On 09/06/2016 11:11 AM, Florian Weimer wrote:
I think this change is only safe because nothing relies on it.
max_align_t is a committee invention with no actual users.
I tried to verify that and ran grep over all the packages in
/usr/portage/distfiles. That did get a number of matches, fairly obvious
and expected ones like gcc and glibc, and several more which match only
because gnulib seems to try to provide a default, but I also found a few
real uses:
grep-2.25/lib/fts.c:
/* Align the allocation size so that it works for FTSENT,
so that trailing padding may be referenced by direct access
to the flexible array members, without triggering undefined
behavior
by accessing bytes beyond the heap allocation. This
implicit access
was seen for example with ISDOT() and GCC 5.1.1 at -O2.
Do not use alignof (FTSENT) here, since C11 prohibits
taking the alignment of a structure containing a flexible
array member. */
len += alignof (max_align_t) - 1;
len &= ~ (alignof (max_align_t) - 1);
libvpx-v1.3.0/nestegg/halloc/src/halloc.c:
/*
* block control header
*/
typedef struct hblock
{
#ifndef NDEBUG
#define HH_MAGIC 0x20040518L
long magic;
#endif
hlist_item_t siblings; /* 2 pointers */
hlist_head_t children; /* 1 pointer */
max_align_t data[1]; /* not allocated, see below */
} hblock_t;
Bernd
