Re: One more anomaly after 5.99.17 upgrade

[Paul Goyette <paul%whooppee.com@localhost>]

On Sun, 13 Sep 2009, Alan Barrett wrote:

> On Sat, 12 Sep 2009, Paul Goyette wrote:
> > [running /etc/rc.d/mountcritremote]
> > /etc/rc.d/mountcritremote reported failure status 1
>
> /etc/defaults.rc.conf sets critical_filesystems_local=/var and
> critical_filesystems_remote=/usr.  Perhaps you didn't override those,
> and you don't have a /usr file system?  mountcritremote redirects the
> output of the "mount" command to /dev/null, so you won't see the actual
> error message when it tries to mount the nonexistent /usr file system.

Yup, that was it!  I never even thought to see what the defaults were 
for critical_filesystems_remote

Without reading the actual code, I cannot comprehend why mountcritremote 
would consider it a failure when attempting to mount an undefined /usr 
partition, yet mountcritlocal doesn't complain about not having a /var 
partition defined!

> I think we need a way of distinguishing between
> "critical_filesystems_remote=/usr because that's the default" and
> "critical_filesystems_remote=/usr because the admin said so", with
> no error in the first case even if /usr doesn't exist.  I have two
> simple ideas:
>
> A) Set critical_filesystems_remote="DEFAULT" and
>   critical_filesystems_remote="DEFAULT" in /etc/defaults/rc.conf,
>   and teach mount_critical_filesystems in rc.subr that "DEFAULT"
>   means "try /var or /usr as appropriate, but it's not an error
>   if it does not exist.
>
> B) Set critical_filesystems_remote="NOERROR:/var" and
>   critical_filesystems_remote="NOERROR:/usr" in /etc/defaults/rc.conf,
>   and teach mount_critical_filesystems in rc.subr that "NOERROR:foo"
>   means "try foo, but it's not an error if it does not exist".

I prefer B.  Option A requires embedding the "real" default value in 
some unexpected location.  Even if there were a comment telling people 
where to go to find the "real" default, it's an extra step...

> > [running /etc/rc.d/savecore]
> > Checking for core dump...
> > savecore - - - no core dump
> > /etc/rc.d/savecore reported failure status 1
>
> Either rc.d/savecore or savecore(8) itself should be changed to make
> "no core dump" not an error.

Yeah!


-------------------------------------------------------------------------
|   Paul Goyette   | PGP DSS Key fingerprint: |  E-mail addresses:      |
| Customer Service | FA29 0E3B 35AF E8AE 6651 |  paul at whooppee.com   |
| Network Engineer | 0786 F758 55DE 53BA 7731 | pgoyette at juniper.net |
| Kernel Developer |                          | pgoyette at netbsd.org  |
-------------------------------------------------------------------------