Brutkey

Alexander Monakov
@amonakov@mastodon.gamedev.place

@siguza@infosec.space @jann@infosec.exchange unfortunately it's even worse if you're concerned with conformance, because the definition of offsetof states:

The type and member designator shall be such that given
static type t;
then the expression &(t.member-designator) evaluates to an address constant.

which completely rules out offsetof(foo, arr[N]) when N is not a compile-time constant, and admits only zero for a constant N.

William D. Jones
@cr1901@mastodon.social

@amonakov@mastodon.gamedev.place @siguza@infosec.space @jann@infosec.exchange (Found this by chance). I feel like I must be missing some context, because max(offsetof(struct foo, arr[0]), sizeof(struct foo)) doesn't return "0" for me?

https://godbolt.org/z/ror4KP3sc

Alexander Monakov
@amonakov@mastodon.gamedev.place

@cr1901@mastodon.social @siguza@infosec.space @jann@infosec.exchange is that regarding the last sentence, "admits only zero for a constant N"? To rephrase, the text of the standard allows only zero in place of N in offsetof(foo, flexible_array_member[N]), not any other integer.

William D. Jones
@cr1901@mastodon.social

@amonakov@mastodon.gamedev.place @siguza@infosec.space @jann@infosec.exchange Yes, about that sentence. I didn't know that e.g. &foo.arr[3] doesn't evaluate to an address constant (feels like it should be allowed, unless the existence of that pointer runs afoul of the "one past the end" behavior).

But I'm not great w/ flexible array members anyway and mostly write C89 when I write C :P.

Siguza
@siguza@infosec.space

@cr1901@mastodon.social @amonakov@mastodon.gamedev.place @jann@infosec.exchange

unless the existence of that pointer runs afoul of the "one past the end" behavior

max(offsetof(struct foo, arr[0]) + N * sizeof(((struct foo*)0)->arr[0]), sizeof(struct foo))