pop3-login: Fix the next command failing after unsuccessful AUTH

If the AUTH command failed (e.g. bad password or proxying failure), the
next command was still processed as AUTH command. For example:

auth plain .
-ERR [AUTH] Invalid base64 data in initial response
quit
-ERR [AUTH] Unsupported authentication mechanism.

diff --git a/src/pop3-login/client-authenticate.c b/src/pop3-login/client-authenticate.c
index 21716d34b43cda52c33e2d732cf4d9127dac7e93..e2158333959a02540ee2f24b4c67d52802d45445 100644 (file)
--- a/src/pop3-login/client-authenticate.c
+++ b/src/pop3-login/client-authenticate.c
@@ -127,6 +127,15 @@ int cmd_auth(struct pop3_client *pop3_client)
                ir = t_strdup(str_c(client->auth_response));
 
        pop3_client->auth_mech_name_parsed = FALSE;
+       /* The whole AUTH line command is parsed now. The rest of the SASL
+          protocol exchange happens in login-common code. We can free the
+          current command here already, because no pop3-login code is called
+          until the authentication is finished. Also, there's currently no
+          single location that is called in pop3-login code after the
+          authentication is finished. For example it could be an auth failure
+          or it could be a successful authentication with a proxying
+          failure. */
+       i_free(pop3_client->current_cmd);
        return client_auth_begin(client, t_strdup(client->auth_mech_name), ir);
 }