M20150601 patch （at socket.c）may cause stack corruption on 16-bit word platforms (e.g. DSP28335)

[ShuaiZhang1998]

In commit M20150601 the code was changed from:

uint8_t taddr[4];
getSIPR(taddr);

to:

uint32_t taddr;
getSIPR((uint8_t*)&taddr);

This works on most platforms, but on 16-bit word architectures (e.g. TI C28x / DSP28335) with optimizations enabled and uint8_t typedef’d as unsigned char, it can cause stack corruption and wrong return addresses.

sizeof(uint8_t[4]); // 4 chars = 64 bits
sizeof(uint32_t); // 2 chars = 32 bits

Example disassembly (DSP28335):
safe:

socket.c:
328042:   B2BD        MOVL         *SP++, XAR1
328043:   AABD        MOVL         *SP++, XAR2
328044:   A2BD        MOVL         *SP++, XAR3
328045:   FE06        ADDB         SP, #6   \\!!!
328046:   5AA5        MOVZ         AR2, @AR5
...
RPC OK

unsafe

socket.c:
328042:   B2BD        MOVL         *SP++, XAR1
328043:   AABD        MOVL         *SP++, XAR2
328044:   A2BD        MOVL         *SP++, XAR3
328045:   FE04        ADDB         SP, #4  \\!!!
328046:   5AA5        MOVZ         AR2, @AR5
...
RPC corrupted (e.g., 0x190001 )

This corrupts the stack, leading to a wrong return address from getSIPR()