svn commit: r334617 - in head: . etc

[Brad Davis]

On Thu, Jun 7, 2018, at 8:46 AM, Eugene Grosbein wrote:
> 07.06.2018 20:17, Brad Davis wrote:
> 
> >>> I don't understand the drama over this.  rc.d startup scripts are
> >>> *binaries*.
> >>
> >> This is plain wrong. Example: before introduction of rcNG we had /etc/
> >> rc.serial
> >> supposed to be user-modified to contain local settings for serial ports 
> >> (uncluding USB serial).
> >> Now it is moved to /etc/rc.d/serial largely unchanged and is still 
> >> supposed to be user-modified.
> > 
> > We can change this script to advise the user to copy it to /usr/local/etc/rc.d.
> 
> Yes, we could. However, /usr/local/etc/rc.d has its limitations 
> comparing to /etc/rc.d:
> it is not possible to run a script from /usr/local/etc/rc.d before 
> FILESYSTEMS
> early/late divider that is critical if one needs to query local UPS over 
> serial port
> to ensure its battery has enough energy (say, above 5%) to delay fs 
> mounts until it charges enough.
> For example, statically linked "apctest" utility placed to /root/bin/ 
> does that just fine
> with some small scripting.
> 
> You see, my point is that you can never know beforehand of all 
> challenges a sysadmin faces in fields.
> And there should be really good reason to break things that work before.
> Like, solving some significant issue we have with current setup. Do we 
> have such?

This is not affected by my changes, you can install *additional* things in /etc/rc.d and they won't be touched just as today.

We can also make the rc subsystem more flexible to handle more things..


Regards,
Brad Davis