Git installs its own Mingw and Msys2 stuff but mostly compiled for a Mingw environment so they consume Windows paths natively instead of using MSYS2/Cygwin path conversion. That's why when you have mixed PATH variable all hell breaks loose with Git.
Doesn't it come with `pacman` too?
MSYS2 is there to just provide the basics so you can develop programs that are Windows native but use some of the tools that have really strong Unix dependence like shells or Make. They depend on the existence of syscalls like `fork` or forward slash being the directory seperator.
Without Cygwin enabling the path, it wouldn't be possible to build GCC for Windows without completely changing its build system. It would be a DOA fork while Mingw and PE32+ support is a part of GCC nowadays.
The nice and genius part of MSYS2 is that it is there to primarily encourage you to develop native Windows software that has better cross-platform behavior rather than Cygwin alone. If Microsoft made a better, free of charge C compiler in early 2000s that is adhering to the standards better, we wouldn't probably need Mingw to build cross-platform apps. Now MSVC is still free of charge for only open source and individuals.
[1] "Cygwin POSIX emulation engine", https://packages.msys2.org/base/msys2-runtime [2] https://github.com/msys2/MSYS2-packages/tree/master/msys2-ru...
It gives you a *nix-like shell/dev environment and tools, but you build native software that runs on Windows systems that don’t have or need to have all/parts of MSYS2/Cygwin installed.
I built a network daemon using the MSYS2 CLANG64 environment and llvm toolchain on Windows 10.
Windows 7 x64 users could download the compiled single-file executable and run it just fine, so long as they installed Microsoft’s Universal C Runtime, which is a free download from Microsoft’s website.
I get your point. Although my point is that there is actually zero need for MSYS at all for this, even as a developer, and especially not with the 'CLANG64' environment. These binaries themselves are built to run in the MSYS2 environment This is how I cross-compile from Windows... to Windows with LLVM-MinGW[1]:
> (gci Env:PATH).Value.Split(';') | sort
> clang-21.exe --version
clang version 21.1.2 (https://github.com/llvm/llvm-project.git b708aea0bc7127adf4ec643660699c8bcdde1273)
Target: x86_64-w64-windows-gnu
Thread model: posix
InstalledDir: C:/Users/dpdx/AppData/Local/Microsoft/WinGet/Packages/MartinStorsjo.LLVM-MinGW.UCRT_Microsoft.Winget.Source_8wekyb3d8bbwe/llvm-mingw-20250924-ucrt-x86_64/bin
Configuration file: C:/Users/dpdx/AppData/Local/Microsoft/WinGet/Packages/MartinStorsjo.LLVM-MinGW.UCRT_Microsoft.Winget.Source_8wekyb3d8bbwe/llvm-mingw-20250924-ucrt-x86_64/bin/x86_64-w64-windows-gnu.cfg
I'm certain I haven't misunderstood the point of MSYS2's CLANG64 and other environments.
> These binaries themselves are built to run in the MSYS2 environment
I'm not sure if you're referring to the toolchain binaries or the binaries one produce's with them.
The CLANG64, etc. environments are 100% absolutely for certain for building software that can run outside of any MSYS2 environment!
You can, of course, build executables specifically intended to run inside those environments, but that’s not the primary use case.
> (gci Env:PATH).Value.Split(';') | sort
I don't want to use PowerShell or Cmd.exe when doing dev stuff on Windows. I want to do CLI work and author scripts in and for modern Bash, just like I would for Linux and macOS. I want to write Makefiles for GNU make, just like...
Now, sometimes there are bumps and sharp edges you have to deal with via `if [[ -v MSYSTEM ]]; then`, similar in Makefile, cygpath conversion, template/conditional code in sources, and so on. But that's a small price to pay, from my perspective, for staying in the same mental model for how to build software.
There. I think that sums it up.