Check the cipher flags to see if the cipher supports aead

[danielwestendorf]

Supports checking if a cipher supports aead, with an authenticated? helper method.

[danielwestendorf]

Need is explained here.

[danielwestendorf]

The darwin CI suite seems to be failing because aes-128-gcm isn't an available cipher, however, when run locally on my macOS system (with Crystal installed via homebrew) it is available and the spec passes.

[ysbaddaden]

CI seems to run on macOS sierra, which probably has an old OpenSSL version (maybe 0.9.8zh).

Do you know which OpenSSL versions support aes-128-gcm?

[danielwestendorf]

@ysbaddaden I think you're on to something there. I'm trying different macOS/OSX versions to see what the oldest version this change would require is, as I can't seem to find any definitive information about when aes-128-gcm was added to OpenSSL.

[danielwestendorf]

It looks like GCM ciphers aren't supported on macOS v10.12 and lower. When I updated the CI config to xcode: "9.3.0" the specs passed 💯.

For now, I've reverted to xcode: "9.0" as it currently is, and simply rescued the ArgumentError in the spec. This is definitely a code smell, but probably better than bumping the CI macOS version we test against. Once it makes sense for Crystal's CI macOS to v10.13+ the rescue can be removed.

[RX14]

I'd like @bcardiff's thoughts on rescuing ArgumentError in the specs before merge.

[straight-shoota]

Maybe add a check that its actually an expected ArgumentError by verifying the error message. It should be re-raised if it doesn't match.

[danielwestendorf]

Could definitely do that, however if the ArgumentError is originating from anything other than the cipher not being available, it'd raise on the next invocation (line 58/59).

[bcardiff]

I'm late to the party. The changes are fine. At most I would've try to only accept the exception as a success if the error message is certain, or wrap the whole block of the it in a feature check if.