* Fixed one more case of a resulting 0 byte buffer. Also added a test case for

it to the test app. This finally fixes ticket #2594.
* Of course, I neither saw this one before, nor did I accidently reproduce it
  with a test :-)


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@28967 a95241bf-73f2-0310-859d-f6bbb57e9c96

src/add-ons/kernel/network/protocols/tcp/BufferQueue.cpp

@@ -176,9 +176,14 @@ BufferQueue::Add(net_buffer *buffer, tcp_sequence sequence)
 			fList.Remove(remove);
 			fNumBytes -= remove->size;
 			gBufferModule->free(remove);
-		} else {
+		} else if (tcp_sequence(next->sequence) > sequence) {
+			// We have the end of this buffer already
 			gBufferModule->remove_trailer(buffer,
 				sequence + buffer->size - next->sequence);
+		} else {
+			// We already have this data
+			gBufferModule->free(buffer);
+			buffer = NULL;
 		}
 	}
 

src/tests/kits/net/tcp_shell/BufferQueueTest.cpp

@@ -1,3 +1,8 @@
+/*
+ * Copyright 2009, Axel Dörfler, [email protected].
+ * Distributed under the terms of the MIT License.
+ */
+
 #define DEBUG_BUFFER_QUEUE 1
 
 #include "BufferQueue.h"
@@ -129,7 +134,11 @@ main()
 	dump("add 3");
 
 	add(60, 540);
-	dump("add at the end of previous data");
+	dump("added at the end of previous data");
+	
+	add(998, 1002);
+	add(500, 1000);
+	dump("added data covered by next");
 
 	put_module(NET_BUFFER_MODULE_NAME);
 	return 0;