Fedora Core 5 on my various machines.

General overview

Here are some links. With this distribution you are going to need them. Stanton Finley's notes have been quite helpful.

It is late April of 2006 and I am getting busy upgrading some of my machines to Fedora Core 5.
As always, I am doing full installs, keeping my /home partition and reformatting /boot and / as well as the swap partitions.

First off let me say that FC5 presents some rude and unpleasant surprises. In particular, the ability to install EVERYTHING when doing a full install is GONE.
Because this really chaps my hide, and I tend to rant on and on about what a kick in the head this is, I present my comments, investigations, and solution to all this in a separate page:

On the brighter side, the 2.16 kernel is impressive, at least my desktop (Gnome) comes up in a flash, which is nice. The base install the way I do it takes about 4.5 G of disk space.

Reading through the release notes, here are some expected highlights (and things to watch carefully):

Launch into the install

The graphical install works fine, but you have to watch closely for the point at which it wants to repartition your drives, click on a hidden menu and instead tell it to do a custom setup. Then a bit later tell it to do an install rather than an upgrade. I always disable SELinux entirely, your mileage may differ.

Before I do the install and reformat my root partition, I always do:

cd /etc; tar cvf /home/etc.tar.
This allows me to copy lots of things back into the new /etc. In particular I copy sections out of /etc/group, etc/passwd, and /etc/shadow to restore my users with their previous passwords.
You must take care to copy grub.conf since it is in /boot. You may also want to ensure that you keep a copy of /etc/X11/xorg.conf just in case this becomes a link someday. At this time it gets preserved in the etc.tar archive. This is especially important if you have a tricky dual head setup (as I do).

Yum

Nicely, the system comes right out of the box with a nicely working yum configuration. This allows me to solve a number of problems without having to struggle with yum itself, nice.

However, to do updates we should use our local mirror! This is a matter of editing the files in /etc/yum.repos.d and pointing them to mmt.mmto.org instead of the fedora sites.

And we add the livna repository. The thing to do is to go to the livna site and read the stuff under their configuration link. Livna is an extension of Fedora extras, and in particular handles some packages that are tangled up in US patent law. Particular among these are an RPM for the nvidia drivers. Before you will have any luck using yum with livna packages, you will need to kick things off (in particular getting the livna GPG key) via: 

rpm -ivh http://rpm.livna.org/livna-release-5.rpm

And I do the same for the Tim Pickering TEP repository: 

rpm -ivh http://caliente.mmto.org/~tim/RPMS/fedora/FC5/RPMS/i386/tep_repos-5.0-1.fc5_tep.noarch.rpm

(actually the above gets some odd error about file not found by glob, so I use wget to fetch the file and rpm it in by hand).

Later, after doing my exercise to install all the packages on the 5 install CD's, I edit /etc/yum.repos.d to use our local mirror and (on 7/19/2006) do: 

yum update
This updates 716 packages. Prior to the update, the machine I did this on was running a 2.6.15-1.2054 kernel. After the update it wants to boot the 2.6.17-1.2157 kernel.

Ethereal

Well the first problem is where the heck is it?. It would seem this is one of the packages that is not part of the "basic install" or whatever they call it.

I swallow my pride and go ahead and type: 

yum install ethereal
This works, or at least it fetches and installs something, but typing ethereal on the command line just has the system sticking its tongue out at me.
It turns out there are two packages (one depending on the other) and what I should have done is: 
yum install ethereal-gnome
This gets what I want, or at least now I can type ethereal on the command line and get what I know and love.

Nvidia driver shenanigans

Once my install was finished, the first issue that arose was getting my dual head Nvidia setup to run dual head again. This is no surprise and fully expected. The thing to do is to get the Nvidia driver and install it. I go to the Nvidia site and fetch the latest, namely:

As I have done a thousand times before, I make this file executable and run it as a script; The install fails because there is no kernel source (groan!). Now the thing to do (see the release notes) is:

Note that the kernel sources are NOT required in full to build modules, only the headers and such from the kernel-devel package.

Now the install gets into hot water trying to load the module it builds (it references an undefined symbol called print_tainted. A google search on nvidia and print_tainted shows that this is a well documented issue (this is why we don't jump on new releases right away unless we have lots of extra time and energy).

Fedora Core 5 ships with a 2.6.15-1.2054-FC5 kernel. I do the following because a google search indicates that the livna yum repository has an rpm for the nvidia driver that will work with the newer fedora kernel.

This gives me a 2.6.16-1.2096_FC5 kernel, and through some miracle updates the /usr/src/linux link to make it correct, doing yum update kernel-devel seems unnecessary.

Just for the record, when booting the 2.6.15 kernel I was getting the message:

Fatal error inserting acpi_cpufreq .... No such device

Apparently this was harmless and the module it was trying to insert supports dynamic CPU clockspeed changes (on those CPU's that allow it, my Athlon does not).
This message seems to have gone away with the newer kernel.

