Re: regression in a micro update! lib/rtorrent 0.6.18

[Greg Troxel <gdt%lexort.com@localhost>]

Thanks - I had filed an issue in the wrong repo, and just crosslinkd
that and the freebsd one, and pointed out the commit.

snow flurry <snow%flurry.sh@localhost> writes:

> This works fine on Linux+glibc because <netdb.h> includes
> <sys/socket.h> for socklen_t (which I'm surprised to learn is allowed
> by POSIX, even if it shouldn't be depended on).

I am slightly surprised too, but POSIX has a practice of trying to allow
existing behavior when writing a spec, to try to make as many existing
systems conforming, if that is sensible.

Here, I'd say it isn't actually sensible, because it allows linuxy code
(defined as written to the linux system in front of you, without regard
to POSIX and portability :-) to build, instead of failing as the old
POSIX_ME_HARDER would require:
  https://www.gnu.org/prep/standards/html_node/Non_002dGNU-Standards.html

(This is likely not a problem on *BSD only because someone will try to
run on GNU/Linux sooner rather than later.)