On 05/25/2011 12:52 AM, George Spelvin wrote:

quoted

V3:
- added function crc8_create().
- crc8_create() takes polynomial and bit direction as parameters.

May I suggest that crc8_create is A Stupid Idea.  Since the bit order
(and polynomial) will always be compile-time constants, just let the
call sites call crc8_create_[lm]sb_first() directly.

Thanks for be subtle in your suggestion. The crc8 module can have 
multiple callers, which have their own bit order and polynomial. From 
caller perspective they are likely compile-time constants. You are right 
that the extra layer crc8_create is not adding much and callers should 
the bit order specific create functions directly.

That way there's no need for the enum, and specifying something
other than lsb_first and msb_first is a compile-time error rather than
a run-time one

If callers use the specific function they can not specify a bit order so 
no compile-time error.

It might be nice to add some const declarations, and use size_t
for the buffer length:
u8 crc8(u8 const *table, u8 const *pdata, size_t nbytes, u8 crc)
or
u8 crc8(u8 const table[256], u8 const *pdata, size_t nbytes, u8 crc)


Style points you might consider, but I do not consider essential:

Personally, when a function parameter is a pointer to
a fixed-size array, I prefer to declare it as
	void crc8_create_lsb_first(u8 table[256], u8 poly)
for better documentation, but that's your choice.

The CRC8_GOOD_VALUE could be explained if you like.  "If a CRC is inverted
before transmission, the CRC computed over the while (message+crc)
is _table[x], when x is the bit pattern of the modification (almost
always 0xff)."

Great remarks. Will update the code.

Thanks!

Thanks.

Gr. AvS

-- 
Almost nobody dances sober, unless they happen to be insane.
-- H.P. Lovecraft --