Re: [musl] Alignment attribute in headers

[Alexander Monakov <amonakov@ispras.ru>]

[-- Attachment #1: Type: text/plain, Size: 1399 bytes --]


On Sun, 21 Apr 2024, Thorsten Glaser wrote:

> Michael Forney dixit:
> 
> >Something like
> >
> >#if __STDC_VERSION__ >= 201112L
> >/* use _Alignas */
> >#elif defined(__cplusplus) && !defined(__GNUC__)
> >/* use alignas */
> >#else
> >/* use __attribute__((__aligned__(N))) */
> >#end
> 
> Something I noticed recently while doing m68k alignment work:
> 
> The C++ alignas is UB if the specified alignment is smaller
> than what the structure would normally have, so adding cautious
> alignments can explode in one’s face. ☹ Not only is this really
> stupid, but makes it not generally usable, too.
> 
> GCC’s attribute, in contrast, (without __packed__) just gets
> ignored for those cases.

This is inaccurate: you can use the attribute to decrease alignment
of scalar types (but you do need __packed__ when attaching the attribute
to a struct). As I recall, GCC documentation used to be misleading
or wrong about this, but the current wording is fairly clear to me:

    When used on a struct, or struct member, the aligned attribute can
    only increase the alignment; in order to decrease it, the packed
    attribute must be specified as well. When used as part of a typedef,
    the aligned attribute can both increase and decrease alignment, and
    specifying the packed attribute generates a warning. 

    https://gcc.gnu.org/onlinedocs/gcc/Common-Type-Attributes.html

Alexander