recognize more charsets than utf-8/iso-8859-1/us-ascii when parsing message headers with address

as they occur in From/To headers, for example: "From:
=?iso-8859-2?Q?Krist=FDna?= <[email protected]>".  we are using net/mail to parse
such headers. most address-parsing functions in that package will only decode
charsets utf-8, iso-8859-1 and us-ascii. we have to be careful to always use
net/mail.AddressParser with a WordDecoder that understands more that the
basics.

for issue #204 by morki, thanks for reporting!

Author: mjl-

message/addr.go

@@ -0,0 +1,29 @@
+package message
+
+import (
+	"fmt"
+	"net/mail"
+
+	"github.com/mjl-/mox/smtp"
+)
+
+// ParseAddressList parses a string as an address list header value
+// (potentially multiple addresses, comma-separated, with optional display
+// name).
+func ParseAddressList(s string) ([]Address, error) {
+	parser := mail.AddressParser{WordDecoder: &wordDecoder}
+	addrs, err := parser.ParseList(s)
+	if err != nil {
+		return nil, fmt.Errorf("parsing address list: %v", err)
+	}
+	r := make([]Address, len(addrs))
+	for i, a := range addrs {
+		addr, err := smtp.ParseNetMailAddress(a.Address)
+		if err != nil {
+			return nil, fmt.Errorf("parsing adjusted address %q: %v", a.Address, err)
+		}
+		r[i] = Address{a.Name, addr.Localpart.String(), addr.Domain.ASCII}
+
+	}
+	return r, nil
+}

message/addr_test.go

@@ -0,0 +1,11 @@
+package message
+
+import (
+	"testing"
+)
+
+func TestParseAddressList(t *testing.T) {
+	l, err := ParseAddressList("=?iso-8859-2?Q?Krist=FDna?= <[email protected]>, [email protected]")
+	tcheck(t, err, "parsing address list")
+	tcompare(t, l, []Address{{"Kristýna", "k", "example.com"}, {"", "mjl", "mox.example"}})
+}

message/part.go

@@ -489,7 +489,12 @@ func parseEnvelope(log mlog.Log, h mail.Header) (*Envelope, error) {
 
 func parseAddressList(log mlog.Log, h mail.Header, k string) []Address {
 	// todo: possibly work around ios mail generating incorrect q-encoded "phrases" with unencoded double quotes? ../rfc/2047:382
-	l, err := h.AddressList(k)
+	v := h.Get(k)
+	if v == "" {
+		return nil
+	}
+	parser := mail.AddressParser{WordDecoder: &wordDecoder}
+	l, err := parser.ParseList(v)
 	if err != nil {
 		return nil
 	}

message/part_test.go

@@ -603,3 +603,10 @@ func TestNetMailAddress(t *testing.T) {
 	tcheck(t, err, "parse")
 	tcompare(t, p.Envelope.From, []Address{{"", `" "`, "example.com"}})
 }
+
+func TestParseQuotedCharset(t *testing.T) {
+	const s = "From: =?iso-8859-2?Q?Krist=FDna?= <[email protected]>\r\n\r\nbody\r\n"
+	p, err := EnsurePart(pkglog.Logger, false, strings.NewReader(s), int64(len(s)))
+	tcheck(t, err, "parse")
+	tcompare(t, p.Envelope.From, []Address{{"Kristýna", "k", "example.com"}})
+}

sendmail.go

@@ -9,7 +9,6 @@ import (
 	"io"
 	"log"
 	"net"
-	"net/mail"
 	"os"
 	"path/filepath"
 	"slices"
@@ -19,6 +18,7 @@ import (
 	"github.com/mjl-/sconf"
 
 	"github.com/mjl-/mox/dns"
+	"github.com/mjl-/mox/message"
 	"github.com/mjl-/mox/mox-"
 	"github.com/mjl-/mox/sasl"
 	"github.com/mjl-/mox/smtp"
@@ -196,16 +196,12 @@ binary should be setgid that group:
 						}
 						recipient = submitconf.DefaultDestination
 					} else {
-						addrs, err := mail.ParseAddressList(s)
+						addrs, err := message.ParseAddressList(s)
 						xcheckf(err, "parsing To address list")
 						if len(addrs) != 1 {
 							log.Fatalf("only single address allowed in To header")
 						}
-						addr, err := smtp.ParseNetMailAddress(addrs[0].Address)
-						if err != nil {
-							log.Fatalf("parsing address: %v", err)
-						}
-						recipient = addr.Pack(false)
+						recipient = addrs[0].User + "@" + addrs[0].Host
 					}
 				}
 				if k == "to" {

webmail/api.go

@@ -517,13 +517,14 @@ type File struct {
 // parseAddress expects either a plain email address like "user@domain", or a
 // single address as used in a message header, like "name <user@domain>".
 func parseAddress(msghdr string) (message.NameAddress, error) {
-	a, err := mail.ParseAddress(msghdr)
+	// todo: parse more fully according to ../rfc/5322:959
+	parser := mail.AddressParser{WordDecoder: &wordDecoder}
+	a, err := parser.Parse(msghdr)
 	if err != nil {
 		return message.NameAddress{}, err
 	}
 
-	// todo: parse more fully according to ../rfc/5322:959
-	path, err := smtp.ParseAddress(a.Address)
+	path, err := smtp.ParseNetMailAddress(a.Address)
 	if err != nil {
 		return message.NameAddress{}, err
 	}
@@ -1658,12 +1659,12 @@ func recipientSecurity(ctx context.Context, log mlog.Log, resolver dns.Resolver,
 		SecurityResultUnknown,
 	}
 
-	msgAddr, err := mail.ParseAddress(messageAddressee)
+	parser := mail.AddressParser{WordDecoder: &wordDecoder}
+	msgAddr, err := parser.Parse(messageAddressee)
 	if err != nil {
-		return rs, fmt.Errorf("parsing message addressee: %v", err)
+		return rs, fmt.Errorf("parsing addressee: %v", err)
 	}
-
-	addr, err := smtp.ParseAddress(msgAddr.Address)
+	addr, err := smtp.ParseNetMailAddress(msgAddr.Address)
 	if err != nil {
 		return rs, fmt.Errorf("parsing address: %v", err)
 	}