Bug#36952: sysvinit: Don't mount devpts in mountall.sh (?)
On Mon, May 10, 1999 at 11:37:16PM +0200, Miquel van Smoorenburg wrote:
> According to firstname.lastname@example.org:
> > Package: sysvinit
> > Version: 2.76-3.2
> > The libc6 package now has a init.d/devpts.sh script that mounts devpts on
> > /dev/pts. I haven't convinced myself that it should--it may be cleaner to
> > simply rely on fstab to mount it. However, if libc6 insists on mounting it
> > itself, mountall.so should add a nodevpts to the mount line:
> > mount -avt nonfs,noproc,nodevpts
> Sysvinit didn't add /dev/pts to fstab. It doesn't know about all these
> things. Since S35devpts.sh runs just before S35mountall.sh it should
> probably just grep for /dev/pts in /etc/fstab and only mount it
> manually if it's not present in fstab.
> Or even simpler, the pre/post inst of libc6 should fix fstab if it
> wants to
> Doing mount -avt nonfs,noproc,nodevpts is non-intuitive. If devfs is
> in fstab, yet /etc/init.d/mountall.sh doesn't mount it on boot that
> is really weird. Remember the principle of least surprise!
I tend to agree that mountall.sh should really mount all and not have lots
of special cases. I made my suggestion because I saw that an exception had
already been made for the /proc filesystem. Perhaps we should figure out
how to get rid of that exception.
The important point to me (which I did not state clearly in my first
message) is that users not see confusing "already mounted" messages at boot.
I suppose they are not too alarming, but they do appear to reflect a
I hope the maintainers of the relevant startup scripts will get together on
a scheme that will avoid warning messages. For example, agreeing that
_only_ mountall.sh mounts filesystems (other than /) would be great, though
perhaps there are obstacles to this.