more elegant solution to the "out-of-band" problem 7001800

diff --git a/rigctl.c b/rigctl.c
index 286aaf9edc0dcc3167010f3fcaa7289b4a493b1f..ec0f8b57d1fb4b1d02d153f03bd626b93d11921b 100644 (file)
--- a/rigctl.c
+++ b/rigctl.c
@@ -490,28 +490,30 @@ static gpointer rigctl_cw_thread(gpointer data)
     }
     strncpy(local_buf, cw_buf, 30);
     cw_busy=0; // mark buffer free again
-    // these values may have changed
+    // these values may have changed, so recompute them here
     dotlen = 1200000L/(long)cw_keyer_speed;
     dashlen = (dotlen * 3 * cw_keyer_weight) / 50L;
     dotsamples = 57600 / cw_keyer_speed;
     dashsamples = (3456 * cw_keyer_weight) / cw_keyer_speed;
     CAT_cw_is_active=1;
-    // if out-of-band, ptt_update() has no effect
-    // and mox will not appear. In this case, we should
-    // skip the pending CW message to avoid an infinite
-    // loop
-    if (!canTransmit() && ! tx_out_of_band) {
-       fprintf(stderr,"CAT CW skipped -- out of band.\n");
-       continue;
-    }
     if (!mox) {
        // activate PTT
         g_idle_add(ext_ptt_update ,(gpointer)1);
        // have to wait until it is really there
-        while (!mox) usleep(50000L);
-       // some extra time to settle down
+       // Note that if out-of-band, we would wait
+       // forever here, so allow at most 500 msec
+       i=10;
+        while (!mox && (i--) > 0) usleep(50000L);
+       // still no MOX? --> silently discard CW message and give up
+       if (!mox) {
+           CAT_cw_is_active=0;
+           continue;
+       }
+       // some extra time to settle down, in order NOT to loose
+       // the first dit or dah
         usleep(100000L);
     }
+    // At this point, mox==1 and CAT_cw_active == 1
     i=0;
     while(((c=local_buf[i++]) != '\0') && !cw_key_hit && !cw_not_ready) {
         rigctl_send_cw_char(c);