I think my comment from #1868 (comment) is still valid and this should be #if !canImport(Darwin).

mmm, pretty sure that FreeBSD also provides malloc_good_size ... so, I can spell it out as the inverse if you like, but I think that this is better.

I agree that we should be more general where possible. !canImport(Darwin) && !os(FreeBSD) maybe?

According to the man pages at https://man.openbsd.org/FreeBSD-12.0/malloc_good_size (which seem to support multiple OSes) no other OS incl FreeBSD has malloc_good_size even though it says FreeBSD's malloc page shows it uses jemalloc.

Is there support for os(FreeBSD) ?

Yes, there is support for os(FreeBSD)

We talked a bit off PR.

I think that this block relies on (not having) the interface of the API offered by the Darwin module. So since the Darwin module is imported earlier if canImport(Darwin) is true, and we are assuming that the Darwin API is implicitly coming from that module, then !canImport(Darwin) is correct here.

We also looked into FreeBSD docs, and it looks like it doesn't expose exactly that symbol. (jemalloc, its malloc implementation, may have a similar one, but it isn't documented API). Let's just use !canImport(Darwin) here, and let FreeBSD port champions see if a different implementation can be used.