Remove Strawberry GCC compiler in Windows CI #41
Conversation
stefanv
force-pushed
the
fix-windows-ci-path
branch
from
e2fc146 to
23c4ab9
Compare
stefanv
force-pushed
the
fix-windows-ci-path
branch
from
23c4ab9 to
e3c7ff0
Compare
stefanv
force-pushed
the
fix-windows-ci-path
branch
from
e26bbee to
6b421a8
Compare
stefanv
force-pushed
the
fix-windows-ci-path
branch
from
99aaeac to
b29cfa7
Compare
stefanv
force-pushed
the
fix-windows-ci-path
branch
from
982ff3d to
789a8b5
Compare
| Remove-Item -Path C:\Strawberry -Recurse | ||
| Remove-Item -Path C:\ProgramData\Chocolatey\bin\gcc.exe | ||
| Remove-Item -Path C:\msys64 -Recurse -Force | ||
| Remove-Item -Path "C:\Program Files\LLVM" -Recurse |
There was a problem hiding this comment.
How would this work when using devpy in NumPy? Would we need to add this CI step?
There was a problem hiding this comment.
I guess it depends which compiler you want. If you're happy with clang, presumably you can just remove gcc. These steps are only necessary to simulate having the Visual Studio compilers.
There was a problem hiding this comment.
Yup -- this isn't really a devpy thing anyway. It happens for random C/C++ projects that don't have anything to do with python, also.
Using something like https://github.com/ilammy/msvc-dev-cmd is a good idea IMO, at least for GitHub actions. It guarantees that cl.exe is set up correctly and is somewhere in PATH, and if it's in PATH then it will always win as far as Meson is concerned.
There was a problem hiding this comment.
This PR now correctly simulates an identical setup in CI to compare to the case where an end user on a personal laptop has MSVC installed, but never went and installed some other gcc/clang compiler.
i.e. it simulates the common case.
|
rf #38 |
No description provided.