Implement u128/i128 support #13
Conversation
|
r? @japaric Also please don't forget to publish a new version (0.2.1?). |
.travis.yml
Outdated
| @@ -9,13 +9,13 @@ env: TARGET=x86_64-unknown-linux-gnu | |||
| matrix: | |||
| include: | |||
| - env: TARGET=i686-unknown-linux-gnu | |||
| - env: TARGET=x86_64-unknown-linux-gnu | |||
| - env: TARGET=x86_64-unknown-linux-gnu RUST_BETA=1 | |||
There was a problem hiding this comment.
nit: the TRAVIS_RUST_VERSION variable will contain the string 'beta' or 'nightly' when 'rust: beta' or 'rust: nightly' is used here. You can use that var instead of defining a new one.
src/lib.rs
Outdated
| u32 => i128, u128; | ||
| u64 => i128, u128; | ||
| usize => i128, u128; | ||
| u128 => f32, f64, u128; |
There was a problem hiding this comment.
I'm not sure if u128 -> f32 can be really be considered a promotion.
f32::MAX prints 340282350000000000000000000000000000000.
u128::MAX prints 340282366920938463463374607431768211455
So u128::MAX is larger than f32::MAX or IOW u128::MAX doesn't fit in f32. IMO, this conversion should return Error::Overflow. However, rustc (and gcc) seem to be fine with the u128::MAX as f32 operation; it returns inf. That seems wrong to me -- that's not loss of precision; that's a rounding operation with a rounding error of inf.
There was a problem hiding this comment.
I've modified it to make it pass now.
|
The review should be addressed now. |
|
Thanks @est31. @homunkulus r+ |
|
📌 Commit ca5aaef has been approved by |
|
☀️ Test successful - status-travis |
Fixes #12
Unfortunately only commented out tests, due to BurntSushi/quickcheck#162