Re: On linux, potentially setting INIT_SYSTEM=systemd?

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

coypu%sdf.org@localhost writes:

> Hi folks,
>
> I needed to have a tunable for "is this a systemd system?" for a PLIST
> variable, and figured INIT_SYSTEM is the right way to put it.
>
> I'm hesitating to commit this because it might have unexpected effects,
> what should I watch out for?

Good question...

My gut reaction is that defining it based on /run/systemd is going to
result in build output changing based on various things, including
building in a chroot where systemd isn't running.   And hence instead of
this I would recommend that those who want to (well, are suffering from
systemd and choose to cope by) build for systemd set INIT_SYSTEM=systemd
in mk.conf.


> And then I used it like so, to resolve a PLIST issue:
>
> diff --git a/www/yaws/Makefile b/www/yaws/Makefile
> index 4d8f8a4d2b38..7cfa5580d786 100644
> --- a/www/yaws/Makefile
> +++ b/www/yaws/Makefile
> @@ -45,6 +45,13 @@ CONF_FILES=          ${EGDIR}/yaws.conf ${PKG_SYSCONFDIR}/yaws.conf
>  RCD_SCRIPTS=           yaws
>  RCD_SCRIPT_SRC.yaws=   ${WRKSRC}/scripts/netbsd/yaws.sh
>  
> +PLIST_VARS+=   systemd rc.d
> +.if ${INIT_SYSTEM} == "systemd"
> +PLIST.systemd=         yes
> +.else
> +PLIST.rc.d=            yes
> +.endif
> +
>  WWWDIR=                        ${PREFIX}/share/yaws/www
>  
>  OWN_DIRS=              ${VARBASE}/log/yaws

There is prior art about smf.