1.0.84

Important

This release has been yanked, due to the Apple deployment target changes causing unintentional widespread breakage for building C++ code on the target (for projects without an explicit deployment target set). A release without these issues will be made in the near future.

Sorry, and thanks.

Changes

This isn't a complete list, but I've tried to cover everything that is either significant, or could cause problems for people using cc in configurations not covered by our tests.

• The MSRV has been updated to 1.53.0 (#894)
• A completely overhauled implementation of the jobserver protocol (used under feature = parallel), which should be more efficient. (#889).
• Better handling of assemblers on windows targets, especially armasm[64].exe (#867, #868, #869, #857)
• A few fixes for incorrect results from is_flag_supported (#839, #886)
• Support for the *-apple-tvos and *-apple-tvos-sim targets (#881, #704)
• Improved handling of the deployment target on Apple targets, including use of rustc --print deployment-target when available (#848, #872)
• We no longer bundle bitcode when compiling for Apple targets, as they've deprecated it (#812)
• A new remove_flag function to remove flags which have been added with add_flag (#885)
• Changes to how clang and gcc are distinguished, and how -m32/-mx32/-m64 flags are passed (#709)
• Minor adjustments to how -isysroot is passed on Apple targets (#703) and how MSVC version is detected on Windows (#696).

As always, please file bugs if you hit issues. cc is used in many more ways than we can possibly test (especially when cross-compiling or using tier-3 platforms).

Thanks.

New Contributors

• @BlackHoleFox made their first contribution in #848
• @youknowone made their first contribution in #704
• @Arc-blroth made their first contribution in #705
• @osiewicz made their first contribution in #878
• @RubixDev made their first contribution in #839
• @lcruz99 made their first contribution in #881
• @raisfeld-ori made their first contribution in #885

Full Changelog: 1.0.83...1.0.84

1.0.83

What's Changed

• Add method for specifying C/C++ standard version by @hikari-no-yume in #761
• Stop using feature "libc/std" by @kadiwa4 in #860

New Contributors

• @hikari-no-yume made their first contribution in #761
• @kadiwa4 made their first contribution in #860

Full Changelog: 1.0.82...1.0.83

1.0.82

This release corrects a deadlock and potential performance regression present since 1.0.80, and contains no other changes. See #849 for details.

1.0.81

This release exists to fix a bug introduced in 1.0.80 where we would panic if non-UTF8 were written to stderr by the C compiler. See #842 and #841 for more info.

1.0.80

What's Changed

• Bump MSRV to 1.46.0 by @thomcc in #781
• Make windows_registry::VsVers non_exhaustive by @NobodyXu in #783
• Prioritize RUSTC_LINKER over target->cross-compile-prefix table. by @dot-asm in #767
• If a file path references a parent directory, transform it to ensure we keep inside the OUT_DIR by @jmesmon in #786
• Speedup Build::clone: Use Arc instead of String/PathBuf/OsString by @NobodyXu in #782
• Fix --target cflag on FreeBSD when compiling against clang by @mqudsi in #785
• Append freebsd-version to all --target=*-freebsd if executed on FreeBSD. by @dot-asm in #788
• Update some doc comments by @mqudsi in #792
• Prevent cloning on every get_xxx() call by @mqudsi in #793
• Change approach for native and cross-compilation with clang++ on FreeBSD by @mqudsi in #794
• Default to llvm-ar when compiling for wasm. by @dot-asm in #657
• Add Loongarch64 support by @zhaixiaojuan in #637
• Refine CUDA documentation. by @dot-asm in #799
• Improve docs for the C++ usage by @SallySoul in #600
• Link against libc++ on AIX by @bzEq in #803
• Speedup CI by @NobodyXu in #805
• Enable dependabot for GitHub Action Workflows by @NobodyXu in #806
• Be more stringent about handling Android NDK. by @jfgoog in #808
• Optimize Build::compile_objects: Only spawns one thread by @NobodyXu in #780
• add a function to populate flags from an arbitrary environment variable by @nagisa in #818
• Use gnu -o flag for obj out path instead of -Fo when using gcc & g++ on Windows by @Spencer1O1 in #820
• Replace os_pipe with direct implementation by @NobodyXu in #822
• Optimize Build::print: Avoid unnecessary heap alloc by @NobodyXu in #824
• Optimize Error::new: Avoid unnecessary heap alloc by @NobodyXu in #823
• Optimize cc::Build::try_compile: Reuse PrintThread by @NobodyXu in #817
• Remove outdated doc about build parallelism by @NobodyXu in #827
• Use correct ABI on NetBSD/riscv64, and add target entry for same. by @he32 in #815
• Fix Build::compile_objects deadlock on parallel by @NobodyXu in #829
• Cleanup mod com, registry, setup_config and winapi by @NobodyXu in #828
• Optimize Build::get_out_dir: Return Cow<'_, Path> by @NobodyXu in #831
• Optimize Build::env_cache: Store Option<Arc<str>> as key by @NobodyXu in #832
• Optimize Build::compile_objects on feature parallel by @NobodyXu in #833
• Disambiguate Windows run-times when using clang to compile. by @dot-asm in #825
• Harmonize README.md with reality. by @dot-asm in #835
• Replace windows-sys dependency with bindings generation by @NobodyXu in #837
• Handle x86_64h-apple-darwin target by @thomcc in #840
• Bump version to 1.0.80 by @thomcc in #834

