Re: String literals and wide_string literals - how?

[Simon Wright <simon.j.wright@mac.com>]

"Jeffrey R. Carter" <jrcarter@acm.org> writes:

> Simon Wright wrote:
>>
>> It may not be useful but in some applications it's essential to know.
>>
>> And it seems to me that a compiler that thought 2#00110011# was
>> acceptable probably wouldn't itself be acceptable.
>
> Why not? It adheres to the well defined and portable semantics Duff
> mentioned.

It just seems wilfully perverse, that's all. What other bizarre
decisions would the designers have made? Why choose software/hardware
that does that if there's a more reasonable alternative?

> My point was that these kinds of arrays are promoted as a better way
> to access individual bits than logical operations and shifts. But
> unless you have some definition of which bit "Y (Y'First)" accesses,
> they're not useful for that because they're extremely compiler
> dependent, while logical operations and shifts are portable across
> compilers and targets. Some sort of correspondence with the bit
> numbers used in record representation clauses would be OK.

I don't quite see the last sentence -- typically the bit numbering
depends on the endianness, very annoying -- not sure whether new Ada
sorts that one out.

For the rest, I quite agree; _much_ easier to map the C condition 
"(x & (1<<2))" to "(X and 2#00000100#) /= 0".