https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90027
--- Comment #4 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> No, and AFAIK the testcase has always worked on strict-alignment platforms.
To be more explicit:
typedef struct
{
short a;
long b;
short c;
short d;
} __attribute__ ((packed)) S;
S s;
S *p = &s;
long *b = &s.b;
p->b is OK because it's (*p).b and p contains a valid address for its type, but
*b is not OK because b doesn't contain a valid address for its type.
