Age | Commit message (Collapse) | Author |
|
The `read_i8_into` method is useful for client code
that wants to avoid `unsafe`.
PR #147
|
|
PR #146
|
|
These were likely a result of copy & pasting example code. There is
actually an overflowing literal lint that would normally catch this, but
doc tests do not need to output lint warnings by default.
Rust 2018 made overflowing literals deny-by-default, and Rust 2015 just
recently did the same---even though it's a breaking change. That in turn
caused the doc tests to fail, and thus, we finally noticed it.
Fixes #144
|
|
|
|
In older versions of Rust, the build.rs file is not auto-detected, so we
add it to the Cargo.toml. Moreover, the existing build.rs was not
compatible with older Rust versions because of the use of eprintln! and
short-hand struct initialization. We fix that.
Closes #140, Closes #141
|
|
|
|
|
|
I had intended to deprecate these in the 1.2 release, along with the
corresponding methods on ReadBytesExt/WriteBytesExt, but missed them.
|
|
It was a copy & paste error.
Fixes #123
|
|
This adds a build.rs to byteorder that will set a conditional compilation
flag automatically if the current Rust compiler supports 128-bit integers.
This makes the i128 feature itself a no-op. We continue to allow the
feature to be supplied for backwards compatibility.
Addresses https://github.com/TyOverby/bincode/issues/250
|
|
Fixes #127
|
|
Previously the [ was just failing with:
+[ ! z mips64-unknown-linux-gnuabi64 ]
ci/script.sh: 7: [: z: unexpected operator
and thus falling through to the `cargo` version.
Fixes #134
|
|
|
|
PR #133
|
|
PR #131
|
|
|
|
|
|
|
|
Our dev-dependencies already pushed us over the minimum Rust version
supported (1.12.0), so we continue with the status quo of only testing
on stable/beta/nightly, but ensure that we continue to build on Rust
1.12.0.
|
|
Closes #129
|
|
Closes #129
|
|
|
|
PR #128
|
|
|
|
|
|
We no longer need to enable the i128 Rust feature, since it
is stabilized. We leave byteorder's i128 feature in tact to
preserve compilation on Rust 1.12.
|
|
Fixes #120
|
|
|
|
|
|
|
|
We should have done this from the start. Everything can be accomplished
via pointer casts. Benchmarks show negligible differences.
|
|
|
|
|
|
This reorganizes CI. The principle change is that we no longer run tests
on Rust 1.12. Instead, we just check that it builds.
|
|
|
|
|
|
Since nearly 2 years, there is a depcrated attribute, which can be used
https://github.com/rust-lang/rfcs/blob/master/text/1270-deprecation.md
|
|
|
|
|
|
This removes the `unsafe` annotation from
`read_{f32,f64}_into_unchecked` and deprecates the methods because the
`unchecked` name is no longer appropriate. We in turn add
`read_{f32,f64}_into` methods.
|
|
|
|
See also #105
Fixes #103
|
|
|
|
This is the mirror commit to https://github.com/rust-lang/rust/pull/46012
|
|
The similar `big_endian` tests were using an offset to read from the
end of the written `u64`, but the `native_endian` tests were reading
directly, just like the `little_endian` tests. That's of course only
correct when the target actually is little endian.
That `big_endian` offset is now sliced directly, instead of cloning into
another vector, and then this logic is also used in the `native_endian`
test, depending on the current `#[cfg(target_endian)]`.
Fixes #102.
|
|
PNG looks pixelated on Apple's retina screens.
|
|
The `repository` key only takes the username and repository name, not a full URL (the default `service` is GitHub).
The `branch` key defaults to `master` so I removed it for simplicity.
|
|
|
|
They take forever, and nobody ever looks at the output anyway. Just
making sure they compile is enough.
|
|
|