GitLab

When using OTRL_FRAGMENT_SEND_ALL, libotr handles outgoing unmodified plaintext messages confusingly.

With libotr 3 it was the caller's responsibility to send any messages or fragments after otrl_message_sending. There were three cases:

• otrl_message_sending returns an error. We should send nothing.
• otrl_message_sending returns success and sets its messagep argument. We should send *messagep via otrl_fragment_and_send.
• otrl_message_sending returns success and sets messagep to NULL. We should send the original message.

The OTRL_FRAGMENT_SEND_ALL path of libotr 4.0.0 handles 1 (as in not sending anything) and 2 but leaves case 3 to the caller.

This is understandable behaviour from the inside but confusing to the user (of libotr). I at least didn't expect a NULL value from an out parameter which one isn't otherwise interested in to signify a need for action. Moreover I mistook SEND_ALL to mean that libotr would take care of all sending.

Might I suggest the following clarifying addition to the documentation of otrl_message_sending? (Also, I think messagep was meant in the line above, please confirm.)

--- a/src/message.c
+++ b/src/message.c
@@ -192,7 +192,10 @@ void otrl_message_free(char *message)
  * Other fragmentation policies are OTRL_FRAGMENT_SEND_ALL,
  * OTRL_FRAGMENT_SEND_ALL_BUT_LAST, or OTRL_FRAGMENT_SEND_ALL_BUT_FIRST. In
  * these cases, the appropriate fragments will be automatically sent. For the
- * last two policies, the remaining fragment will be passed in *original_msg.
+ * last two policies, the remaining fragment will be passed in *messagep.
+ *
+ * With OTRL_FRAGMENT_SEND_ALL, if the routine returns non-zero and sets
+ * *messagep to NULL, you must still send the original message.
  *
  * Call otrl_message_free(*messagep) if you don't need *messagep or when you're
  * done with it. */

Cf. also http://bugs.bitlbee.org/bitlbee/ticket/1109#comment:9

(original: http://sourceforge.net/p/otr/bugs/26/)

(from redmine: created on 2014-04-18)