LWIP standalong server bug?

[Altera_Forum]

Hi, 

 Examining the example lwip_server, in particular httpd.c.  When tcp_close is called the pcb is deallocated, so surely a tcp_new,bind must be performed again?  in the LWIP source tcp.c you can see the memp_free?  is there any good documentation for the LWIP rawapi, the little rawapi.txt file does not talk about pbuf management - such as is done within http_recv() for example?  is the pbuf_free() necessary? and should the tcp_recved(pcb, p->tot_len) not actually be tcp_recved(pcb,p->len) ? tot_len is the sze of the buffer and all other chained ones? (or so the LWIP source comments say?)  Are there any other examples of LWIP rawapi apps?  Thanks!  http://forum.niosforum.com/work2/style_emoticons/<#EMO_DIR#>/huh.gif  Thanks!

[Altera_Forum]

I&#39;m also trying to use the standalone LwIP Raw API in a more extensive application than the http-server example. 

I share the complaint that you made : the &#39;rawapi.txt &#39; documentation is far from complete.  But to be maybe of a little help to you : the thesis written by Adam Dunkels (the original designer of LwIP) is available on the internet. in Chapter 4 it discusses in general how the TCP/IP stack works and you can find some more details about memory management and the use of the pbufs.  You can find the thesis at   http://www.sics.se/~adam/thesis.pdf (http://www.sics.se/~adam/thesis.pdf)  I hope this answers some questions...