Re: [OE-core] [PATCH] openssh: Add SSHD_SERVICE_TYPE variable

[Richard Purdie <richard.purdie@linuxfoundation.org>]

On Mon, 2023-12-04 at 16:54 +0800, Yu, Mingli wrote:
> From: Mingli Yu <mingli.yu@windriver.com>
> 
> There are two types of sshd server now, one is based on socket
> activation(sshd.socket) and another is service activation(sshd.service).
> And the default sshd service type is based on socket by default as below.

Firstly, I'm a little frustrated about how these changes are being
proposed in pieces. The original patches made no mention that this was
eventually going to be controlled by a new variable. I did wonder if
something like this was going to happen but the didn't push back. Now
we're "pinned" into a corner with the direction this code takes now
we've already merged the on target configuration change.

> But it's more convenient to determine the service type at build time
> if there are so many devices.

Why did we merge a patch which made this an on target decision then?

> So add SSHD_SERVICE_TYPE variable to enable sshd.socket or sshd.service
> at build time and we still enable sshd.socket by default now.

No, we're not adding randomly named variables which take magic options.

I cannot look at SSHD_SERVICE_TYPE and know that "1" means socket
activation and not setting it means service activation (or whatever,
the fact I can't check what I've written easily proves my point). What
happens if I set it to "true", or "socket", or "apples"?

I'd suggest thinking about a PACKAGECONFIG option such as systemd-sshd-
socket-mode and systemd-sshd-service-mode which at least uses naming
users are mode used to and says what it does.

Cheers,

Richard