#1388. FFI test failures fixed

[sgrekhov]

No description provided.

[eernstg]

I'll take this opportunity to learn a bit more about how various number types are handled using FFI, so I can't yet comment on the PR as such.

Just one thing: I assume that the platform is the vm (no other platforms support FFI? yet?), and in that case it might be useful to consider int.toUnsigned: (-42).toUnsigned(32) yields the 64 bit zero extending interpretation of the bit pattern which is the 32 lowest bits of the representation of the receiver, so that's a way to get hold of (or check) those big numbers that arise because of signed/unsigned/32-bit/64-bit and whatnot conversions.

[eernstg]

OK, I think this one is done. Landing.

[eernstg]

LGTM

[dcharkes]

Just one thing: I assume that the platform is the vm (no other platforms support FFI? yet?), and in that case it might be useful to consider int.toUnsigned: (-42).toUnsigned(32) yields the 64 bit zero extending interpretation of the bit pattern which is the 32 lowest bits of the representation of the receiver, so that's a way to get hold of (or check) those big numbers that arise because of signed/unsigned/32-bit/64-bit and whatnot conversions.

That could indeed simplify this test. I learn something new every day!