Dear All,
It is my exp that what TCP/IP stack code is puiblished as opensource by nichelite at NXP site, seems very buggy.
I spent days to make it correct but a very little support is there, and these (nichelite) folks ask you to pay lots of $$$ for such low end stack support.
Batter to try uTasker. I am tryting to get aroud with this.
Pls avoid NicheLite anymore.. !
Could you be more specific - in what way(s) did you find it "buggy"?
1) Getting disconnected from LAN at IP level while physical Connectivity is up (Checked by DP Phy read).
2) At the time of network stress (two three icmp) request running in parallel, it shows Task blocking and hang..
Although this can be partially avoided (minimize) using pretask task with larger memory stack for tasks but , still after long hours it will give the same problem.
3) Heap and stack management is not transparent to users - not enough documented (not documented at all). So, only way to dig is waste days by understand the code.
4) TCP Socket management is very poor, Many times stack is not able to close the socket or remote connection termination.
5) Biggest flow is that packet management (allocation/ retransmission/ de allocation) mechanism of the stack is under doubt , as I made long observations of retransmitting the TCP packets again and again. Duplicate allocation of the same packets and on long hours its BigPack and LilPack buffers getting overflows.
All my observations on MCB-2300 board with keil compiler code what they have provided.
Have you notified NXP about this?
Or, at least, posted it on their forum?
No i will have