Suggestion: remove libnuma dependency in favor of syscall
We added https://github.com/dotnet/runtime/blob/cd66519f4d879a9165c4a2cc3b7a4c804a3ff668/src/coreclr/gc/unix/numasupport.cpp in .NET runtime a while back (net8.0 time-frame) to remove this dependency.
libnuma is licensed under GPL/LGPL 2.0 which may not be problematic, but it's just about a single syscall followed by walking a directory path, so not worth the dependency pressure.
Noticed in https://github.com/dotnet/dotnet-buildtools-prereqs-docker/pull/1522; some images are still pulling libnuma for msquic.
Hi, thanks for opening this. To clarify, is this something that you did in .NET and recommend us to also do, or is doing this something that would help what you do in .NET?
Hi @guhetier, My suggestion is for this project to adopt the same NUMA implementation that .NET uses. Doing so would remove the dependency on libnuma and eliminate the need for msquic consumers to install it separately.
Thanks for the clarification. Removing a build time dependency could be interesting. It is unlikely that we can get to it soon though, but if that is of interest to you, please feel free to open a PR, contributions are welcome!