On Mar 14, 2008, at 12:30 PM, Matthew Toseland wrote:
> On Friday 14 March 2008 15:40, Robert Hailey wrote:
>>
>> [...]
>>
>> Unless I read this wrong you have inverted the return value
>> (isLocalAddress()==!shouldThrottle()), but did not negate the
>> condition where it is used.
>>
>> The code in PacketThrottle now reads:
>>
>> if (should NOT throttle) {
>> log("throttling");
>> ...
>> }
>
> You're right... I don't understand why the simulation worked in that
> case...
Well... the real problem would have been if it had got to the live
network (no throttling where it needs it), in the sim it was just
throttling where it didn't need to.
Along the same lines... shouldn't we not report that they have already
been reported? Unless this is how we get fast local transfers in the
first place (mis-reporting). Something like this:
Index: src/freenet/io/xfer/PacketThrottle.java
===================================================================
--- src/freenet/io/xfer/PacketThrottle.java (revision 18536)
+++ src/freenet/io/xfer/PacketThrottle.java (working copy)
@@ -210,15 +210,17 @@
Logger.minor(this, "Congestion control wait time:
"+waitTime+"
for "+this);
MyCallback callback = new MyCallback();
try {
+ int alreadyReportedBytes=0;
if(((PeerNode)peer).shouldThrottle()) {
if(logMINOR) Logger.minor(this, "Throttling
"+peer.shortToString()+" : "+packetSize+" for "+this);
long startTime = System.currentTimeMillis();
overallThrottle.blockingGrab(packetSize);
+ alreadyReportedBytes=packetSize;
long delayTime = System.currentTimeMillis() -
startTime;
((PeerNode)peer).reportThrottledPacketSendTime(delayTime);
} else if(logMINOR)
Logger.minor(this, "Not throttling
"+peer.shortToString()+" for
"+this);
- peer.sendAsync(msg, callback, packetSize, ctr);
+ peer.sendAsync(msg, callback, alreadyReportedBytes,
ctr);
} catch (RuntimeException e) {
callback.fatalError();
throw e;
--
Robert Hailey
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<https://emu.freenetproject.org/pipermail/devl/attachments/20080314/f1d7ee4c/attachment.html>
