RE: RE: Re: CRC problem with 8 bit truncated polyno mial

From: Vladimir Prus (yg-boost-users_at_[hidden])
Date: 2003-07-09 10:08:22

Jim.Hyslop wrote:

> Vladimir Prus wrote:
>> Sorry, I don't understand what you're saying...
>> 
>> std::cout << "Optimal CRC-8 : "  << std::hex << (unsigned)crc_8()
>>      << std::endl;
>> 
>> is the same as
>> 
>> std::cout << "Optimal CRC-8 : "  << std::hex << (unsigned int)crc_8()
>>      << std::endl;
> Sorry, I wasn't very clear in my last message.
> 
> The results can be unexpected, depending on whether the result of crc_8()
> (or more precisely, operator() ) is signed or unsigned. If operator()
> returns an unsigned value, there will be no problems, and the output will
> be ea.
> 
> On the other hand, if operator() returns a signed value, then it will
> first be promoted to a signed int, then converted to an unsigned int.
> Using the original example values posted by Victoria, the output would be
> ffffffea. In this scenario, the return value must be first cast to an
> unsigned char, then to an unsigned int:
>   ... << static_cast<unsigned>( static_cast<unsigned char>( crc_8() ) )

I now understand what you meant. Yep, I was hit by this problem a couple of
times.

> I just had a quick look at the crc.hpp file, and operator() returns an
> unsigned value, thus banishing my caveat to the dusty corners where all
> academic footnotes end up ;-)

:-)
I've implicitly though return type is always unsigned. Using signed types
for bit-manipulation is way too messy.
 
- Volodya