Ponting ioLibrary_v100 to 16bit, bugs

Question

All, I have come across 3 bug today in code from ioLibrary_v100 that has issues when ported to 16bit. Function “wiz_send_data”, in W5500.c: uint16_t ptr = 0; needs changed to a uint32. If not the upper 8 bits gets truncated during addrsel = (ptr << 8) + (WIZCHIP_TXBUF_BLOCK(sn) << 3); This causes garbage to be sent out during a send data. Also changed it in wiz_recv_data, although it was working. Function “inet_ntoa” and “inet_ntoa”_pad, in sockutil.c: sprintf((char*)addr_str,“%d.%d.%d.%d”,(int32_t)(addr>>24 & 0xFF),(int32_t)(addr>>16 & 0xFF),(int32_t)(addr>>8 & 0xFF),(int32_t)(addr & 0xFF)) int32’s needs changed to int16’s. It may be compiler dependent, but for me I had to change it to int16’s or I was getting extra characters out of the sprintf function. Function “loopback_tcps”, in loopback.c: ret is defined as int32, but in this line: SendLen = TCPSend(sock, data_buf, ret) function TCPSend is expecting the length to be uint32. Had to add a type cast like this: SendLen = TCPSend(sock, data_buf, (Uint32)ret) with out the type cast garbage was passed to TCPSend. Maybe this will help someone, and maybe for future updates. Thanks, Jim

JimMartz · Answer

Come across another one today. Function “loopback_udp”, in loopback.c: The destination IP is defined as, uint32_t destip = 0; The problem is that the UDPRecv function places the IP address in 4 consecutive memory locations, in the case of a 16bit processor, it uses 4 16bit words, this don’t fit into the 32bit address defined. After the UDPRecv function the memory is Eg. 0x00A800C0 (192.0.168.0). To resolve this I defined the destination IP like: Uint8 destip[4] = { 0 }; There are pointer castings on the function, (Uint8*)&destip but this did not seem to matter to my compiler. Jim,