chore: release v0.2.3+llvm-a284bdb31146

[github-actions]

🤖 New release

• rustc_apfloat: 0.2.3+llvm-462a31f5a5ab -> 0.2.3+llvm-a284bdb31146

Changelog

0.2.3+llvm-a284bdb31146 - 2026-09-20

Other

• Cache LLVM downloads separately
• Switch Cargo package from exclude to include
• Add tests for ONE similar to what we have for ZERO
• add const ONE: ... to Float
• Add more detailed documentation for 8-bit float types
• Make public enums #[non_exhaustive] where relevant
• Bump LLVM commit to llvm/llvm-project@a284bdb31146
• Switch min and max to IEEE 754-2019 {minimum,maximum}Number semantics
• Update references to old versions of IEEE-754
• Add getExactLog2Abs
• Add get_exact_log2
• Add TF32
• Add is_sign_negative and is_sign_positive
• Bump LLVM commit to llvm/llvm-project@687bd77e2c26
• Add IeeeFloat::is_representable_as_normal_in
• Add bits_to_f8_to_bits and f8_to_bits_to_f8 tests
• Add E4M3B11FNUZ
• Bump LLVM commit to 71f1ea2c15bbc5b604e20d71191a16d57d106dc0
• Add Float::make_quiet method
• Resolve a test failure from the new 8-bit float types
• Fix the build on Apple platforms
• Improve logic for negative zero handling
• Add NaN-in-negative-zero formats
• Resolve the {min,max}_value lint and silence a Cargo naming lint
• Make StatusAnd::unwrap #[inline] and #[track_caller]
• Fix -xux to -eux in the download script
• Enable checking status on host floats
• Add Float::max_int_bits
• Replace mix/max combo with Ord::clamp
• Bump LLVM commit to llvm/llvm-project@e60b91d
• Skip decoding README.txt files
• Add a subcommand for generating corpus input
• Create a justfile for invoking common commands
• Change the save directory to fuzz/runs rather than target
• Add an assembly host implementation on x86
• Adjust print and exit options
• Download LLVM after the cache action
• Move host floats to a separate module
• Move Op and Airity into main.rs
• Extract FpKind out of the macro
• Move main toward the top of the file
• Ignore ICE output in git
• Move the bruteforce entrypoint to its separate module
• Port exhaustive tests to the new fuzz setup
• Switch from panics to an error that we can panic from later
• Skip running cxx versions of f8e4m3fn fma
• Improve flexibility of the fuzz binary
• Extract brute force to a separate module
• Download and cache LLVM for the fuzz build
• Build and test all crates in the workspace
• Use the 2024 edition
• Check in generated Rust source
• Replace generated C++ with a source file
• Add a check-cfg for cfg(fuzzing)
• Remove comments about u128 ABI issues
• Replace shelling out in build.rs with Rust code
• Check in relevant files rather than creating
• Add caching for the test job
• Set a MSRV of 1.65 and test it in CI
• Bump actions/checkout to the latest version (v6)
• Pin Ubuntu versions to 24.04
• Add a timeout for all jobs
• Add a script for downloading the correct LLVM version
• Deny warnings and enable backtraces
• Also run on pushes to main
• Invoke bash directly rather than using sh
• Don't include .github in the shipped crate
• clarify comments for min/max operations

---

This PR was generated with release-plz.