Fixed close handles with Alpine and MUSL

[dchansen]

The MUSL C library does not provide the close_range in unistd.h and Alpine linux therefore cannot compile boost::process::v2, at close_handles.ipp fails to compile.

This PR adds a check to only use close_range if we are using GLIBC.

[klemens-morgenstern]

Looks good, will merge if all tests pass.

[klemens-morgenstern]

Ok, I don't think this is correct, as it doesn't handle libc++. I'll need to add a MUSL test to CI.

[t43rr7]

Hello, @dchansen, I've encountered the same issue, but with glibc. The close_range checks here would be a bit more complicated. I can include your changes in my patch, so it could resolve 2 issues with single PR.

[dchansen]

@t43rr7 That looks good to me. Could you make a PR for my branch? then I will merge it in.

[t43rr7]

@dchansen I've opened PR to your branch. Looks like your branch's HEAD a bit behind origin develop, so there're a bit more changes that expected. Anyway, the most interesting diffs here. I've also added raw system call support which should work even on MUSL library.

Please, verify this changes on your setup, because I have no possibility to do it fast (with MUSL). Thanks!

UPD: Same changes can be verified in this PR

[t43rr7]

@klemens-morgenstern please, look at the new changes

[klemens-morgenstern]

Can't we just put a close_range wrapper around the syscall into a detail namespace and use using ::close_range otherwise? That way we don't need to reimpl close_all.

[t43rr7]

Do you mean something like "code below"?

int close_range_wrapper(...) {
#ifdef HAS_CLOSE_RANGE
    return ::close_range(...);
#elif defined(HAS_CLOSE_RANGE_SYSCALL) 
    return ::syscall(SYS_close_range, ...);
#endif
}

I like this idea :) But this PR is not from my fork, I can do it in similar opened PR. How do you feel about this?

[t43rr7]

@klemens-morgenstern done in this commit

Please, see PR

[t43rr7]

@klemens-morgenstern are there any news?