Re: [Dev-C++] Character variables

[Per W. <pw...@ia...>]

And if you check closer, for shift operations it's the >> that is
implementation specific, i.e. if it will do an arithmetic or logic shift
right for signed numbers.

The << is always ok, i.e. will always shift in a zero. Is's only when you
try to use it to multiply with 2, you have to note that because of
overflow, you might shift away your sign bit, resulting in a bit of
confusion, i.e. [16-bit integers]:
 -20000 (0xB1E0) << 1 results in 25536 (0x63c0)
but then all operators will give more or less strange results when
operated outside the available numeric range.

About compatibility with strange architectures, I don't expect us to have
to see too many machines with 9-bit characters and 36-bit integers
anymore, as existed for some big old iron. Such a computer manufacturer
would now have too hard a time trying to attract both compiler vendors and
actual end users.

/Per W

On Wed, 5 Mar 2003, Ioannis Vranos wrote:

> -------Original Message-------
>
> From: Per Westermark
> Date: =D4=E5=F4=DC=F1=F4=E7, 5 =CC=DC=F1=F4=E9=EF=F2 2003 2:04:36 =EC=EC
> To: Ioannis Vranos
> Cc: Dev-C++ Mailing List
> Subject: Re: [Dev-C++] Character variables
>
> The shift operators are of course defiend for both signed and unsigned
> data types.
>
> It's just that for >> you may not assume if it will shift in a '0' or a
> '1' in the most significant bit fir signed integers, i.e.
>
> -3 (0xfffd) >> 1 =3D 32766 (0x7ffe)
> -3 (0xfffd) >> 1 =3D -2 (0xfffe)
>
> depending on implementation.
>
> However, when shifting in characters into a 32-bit integer, I don't have
> to care about this, since << always shifts in 0 in the least significant
> bit.
>
>
> =3D=3D> Yes bitwise operators can be applied to integral types signed and
> unsigned however in signed types the behaviour is implementation
> dependent/undefined. This is explicitly mentioned in the C99 standard:
>
> =3D=3D> "Some operators (the unary operator ~, and the binary operators <=
<, >>,
> &, ^, and |,
> collectively described as bitwise operators) are required to have operand=
s
> that have
> integer type. These operators return values that depend on the internal
> representations of
> integers, and have implementation-defined and undefined aspects for signe=
d
> types."
>
>
> =3D=3D> The above is not mentioned explicitly in the C++98 standard, but =
it is
> common sense. That's why i spoke about bad magic. :)
>
>
>
> Ioannis Vranos
>
> * Programming pages: http://www.noicys.freeurl.com
> * Alternative URL: http://run.to/noicys