[Issue 23573] std.bitmanip.bitfields doesn't respect native endianness

[d-bugmail at puremagic.com]

https://issues.dlang.org/show_bug.cgi?id=23573

--- Comment #4 from Iain Buclaw <ibuclaw at gdcproject.org> ---
(In reply to johanengelen from comment #3)
> (In reply to elpenguino+D from comment #2)
> > 
> > D doesn't need two undefined flavours of bitfields. Some people find it
> > useful to be able to read bitfields from files or the network regardless of
> > cpu architecture.
> 
> OK, fair point.
> But then the bugreport should be rewritten without comparing it to C
> bitfields, and instead specify exactly the bit layout as desired in bytes.
> It's not clear to me what is meant, because (I think) the OP says that the
> order of bits _within_ a byte is reversed depending on endianness, which
> does not give the architecture independent behavior that you want.
Well, the bug is that it *nearly* ended up in Phobos without checking whether
it actually worked.

https://github.com/dlang/phobos/pull/8478/files#diff-0f4bb61407d729d1f6d76e25e15eb84b89ceae014ccc887149c53fb2290a29ccR377-R392

std.numeric also uses bitfields to extract the sign, exponent, and significand
out of a float and double too.

    mixin(bitfields!(
        T_sig, "significand", precision,
        T_exp, "exponent"   , exponentWidth,
        bool , "sign"       , flags & flags.signed ));

It is not obvious to the observer that the mixin will generate code to match
the layout for both big and little endian (depending on which is in effect).

--