Revise comment on when we should switch from writing back to reading

on a TCP stream.  (We don't currently, ever, and we probably should
eventually, but sometimes closing the connection is required.)

git-svn-id: svn://anonsvn.mit.edu/krb5/trunk@18234 dc483132-0cff-0310-8789-dd5450dbe970

diff --git a/src/kdc/network.c b/src/kdc/network.c
index f7bb958f3e7c0668474ab3f3426f61e36d5d8468..42c487a2c8951de69bfd5d005d46b32ee3740e2e 100644 (file)
--- a/src/kdc/network.c
+++ b/src/kdc/network.c
@@ -951,7 +951,10 @@ process_tcp_connection(struct connection *conn, const char *prog, int selflags)
        }
        if (conn->u.tcp.sgnum == 0) {
            /* finished sending */
-           /* should go back to reading */
+           /* We should go back to reading, though if we sent a
+              FIELD_TOOLONG error in reply to a length with the high
+              bit set, RFC 4120 says we have to close the TCP
+              stream.  */
            goto kill_tcp_connection;
        }
     } else if (selflags & SSF_READ) {