Rich Felker

I don't see what the motivation for adding Linux 2.6.16 is. There is no reason to want it to match the kernel version used on the host you're compiling on...

In "The motivation is to create a toolchain that targets CentOS 5", do you mean motivation for using old kernel headers? This is not necessary. New kernel headers work perfectly...

I think this is a duplicate but I don't see the particular issue it's a duplicate of so maybe it's just related. The symlink there is a symlink to be...

Strictly speaking libc++ could be used with GCC too, but I don't know what's involved to get it to work. Building & using particular software with musl is outside the...

Yes, I am. It's hard to bootstrap (you need a host compiler that supports Ada to begin with) which means you can't use an existing lineage of MCM toolchains to...

For `getentropy`, there is no portable implementation or safe fallback that widely works. What we can easily do though is just declare it ourselves (`int getentropy(void*, size_t);`) so as not...

It's intended that the upstream headers be usable with the patches applied by mcm (if any). If there are problems with that, they should be both reported upstream (kernel) and...

The Linux version here has nothing to do with the kernel version you intend to run. It's just the one to pull headers from, if you want your output toolchain...

This was a known bug in Linux from that era, that it was not future-proof against compiling with GCC newer than the newest it knew about at the time. In...

I don't think straight setting of CXXFLAGS here is right because that will either override, or be overridden by, any user settings of CXXFLAGS. In particular possibly-desirable things like CXXFLAGS="-g0...