Re: [gentoo-user] trouble starting bash

[Helmut Jarausch <jarausch@igpm.rwth-aachen.de>]

On  7 Feb, David Relson wrote:
> On Sun, 7 Feb 2010 02:20:19 -0800
> James Ausmus wrote:
> 
>> On Sat, Feb 6, 2010 at 8:07 PM, David Relson
>> <relson@osagesoftware.com>wrote:
>> 
>> > On Sat, 6 Feb 2010 19:13:33 -0500
>> > Willie Wong wrote:
>> >
>> > > On Sat, Feb 06, 2010 at 06:29:27PM -0500, David Relson wrote:
>> > > > Your replies are much appreciated as we're in an area of Linux
>> > > > about which I'm poorly informed.
>> > > >
>> > > > Output (below) of "rc-status sysinit" indicated devfs stopped,
>> > > > so I started devfs (which didn't change /dev/pt*), then
>> > > > restarted udev (which didn't affect /dev/pt*).
>> > >
>> > > Right, but can you ssh in to the machine now (or open a terminal
>> > > emulator in X)?
>> > >
>> > > /dev/pts is just the mount point for the devpts pseudo
>> > > filesystem. In modern versions of linux the pts devices are
>> > > created on-the-fly when requested (as opposed to other versions
>> > > and some modern unixes where there will be a fixed number of
>> > > device nodes under /dev/pts or equivalent). All that just goes to
>> > > say that if /dev/pts is empty right after you restart the devfs
>> > > service, it is normal. A device file should be created
>> > > automatically now when userspace programs demand it. (E.g. if you
>> > > now ssh in, and if it succeeds, ls /dev/pts should show one
>> > > entry.)
>> > >
>> > > Try it, let me know if the problem is still there.
>> >
>> > Nope.  Both ssh and X terminal emulators are still broken.  No
>> > change in behavior.
>> >
>> > FWIW, most of the entries in /dev are timestamped 02/02 23:34 which
>> > is when I updated udev earlier this week. Today's upgrade/downgrade
>> > emerge hasn't affected the timestamps.
>> >
>> > A comparison of /etc/udev/rules.d to a saved copy didn't show
>> > much.  The only puzzling difference is:
>> >  --- 90-hal.rules      (revision 51)
>> >   +++ 90-hal.rules     (working copy)
>> >   @@ -1,2 +1,2 @@
>> >    # pass all events to the HAL daemon
>> >   -RUN+="socket:/org/freedesktop/hal/udev_event"
>> >   +RUN+="socket:@/org/freedesktop/hal/udev_event"
>> >
>> > removing the "@" and restarting udev hasn't helped.  Since the rule
>> > is hal related, I also restarted hald -- which hasn't helped.
>> >
>> >
>> What happens if you do:
>> 
>> mount -t devpts none /dev/pts
>> 
>> Does the problem go away?
>> 
>> -James
> 
> Eureka!  Problem fixed.
> 
> Looking in /etc/mtab, the last line is:
> 
>    none /dev/pts devpts rw 0 0
> 
> Perhaps the mount devpts command should have been issued as part of
> emerging udev, openrc, or sysinit ???  Should this be reported to
> b.g.o.??
> 
> David
> 

I have the following line in my /etc/fstab (I can't remember if I put it
there myself or not)

devpts               /dev/pts             devpts     mode=0620,gid=5       0 0

Since a "mount -a" is issued quite early during boot, this is mounted,
as well.

Helmut.

-- 
Helmut Jarausch

Lehrstuhl fuer Numerische Mathematik
RWTH - Aachen University
D 52056 Aachen, Germany