[5469] in RedHat Linux List
Some things that could use attention
daemon@ATHENA.MIT.EDU (Paul Gortmaker)
Thu Nov 21 17:48:20 1996
From: Paul Gortmaker <Paul.Gortmaker@anu.edu.au>
To: redhat-list@redhat.com
Date: Fri, 22 Nov 1996 03:43:50 +1100 (EST)
Resent-From: redhat-list@redhat.com
Reply-To: redhat-list@redhat.com
Here are some things I noted about RH 4.0 that people here might want
to take note of. Please take this as constructive critisism, as I know
it is difficult to create a complete system and then test all possible
configurations, package interactions, etc.
1) If you specify no mouse at install, you don't get an /etc/sysconfig/mouse
(Not even one with all the lines commented out). Then of course the gpm
script in /etc/init.d complains. I'd suggest that for no-mouse systems,
the /etc/sysconfig/mouse contains a line like "NOMOUSE=yes" and that the
gpm script catch this and exit cleanly. It could also contain a comment
stating something like "You should be using "mouseconfig" to edit this"
to help point people like me doing "hand-editing" at the right util.
2) During the install it asked to do an "X -probeonly" and I said yes, to
see what it came up with. It came up with an error (I had specified a
generic S3-911/924 and generic VGA) and it said "start again" or something
like that but when I hit "Ok" it just continued with the install?!? Later
on I realized it had failed to create an /etc/X11/XF86Config file. After
I copied the XF86Config I had been using previously it all worked okay.
Note that this is RH4.0 via NFS without the Metro-X server.
3) There is an *ugly* hack in /etc/init.d/rc.sysconfig to handle SCSI
tape drive modules. It barfs if /proc/scsi/scsi doesn't exist. If such
a hack is *really* required it should do a if [ -f /proc/scsi ] or
someting like that 1st.
4) The UID/GID of "nobody" is 99. I thought that the value of 2^16-2
was traditionally used for sane handling of the "root_squash" NFS
mount option.
5) In rc.sysconfig (again) depmod is called without an "if [ -x " to
check if it exists 1st. In addition, it should be wrapped in an
additional "if [ -f /proc/ksyms" so that it doesn't moan when a
non module kernel is used. Also, launching kerneld should be in
a similar wrapper.
6) After the install is complete, perhaps it could run updatebd so that
the user doesn't get an error message which might appear like a
configuration error. An alternative is to have the locate rpm include
a zero byte /var/lib/locatedb with a timestamp of 01/01/1970 so the
user gets a sensible warning instead of an error message.
7) Why does the /bin/login process stay hanging around after launching
the users shell? On an 8MB machine, a single /bin/login accounts
for 10% of memory.
8) Mingetty assumes some things about the tty that it shouldn't. For
example, if you kill the above /bin/login process, the tty is left
in a useless state where a carriage return is physically printed as
a ^M making it impossible to login. It should set it to a known
sane state and not assume that it is in a sane state.
9) Holding down <Ctrl><Alt><Del> as opposed to just hitting it once
causes endless disk activity, and the machine won't reboot until
you let go. Okay, you say "don't do that then" -- true, but it is
a DOS-ism habit that I've seen novices do, and the machine's response
doesn't look very professional.
10) The behaviour of "rpm -e" and "rpm -i" is not consistent with respect
to suffix stripping. With the -i or -U options, it happily accepts
names with ".i386.rpm" suffix, wheras -e doesn't. The -q option
is also like the -e option, in that it doesn't strip. An example:
[root@ratbag i386]# rpm -i gpm-1.10-7.i386.rpm
package gpm-1.10-7 is already installed
error: gpm-1.10-7.i386.rpm cannot be installed
[root@ratbag i386]# rpm -e gpm-1.10-7.i386.rpm
package gpm-1.10-7.i386.rpm is not installed <---- blatant lie :)
11) An "rpm -e ash" results in an error of the form:
/bin/sh: [ : == : binary operator expected
12) The "procinfo" program will not run without /proc/modules and
hence is useless on a non-module kernel (not a RH specific problem
really, but I thought I'd mention it.)
13) There are a few dependencies missing that I noticed:
XFree86-devel should depend on: XFree86
XFree86-fonts should depend on: XFree86
Fvwm95-icons should depend on: Fvwm95 (or fvwm)
sendmail-cf should depend on: sendmail
sendmail-doc should depend on: sendmail
rpm should depend on: cpio *IMPORTANT*
The last one is really bad, as it let me "rpm -e cpio" without
any warning. And of course you then can't use rpm to put cpio
back as rpm needs cpio etc. etc... Ooops. (BTW, the error message
that rpm gives when cpio is not there isn't particularly helpful. It
just says "installation failed." :( Ugh.)
14) An "rpm -e etcskel" will never work without an error of
directory not empty, since the dotfiles in /etc/skel that don't
get deleted cause the rmdir to fail.
15) Access to the descriptions in the rpm packages during the initial
install/selection phase would be useful. This is one of the advantages
of the Debian install method over RH. For example, I had no idea what
"TheNextLevel" was (okay, so sue me). A pop-up dialog window with
the contents of "rpm -q -i foo.rpm" would be very nice. (Perhaps
there is already a way to do this that I overlooked...)
Okay, I think that is enough nit-picking for now. Hopefully somebody
finds this feedback useful. Feel free to CC me any replies, as I am not
on this list.
Paul.
--
PLEASE read the Red Hat FAQ, Tips, Errata and the MAILING LIST ARCHIVES!
________________________________________________________________________
http://www.redhat.com/RedHat-FAQ http://www.redhat.com/RedHat-Errata
http://www.redhat.com/RedHat-Tips http://www.redhat.com/mailing-lists
------------------------------------------------------------------------
To unsubscribe: mail -s unsubscribe redhat-list-request@redhat.com < /dev/null
