Re: fixing slink boot-floppies for sparc
On Mon, 5 Apr 1999, J. S. Connell wrote:
> [recipient list trimmed]
> On Mon, 5 Apr 1999, Joshua Uziel wrote:
> > What happens with xdm? Does it run the X server as root and
> > keep it that way? Or is it restarted when the user logs in?
> > (I haven't used xdm in like 3 years.)
> Well, in all cases, the wrapper which invokes the X server is setuid root.
> It oughtn't matter what the group/other permissions are, as long as
> /dev/fb* are owned by root. Then again, I haven't done any experimentation
> in this respect - I didn't actually have to change the permissions.
Then perhaps something got unset from being SUID root on my
system during installation. Looking at /usr/X11R6/bin/startx,
the last line exec's xinit ... ls -l /usr/X11R6/bin/xinit shows:
-rwxr-xr-x 1 root root 10016 Feb 23 22:58 xinit
Which is clearly not SUID root... I have no reference point,
though so I don't know whether this is correct or not, and
this may be the key to our problems.
> > > The /dev/mouse link has to be created: "ln -sf sunmouse /dev/mouse"
> > >
> > > I think everyone agrees with this. The issue is where should be fix
> > > it. Should we just hack the boot-floppies to do this? Is this a
> > > base-files bug? A xserver-sun bug? (I don't think the latter)
> > Does anything other than X use /dev/mouse? How about gpm? I'd
> > say if X is the only thin to use /dev/mouse then it's the latter
> > else the former.
> Nothing actually explicitly refers to /dev/mouse unless you tell it to, I
> think, although I believe many things default to it.
> The reason (as I understand it) is that on Intel boxes there are a whole
> pile of different mouse devices - various kinds of bus mouse plus serial
> mice - so /dev/mouse evolved as a convenience. Most useful when the
> installer asks you "What kind of mouse do you have?" and automagically
> configures the link for you.
> On Suns, though, there is only one kind (I think - maybe the PCI keyboard/
> mouse stuff does things differently?), which is /dev/sunmouse. If that's
> the only mouse device, then perhaps the base ought to contain the
> /dev/mouse -> /dev/sunmouse symlink.
Sure, why not? Or then, if the *only* mouse device is /dev/sunmouse
then /dev/mouse could be the device and /dev/sunmouse can be
skipped altogether... I'm indifferent... it's truly a matter of
preference. I for one would probably just do the symlink method.
Also, on the topic of of which kernel to package... yeah, I'd vote
for the 2.2.5 kernel being used. I have a happy 2.2.5 SS10:
Linux sparcy 2.2.5 #1 Mon Mar 29 01:04:53 PST 1999 sparc unknown