ssh_systemd_notify: remove pretended abstract unix sockets "support"

[turistu]

as per https://lwn.net/Articles/993705/

and while at it, git rid of some prissy cargo-cult too

[djmdjm]

If there's a bug in the implementation then I'd rather it be fixed rather than removed.

and while at it, git rid of some prissy cargo-cult too

This is unhelpful language.

[turistu]

If there's a bug in the implementation then I'd rather it be fixed rather than removed..

Of course, you're not telling what "fixing" it would entail -- but neither accepting just a subset a unix-domain abstract sockets addresses (e.g. those who contain just a single nul byte), nor adding and documenting an ad-hoc string escaping mechanism (just for the sake of sending a literal READY=1 and RELOADING=1 message to systemd in a non-default configuration) would be "fixing" anything at all.

and while at it, git rid of some prissy cargo-cult too

This is unhelpful language.

Ok, keep it broken then

[djmdjm]

https://lists.mindrot.org/pipermail/openssh-unix-dev/2026-January/042348.html

[turistu]

https://lists.mindrot.org/pipermail/openssh-unix-dev/2026-January/042348.html

That's pointless and broken.

Pointless because one would have to compile a modified systemd in order to use an abstract address for the notify socket, so it's not serving any real-life purpose.

Broken because that would only accept an arbitrary subset of abstract socket addresses, namely only those that do not contain more than a single nul byte. So it's not even worth as example code.