New Contributors

• @zhaixiaojuan made their first contribution in #637
• @SallySoul made their first contribution in #600
• @bzEq made their first contribution in #803
• @jfgoog made their first contribution in #808
• @Spencer1O1 made their first contribution in #820
• @he32 made their first contribution in #815

Full Changelog: 1.0.79...1.0.80

1.0.79

Version 1.0.79

Changelog

• The riscv32imc-esp-espidf target is now supported (#776)
• Build::is_flag_supported now checks the exit status of the compiler in addition to stderr output, improving reliability (#757)
• When building for the aarch64-apple-ios-sim target -arch arm64 is now correctly passed as two arguments instead of one (#759)
• Improved control over ar and ranlib (#763)
  • New functions to return the ar (Build::get_archiver) and ranlib (Build::get_ranlib) binaries used for this build have been added.
  • Several related environment variables involving the archiver and ranlib variant (ARFLAGS, RANLIBFLAGS, AR, and RANLIB) are now respected automatically if not overridden.
  • Some other changes too, see #763 for details.
• We will now correctly locate msbuild under VS2022 (#773)

---

Additionally, a number of smaller changes which shouldn't impact users have been made (in particular #769 and #777); see the commit history for complete details.

Thanks to everybody who contributed to this release!

1.0.78

Version 1.0.78

Changelog

• Now, only .asm files are passed to masm on windows targets, bringing things back in line with how cc-rs<1.0.77 handled it (#755).
• Absolute paths in source are now mapped as relative in OUT_DIR (#684)
• Several improvements were made to CUDA support (#712)
• llvm-lib.exe is now used instead of lib.exe when clang-cl is used in an MSVC environment (#758)

---

Thanks to everybody who contributed to this release!

1.0.77

Version 1.0.77

Changelog

• Added a new Build::asm_flag function, which allows providing flags that are only used when compiling assembly files (for example, if your C compiler rejects flags passed to the assembler when not used as an assembler). (#752)
• Ensure that the version of DWARF debuginfo we emit is consistent with what rustc uses on the given target. (#694)

---

Thanks to everybody who contributed to this release!

1.0.76

Version 1.0.76

Changelog

• When compiling with clang-cl, we now only use -- to separate flags/options from input files when not using the assembler, which doesn't support that option. This is a fix for a regression introduced in 1.0.74 (by #514) (present in 1.0.75 as well). (#747)

---

This is essentially a bugfix release, so that's all!

1.0.75

Version 1.0.75

Changelog

• When Rust debuginfo is enabled, debuginfo is passed to MSVC assemblers (#742)
• When locating an ar executable, *-ar is now preferred to *-gcc-ar (#741)
• On MSVC 15+, the atlmfc directory is now correctly located (#699)

---

Additionally, a number of smaller changes which shouldn't impact users have been made; see the commit history for complete details.

Thanks to everybody who contributed to this release!