Re: [musl] Alignment attribute in headers

[Markus Wichmann <nullplan@gmx.net>]

Am Sun, Apr 21, 2024 at 12:23:58PM +0200 schrieb Jₑₙₛ Gustedt:
> Markus,
>
> on Sun, 21 Apr 2024 09:38:38 +0200 you (Markus Wichmann
> <nullplan@gmx.net>) wrote:
>
> > Am Sun, Apr 21, 2024 at 09:16:05AM +0200 schrieb Jₑₙₛ Gustedt:
> > > Since this is unified starting with C23 and I think we morally
> > > should have C conformance first and fallbacks only if imperatively
> > > needed I would go for
> > >
> >
> > Ugh. Let the bike shedding begin. I will tell you that moral arguments
> > about software don't make a lot of sense to me, though.
>
> I am not a native speaker,

Me neither, but I do spend an embarassing amount of time on YouTube.

> but I think this is generally used as
> figure of speech for "there got reasons to do something".
>

I have only ever heard this phrase in technical discussions when the
pundit had run out of technical arguments. And usually only for personal
preferences. I personally also prefer standard solutions over
implementation-specific hacks (like the attribute syntax), but I cannot
blind myself to the requirements of the implementation.

Plus, I dislike conditional inclusion in general, but in this case, we
do require all three versions to be present anyway and are merely
arguing about precedence. Well, it won't make any difference to a human
reader, as they will have to process all seven lines anyway, and the
compiler doesn't care for any line except the one that does get included
in the end. I would much prefer to dispense with the entire attribute,
but that would break ABI. So the least bad form of this very bad code
would be my proposal.

BTW, in case it got missed earlier: The explicit padding is still
necessary for compat with non-GNU pre-C11 compilers! At least the layout
will be compatible this way.

Ciao,
Markus