Cygwin Runtime

Project

• Cygwin may refer both to:

  • Cygwin runtime (cygwin1.dll), required for aarch64-pc-cygwin GNU toolchain, available at https://www.cygwin.com/git.html, which provides the necessary POSIX APIs and the standard library implementations for building and running POSIX-based software on Windows,

  • or the whole distribution of POSIX-based software (https://www.cygwin.com/packages.html) compiled to run on Windows available at https://cygwin.com/packaging/repos.html.

• This project’s scope is limited to Cygwin runtime (cygwin1.dll) only. Porting of individual Cygwin packages hasn’t started yet as the aarch64-pc-cygwin GNU toolchain and the runtime are not available upstream and it would deserve separate project pages for individual packages or some reasonable groups of packages.

• The Cygwin runtime source code consists for two main sub-projects, newlib and winsup:

  • The newlib sub-project contains implementation of standard C library (libc), and implementation of standard C mathematical library implementation (libm).

  • The winsup project is where the most of the magic of POSIX APIs and Cygwin environment implementations happen.

  • Fun fact: The newlib is a least 26 years old fork of GCC codebase (that far the git history goes) being occasionally synchronized to the upstream.

Risks

• Windows on Arm64 require binaries to use https://learn.microsoft.com/en-us/cpp/build/reference/highentropyva-support-64-bit-aslr?view=msvc-170 and https://learn.microsoft.com/en-us/cpp/build/reference/dynamicbase?view=msvc-170 which is known to be poorly implemented in Cygwin (e.g., https://cygwin.com/pipermail/cygwin/2021-February/247922.html).

• Porting POSIX fork implementation will be the biggest challenge as it requires address space and stack manipulation and is affected by the high-entropy ASLR and dynamic base issues as well.

Status

• Experiment attempting to build aarch64-pc-cygwin GNU toolchain and the cygwin1.dll runtime capable of building functional Hello World! executable has been conducted within https://github.com/Windows-on-ARM-Experiments/mingw-woarm64-build/pull/123, the insights gained during that experiment has been documented at https://github.com/Windows-on-ARM-Experiments/mingw-woarm64-build/issues/183.

Next Steps

• The Hello World! experiment did lot of hacks, like commenting out the code that is not needed to run the binary, that needs to be implemented in a way that would be acceptable for upstreaming.

• Then, the next target should be compilation of bash.exe executable. It might show up that this goal si too stretched and some less ambitious target should be chosen instead.