Re: C23 thoughts and opinions

Michael S <already5chosen@yahoo.com> writes:

On Fri, 31 May 2024 19:03:10 +0100
bart <bc@freeuk.com> wrote:
>

OK, thanks.  But I forget to ask what results you got from running the
program.  Because if I try your code, using hello.c and hello.exe as
test binary/source data, I get this output:
>
_binary_test_bi_start                    00007ff6497620e0
140695771160800 _binary_test_bi_end
00007ff649762ae0 140695771163360 _binary_test_bi_size
    00007ff509750a00 140690402380288 _binary_bin_to_list_c_start
        00007ff649762ae0 140695771163360 _binary_bin_to_list_c_end
            00007ff649762b26 140695771163430
_binary_bin_to_list_c_size               00007ff509750046
140690402377798
>
The sizes should have been 2560 and 70 respectively;  those values are
bit bigger than that.

>
That's strange.  I got expected results:
_binary_test_bi_start                    000000013FDD30C0 5366427840
_binary_test_bi_end                      000000013FDD67AC 5366441900
_binary_test_bi_size                     00000000000036EC 14060
_binary_bin_to_list_c_start              000000013FDD67AC 5366441900
_binary_bin_to_list_c_end                000000013FDD711F 5366444319
_binary_bin_to_list_c_size               0000000000000973 2419
>

However I see that you also have start and end addresses, which
sounds a much better way of determining the size.  (In that case, what
are those *size symbols for?).

>
I'd guess, *_size is here for the benefit of less smart compilers that
can not figure out that *_end - *_start is a connst expression
and can not compile code like:
>
static ptrdiff_t bar = _binary_test_bi_end - _binary_test_bi_start;

I wouldn't expect any C compiler to accept that.  It's not a constant
expression under the rules of the C standard, and there is no plausible
way to generate code for it in the context of compiling one translation
unit.  Neither gcc nor clang accepts it, even run without asking for
any standard compliance.  An implementation could allow it as an
extension, but there seems little reason to do so, because it would be
a lot of work to implement, and offers very little utility.