[notmuch] [PATCH 2/2] notmuch-reply.c: Handle munged `Reply-To' headers.

[Jed Brown]

Some mailing lists engage in the evil practice of changing the Reply-To
header so that replies from all mailers go to the list by default, at
the expense of not responding to the person who actually sent the
message.  When this is detected, we reply to `From' and remove the
duplicate response to the mailing list.  Consider a reply to the
following message.

  From: Some User <some.user at example.com>
  To: Sample users list <sample-users at sample.org>
  Reply-To: Sample users list <sample-users at sample.org>

Prior to this patch, `notmuch reply' produces

  To: Sample users list <sample-users at sample.org>,
      Sample users list <sample-users at sample.org>

and after the patch,

  To: Some User <some.user at example.com>,
      Sample users list <sample-users at sample.org>

Signed-off-by: Jed Brown <jed at 59A2.org>
---
 notmuch-reply.c |   47 ++++++++++++++++++++++++++++++++++++++++++++++-
 1 files changed, 46 insertions(+), 1 deletions(-)

diff --git a/notmuch-reply.c b/notmuch-reply.c
index b91a830..9b9e9ab 100644
--- a/notmuch-reply.c
+++ b/notmuch-reply.c
@@ -182,6 +182,39 @@ add_recipients_for_string (GMimeMessage *message,
     return add_recipients_for_address_list (message, config, type, list);
 }
 
+/* Some mailing lists munge the Reply-To header despite it being A Bad
+ * Thing, see http://www.unicom.com/pw/reply-to-harmful.html
+ *
+ * This function detects such munging so that reasonable headers can be
+ * generated anyway.  Returns 1 if munged, else 0.
+ *
+ * The current logic is fairly naive, Reply-To is diagnosed as munged if
+ * it contains exactly one address, and this address is also present in
+ * the To or Cc fields.
+ */
+static int
+mailing_list_munged_reply_to (notmuch_message_t *message)
+{
+    const char *header, *addr;
+    InternetAddressList *list;
+    InternetAddress *address;
+    InternetAddressMailbox *mailbox;
+
+    header = notmuch_message_get_header (message, "reply-to");
+    list = internet_address_list_parse_string (header);
+    if (internet_address_list_length (list) != 1)
+	return 0;
+    address = internet_address_list_get_address (list, 0);
+    if (INTERNET_ADDRESS_IS_GROUP (address))
+	return 0;
+    mailbox = INTERNET_ADDRESS_MAILBOX (address);
+    addr = internet_address_mailbox_get_addr (mailbox);
+    /* Note that strcasestr() is a GNU extension, strstr() might be sufficient */
+    if (strcasestr (notmuch_message_get_header (message, "to"), addr) == 0 ||
+	strcasestr (notmuch_message_get_header (message, "cc"), addr) == 0)
+	return 1;
+    return 0; }
+
 /* Augments the recipients of reply from the headers of message.
  *
  * If any of the user's addresses were found in these headers, the first
@@ -192,7 +225,7 @@ add_recipients_from_message (GMimeMessage *reply,
 			     notmuch_config_t *config,
 			     notmuch_message_t *message)
 {
-    static const struct {
+    struct {
 	const char *header;
 	const char *fallback;
 	GMimeRecipientType recipient_type;
@@ -205,6 +238,18 @@ add_recipients_from_message (GMimeMessage *reply,
     const char *from_addr = NULL;
     unsigned int i;
 
+    /* When we have detected Reply-To munging, we ignore the Reply-To
+     * field (because it appears in the To or Cc headers) and use the
+     * From header so that person will get pinged and will actually
+     * receive the response if not subscribed to the list.  Note that
+     * under no circumstances does this fail to reply to the address in
+     * the Reply-To header.
+     */
+    if (mailing_list_munged_reply_to (message)) {
+	reply_to_map[0].header = "from";
+	reply_to_map[0].fallback = NULL;
+    }
+
     for (i = 0; i < ARRAY_SIZE (reply_to_map); i++) {
 	const char *addr, *recipients;
 
-- 
1.6.5.3