TCP Sockt OLD DATA in recv buf nevr clears [SLVD?]

[KickinAezz]

for TCP_NODELAY =>
#include <netinet/tcp.h>


About the delay what range of values did you try?

Seems like TCP_NODELAY is the one that did the trick.
Disabled it; It occasionally lagged waiting for data.





Here's the updated one to try:
http://www.sendspace.com/file/d7e2mq


Would anyone mind dissecting sceInetRecv Kindly? Where does it read from?
[We could perhaps find the hardware address of wifi buffer and clean it out]

[SANiK]

From what you say, I guess that, what you see there is NAGLE-ING

You can't get rid of it because it's not up to you but it's up to the sender of the data if they want nagle-ing or not.

The only thing you can do is to write a system that processes each packet and takes out the most up to date information.

BUT, if you do recv() and then do recv() again and then you see the same data and one new data piece - then that's something else

[KickinAezz]

From what you say, I guess that, what you see there is NAGLE-ING

You can't get rid of it because it's not up to you but it's up to the sender of the data if they want nagle-ing or not.

The only thing you can do is to write a system that processes each packet and takes out the most up to date information.

BUT, if you do recv() and then do recv() again and then you see the same data and one new data piece - then that's something else

Yepp.

Now I need to focus on OpenXDK side and optimize the tcp stack once I finalize my other project.