gmacros: Don’t use __alignof__ in G_ALIGNOF implementation

It has different semantics from _Alignof and our G_STRUCT_OFFSET
fallback. See the comments in the diff for details.

Signed-off-by: Philip Withnall <withnall@endlessm.com>

https://gitlab.gnome.org/GNOME/glib/issues/1055
This commit is contained in:
Philip Withnall 2018-12-21 13:09:57 +00:00
parent ea0da960ab
commit b7d2eeed17
2 changed files with 10 additions and 3 deletions

View File

@ -1870,11 +1870,16 @@
* G_ALIGNOF
* @a: a type-name
*
* Return the minimum alignment required by the platform ABI for values of the given
* Return the minimal alignment required by the platform ABI for values of the given
* type. The address of a variable or struct member of the given type must always be
* a multiple of this alignment. For example, most platforms require int variables
* to be aligned at a 4-byte boundary, so `G_ALIGNOF (int)` is 4 on most platforms.
*
* Note this is not necessarily the same as the value returned by GCCs
* `__alignof__` operator, which returns the preferred alignment for a type.
* The preferred alignment may be a stricter alignment than the minimal
* alignment.
*
* Since: 2.60
*/

View File

@ -406,12 +406,14 @@
#endif
/* Provide G_ALIGNOF alignment macro.
*
* Note we cannot use the gcc __alignof__ operator here, as that returns the
* preferred alignment rather than the minimal alignment. See
* https://gitlab.gnome.org/GNOME/glib/merge_requests/538/diffs#note_390790.
*/
#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 201112L && !defined(__cplusplus)
#define G_ALIGNOF(type) _Alignof (type)
#elif defined(__GNUC__)
#define G_ALIGNOF(type) (__alignof__ (type))
#else
#define G_ALIGNOF(type) (G_STRUCT_OFFSET (struct { char a; type b; }, b))
#endif