1.12.0 (29 September, 2016) #

Highlights #

• rustc translates code to LLVM IR via its own “middle” IR (MIR). This translation pass is far simpler than the previous AST->LLVM pass, and creates opportunities to perform new optimizations directly on the MIR. It was previously described on the Rust blog.
• rustc presents a new, more readable error format, along with machine-readable JSON error output for use by IDEs. Most common editors supporting Rust have been updated to work with it. It was previously described on the Rust blog.

Compiler #

• rustc translates code to LLVM IR via its own “middle” IR (MIR). This translation pass is far simpler than the previous AST->LLVM pass, and creates opportunities to perform new optimizations directly on the MIR. It was previously described on the Rust blog.
• Print the Rust target name, not the LLVM target name, with --print target-list
• The computation of TypeId is correct in some cases where it was previously producing inconsistent results
• The mips-unknown-linux-gnu target uses hardware floating point by default
• The rustc arguments, --print target-cpus, --print target-features, --print relocation-models, and --print code-models print the available options to the -C target-cpu, -C target-feature, -C relocation-model and -C code-model code generation arguments
• rustc supports three new MUSL targets on ARM: arm-unknown-linux-musleabi, arm-unknown-linux-musleabihf, and armv7-unknown-linux-musleabihf. These targets produce statically-linked binaries. There are no binary release builds yet though.

Diagnostics #

• rustc presents a new, more readable error format, along with machine-readable JSON error output for use by IDEs. Most common editors supporting Rust have been updated to work with it. It was previously described on the Rust blog.
• In error descriptions, references are now described in plain English, instead of as “&-ptr”
• In error type descriptions, unknown numeric types are named {integer} or {float} instead of _
• rustc emits a clearer error when inner attributes follow a doc comment

Language #

• macro_rules! invocations can be made within macro_rules! invocations
• macro_rules! meta-variables are hygienic
• macro_rules! tt matchers can be reparsed correctly, making them much more useful
• macro_rules! stmt matchers correctly consume the entire contents when inside non-braces invocations
• Semicolons are properly required as statement delimiters inside macro_rules! invocations
• cfg_attr works on path attributes

Stabilized APIs #

• Cell::as_ptr
• RefCell::as_ptr
• IpAddr::is_unspecified
• IpAddr::is_loopback
• IpAddr::is_multicast
• Ipv4Addr::is_unspecified
• Ipv6Addr::octets
• LinkedList::contains
• VecDeque::contains
• ExitStatusExt::from_raw. Both on Unix and Windows.
• Receiver::recv_timeout
• RecvTimeoutError
• BinaryHeap::peek_mut
• PeekMut
• iter::Product
• iter::Sum
• OccupiedEntry::remove_entry
• VacantEntry::into_key

Libraries #

• The format! macro and friends now allow a single argument to be formatted in multiple styles
• The lifetime bounds on [T]::binary_search_by and [T]::binary_search_by_key have been adjusted to be more flexible
• Option implements From for its contained type
• Cell, RefCell and UnsafeCell implement From for their contained type
• RwLock panics if the reader count overflows
• vec_deque::Drain, hash_map::Drain and hash_set::Drain are covariant
• vec::Drain and binary_heap::Drain are covariant
• Cow<str> implements FromIterator for char, &str and String
• Sockets on Linux are correctly closed in subprocesses via SOCK_CLOEXEC
• hash_map::Entry, hash_map::VacantEntry and hash_map::OccupiedEntry implement Debug
• btree_map::Entry, btree_map::VacantEntry and btree_map::OccupiedEntry implement Debug
• String implements AddAssign
• Variadic extern fn pointers implement the Clone, PartialEq, Eq, PartialOrd, Ord, Hash, fmt::Pointer, and fmt::Debug traits
• FileType implements Debug
• References to Mutex and RwLock are unwind-safe
• mpsc::sync_channel Receivers return any available message before reporting a disconnect
• Unicode definitions have been updated to 9.0
• env iterators implement DoubleEndedIterator

Cargo #

• Support local mirrors of registries
• Add support for command aliases
• Allow opt-level="s" / opt-level="z" in profile overrides
• Make cargo doc --open --target work as expected
• Speed up noop registry updates
• Update OpenSSL
• Fix --panic=abort with plugins
• Always pass -C metadata to the compiler
• Fix depending on git repos with workspaces
• Add a --lib flag to cargo new
• Add http.cainfo for custom certs
• Indicate the compilation profile after compiling
• Allow enabling features for dependencies with --features
• Add --jobs flag to cargo package
• Add --dry-run to cargo publish
• Add support for RUSTDOCFLAGS

Performance #

• panic::catch_unwind is more optimized
• panic::catch_unwind no longer accesses thread-local storage on entry

Tooling #

• Test binaries now support a --test-threads argument to specify the number of threads used to run tests, and which acts the same as the RUST_TEST_THREADS environment variable
• The test runner now emits a warning when tests run over 60 seconds
• rustdoc: Fix methods in search results
• rust-lldb warns about unsupported versions of LLDB
• Rust releases now come with source packages that can be installed by rustup via rustup component add rust-src. The resulting source code can be used by tools and IDES, located in the sysroot under lib/rustlib/src.

Misc #

• The compiler can now be built against LLVM 3.9
• Many minor improvements to the documentation.
• The Rust exception handling “personality” routine is now written in Rust

Compatibility Notes #

• When printing Windows OsStrs, unpaired surrogate codepoints are escaped with the lowercase format instead of the uppercase
• When formatting strings, if “precision” is specified, the “fill”, “align” and “width” specifiers are no longer ignored
• The Debug impl for strings no longer escapes all non-ASCII characters