[Commit-gnuradio] [gnuradio] 05/06: Fix an error in which the wrong header is allowed through

[git]

This is an automated email from the git hooks/post-receive script.

jcorgan pushed a commit to branch maint
in repository gnuradio.

commit cba91e3a25d27e39379fe41381beaf55a52ab43f
Author: Richard C. Bell <address@hidden>
Date:   Mon Jun 1 18:35:52 2015 -0700

    Fix an error in which the wrong header is allowed through
    
    Added break statements to every case. This fixed issues we were having
    with this block in which the wrong header would be allowed through,
    sometimes seconds after start, sometimes hours after start,
    depending on the HPD settings used. With the break statements,
    general_work must be called each time a new state begins, which removes
    any sample offset issues that could be happening in the original fall
    through method. I suspect the original methods issue is caused by the
    check_items_available function, though I have not confirmed this. I've
    tested this fix against the current built-in block implementation in my
    own custom flowgraph, and in the rx_ofdm.grc example. The fix works in
    both cases.
---
 gr-digital/lib/header_payload_demux_impl.cc | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff --git a/gr-digital/lib/header_payload_demux_impl.cc 
b/gr-digital/lib/header_payload_demux_impl.cc
index 160f540..89428fa 100644
--- a/gr-digital/lib/header_payload_demux_impl.cc
+++ b/gr-digital/lib/header_payload_demux_impl.cc
@@ -221,7 +221,13 @@ namespace gr {
          in += d_itemsize;
          nread++;
          d_state = STATE_FIND_TRIGGER;
-         // Fall through
+         // The following break was added to this state as well as 
STATE_FIND_TRIGGER
+      // and STATE_HEADER. There appears to be a bug somewhere in this code 
without
+      // the breaks that can lead to failure of this block. With the breaks in 
the code
+      // testing has shown more stable performance with various block 
paramters.
+      // If an offset calculation bug is found and fixed, it should be 
possible to 
+      // remove these breaks for some performance increase.
+    break; 
 
        case STATE_FIND_TRIGGER:
          trigger_offset = find_trigger_signal(nread, noutput_items, 
input_items);
@@ -234,7 +240,7 @@ namespace gr {
          consume_each (trigger_offset);
          in += trigger_offset * d_itemsize;
          d_state = STATE_HEADER;
-         // Fall through
+    break;
 
        case STATE_HEADER:
          if (check_items_available(d_header_len, ninput_items, noutput_items, 
nread)) {
@@ -262,7 +268,7 @@ namespace gr {
          consume_each (nread);
          in += nread * d_itemsize;
          d_state = STATE_PAYLOAD;
-         // Fall through
+    break;
 
        case STATE_PAYLOAD:
          if (check_items_available(d_curr_payload_len, ninput_items, 
noutput_items, nread)) {
@@ -274,7 +280,7 @@ namespace gr {
            set_min_noutput_items(d_output_symbols ? 1 : (d_items_per_symbol + 
d_gi));
            d_state = STATE_FIND_TRIGGER;
          }
-         break;
+       break;
 
        default:
          throw std::runtime_error("invalid state");