At 10:01 11/18/2001 +0000, Ian Chapman wrote:
>Maybe I'm missing something here, but it seems to me that this approach
>only works correctly for the cases where |c0-c1| < 128 (perhaps this is
>Gerhard's point about overflow not being taken into account).

This is correct. This calculation is in the "signed char domain", where
values with |value|>128 are not defined and results that fall into this
range may not behave as expected.

>Nasty, eh?
>
>Of course, this code would work as intended in a compiler which treats
>"chars" as 16-bit values (signed or unsigned).  One of the reasons why I
>have traditionally disliked 'C' in an embedded environment is that I
>have had to put a lot of effort into "working around" assumptions made
>by the compiler in order to get sensible code.

I don't think a lot of effort is needed. Whenever I start to use a new
compiler, I read once through the docs and "translate" the compiler's types
into my own types (i8, u8, ...). And the other thing you have to take into
account is that C (as low level language) does not handle overflow on
integer types. On any C compiler I've seen, an addition like (255 + 1) on
an 8bit unsigned type results in 0, not in an overflow. C mostly behaves
pretty much as assembler would, too. (Which is the reason why I prefer it
in embedded environments... :)

ge

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads