
 	mark_changed($mime);
 	return 1;
 }
@@ -220,8 +226,7 @@ sub replace_body {
 sub recheck_type_ok {
 	my ($part) = @_;
 	my $s = $part->body;
-	((bytes::length($s) < 0x10000) &&
-		($s =~ /\A([\P{XPosixPrint}\f\n\r\t]+)\z/))
+	((length($s) < 0x10000) && ($s =~ /\A([[:print:]\s]+)\z/s));
 }
 
 1;