Fix undefined-behavior on MacOSX structs in stdbuilds

[gnzlbg]

Some MacOSX structs have an incorrect layout that results in undefined behavior. This is because on x86_64 the MacOSX kernel headers define these using #pragma pack 4.

This PR fixes their layout using repr(packed(4)) . Since it is only available on nightly, it is only enabled for stdbuilds .

[rust-highfive]

r? @alexcrichton

(rust_highfive has picked a reviewer for you, use r? to override)

[gnzlbg]

@alexcrichton this would require updating the stable Rust version used in CI. Is there a policy about that?

[alexcrichton]

What would it require updating to?

[gnzlbg]

I think attr_literals but I'll bisect the first version that works.

[gnzlbg]

@alexcrichton we would need to update from 1.0.0 to 1.13.0 meaning newer libc versions wouldn't compile with older compilers anymore. This is pretty much a backwards incompatible change, might need to change libc from 0.2 to 0.3 :/

[gnzlbg]

An alternative could be tuning repr(packed) syntax to something that is compatible with 1.0, maybe instead of repr(packed(N)) we could add repr(packed1), repr(packed4), etc.

[gnzlbg]

@alexcrichton I can also put these two structs into two different files, and conditionally include one or the other depending on the unstable cargo feature.

[alexcrichton]

Nah Rust 1.13.0 is ancient at this point (aka pre-stable-Serde) so I doubt anything before that is still in use. If you'd like to raise the minimum version in CI to 1.13.0 I think we'll be good to go!

[gnzlbg]

@alexcrichton done !

[alexcrichton]

@bors: r+

[bors]

📌 Commit 77837a0 has been approved by alexcrichton

[bors]

⌛ Testing commit 77837a0 with merge 69769fb...

[bors]

☀️ Test successful - status-appveyor, status-travis
Approved by: alexcrichton
Pushing 69769fb to master...