Miscompilation under target-cpu >= haswell

[rustcrypto]

Miscompilation under target-cpu >= haswell

[newpavlov]

@tarcieri
Should we use rustsec to notify users of this crate?

[newpavlov]

Looks like the problem originates in the aes-soft crate. I can reproduce it by simply running cargo test --release with the reported target-cpu flags. IvyBridge and later have AES-NI, so they use aesni. But interestingly enough running:

RUSTFLAGS='-C target-cpu=ivybridge' cargo test --release

for aes-soft does not result in the miscompilation.

UPD: Ah, you already use aes-soft in your example.

Can you create a Rust issue and link this bug there? I will keep it open until the issue gets resolved.

[tarcieri]

@newpavlov yes, sounds good to me

[anonymous6727]

The LLVM and Rust issues were resolved.

I've tested the code and commands from the opening post, and they all succeed for me.

• aes-soft v0.3.3
• nightly-x86_64-unknown-linux-gnu AND stable-x86_64-unknown-linux-gnu

So I think this can be closed.