After doing this (getting the new kernel) I can build the nvidia driver using the file provided by Nvidia, but cannot load it. One hint says that it is looking in /usr/X11R6/lib instead of /usr/lib/xorg for modules and I need to fiddle with the ModulePath, (ModulePath is set in /etc/X11/xorg.conf it turns out, but I didn't know this at the time .... the thing to do (maybe) would be to change entries with /usr/X11R6/lib to /usr/lib/xorg, but we took a different tack, namely to give up on the driver from the Nvidia site and see what gives with the nvidia rpm on the livna site.

Actually I do the first and not the second since the system I am now working on is not SMP. This gets the driver from the livna site. I have been warned that the nvidia driver will run into trouble with SElinux, but of course I turned that off entirely during the install. This driver loads and runs and works. Now I just need to fiddle my xorg.conf to get dual head.

It turns out the fiddling boils down to just adding the following lines to the Device section of the /usr/X11/xorg.conf file: 

# extra stuff follows to get dual head.
        Option "TwinView"
        Option "SecondMonitorHorizSync"   "30.0 - 65.0"
        Option "SecondMonitorVertRefresh" "50.0 - 120.0"
        Option "MetaModes"                "1280x1024,1280x1024
        Option "TwinViewOrientation"      "LeftOf"

Initial impressions

And away we go! I still need to check on the new removable media support (it would be nice is flash cards and flash sticks were cleanly supported).

Core 5 does seem noticeably faster (at least the desktop loads up a lot faster) than FC3 was on my Athlon 1900+ system with 512M of RAM. Even my doggy laptop seems faster, so the 2.16 kernel seems to be a big win, even if the install is virtually ruined.

Install on a dual core AMD-64 system

I thought long and hard and decided NOT to install the 64 bit architecture on this one. I might if I anticipated lots of number crunching, or if that was the only way to get SMP support for the dual core, but neither of these seem to be the case and this will just save me lots of hassles, and I feel lazy in the absence of some compelling reason.

The install was simple and straightforward. I will have to cope with the Nvidia dual head issues at some point, but for now the install is done with as much junk selected as I can (excluding KDE and weird language support). I am using 4.5G on my ample 40G root partition.

Printing.

First off, make your life easy and use the GUI printer dingus. I find the "System" menu on my Gnome panel, select the Administration submenu, and then "Printing". This asks for my root password in order to run printconf-gui. I set up two queues (and ignore the browsed queues which look promising, but in fact do not work). by just hitting the new button and entering the following information into the series of forms that I am presented with: 

lw19
host lw19 port 9100
Networked Jet Direct
HP Laser Jet 4 plus

mmtcps
host mmt-cps port 9100
Networked Jet Direct
HP Color LaserJet 4600

Now I discover the printers are stalling with a message that I need to load A4 paper, so I go back to the printer configuration dingus and hit edit for each queue, select driver options, and change page size from A4 (the default apparently) to either Letter or US Letter, this makes things good.

If you get entirely frustrated with all this, just use netcat (nc) and send your ascii or postscript directly to port 9100 on the printer of choice.

Wine

This one seems really easy. Wine has made it into fedora extras, so I can just do: 
yum install wine
This Gives me wine-0.9.17-1.fc5 (and 9 other wine package of the same vintage). It runs winemine right away (this looks good!). However it gets a page fault trying to run NGS TOPO!, which is not good since this is the only thing I ever use Wine for.

Mail (sendmail)

Nothing really new here. I make sure I have a .procmailrc file in my /home directory (no .forward is needed, it would seem that sendmail is set up under fedora to just feed things to procmail if there is a .procmailrc, which is great.

I need to fiddle with the sendmail configuration and start sendmail on port 25. To do this I go to /etc/mail and edit sendmail.mc, commenting out one line: 

dnl DAEMON_OPTIONS(`Port=smtp,Addr=127.0.0.1, Name=MTA')dnl
I also add all the aliases (and virtual hosts that I want to be able to receive mail for) to the file: 
local-host-names
After this restart sendmail does the make to rebuild sendmail.cf and a quick test shows that I can receive mail.

This isn't really the place for this, but what the heck. In the midst of this upgrade I shuffled a couple of machines around and changing hostnames. After this, attempts to send mail to one machine which had not been upgraded, but which had its name changed just got the error: 

mail loops back to me (MX problem?) Local configuration error
It turns out this was a simple matter of adding the new machine name (and deleting the old) from the local-host-names file just described, and restarting sendmail.

Spam

Now that I have mail back on again, the spam is pouring in. There are several things to do: 
chkconfig spamassassin on
service spamassassin restart
And perhaps most important of all, create a file /etc/procmailrc that looks like this: 
DROPPRIVS=yes

:0fw
* < 256000
| /usr/bin/spamc

Flash

There are a lot of websites out there that need the flash plugin from Adobe (was Macromedia). It turns out this is easy via Yum, there is a special yum repository for this all set up via /etc/yum.repos.d/flash.repo. (This repository is in Spain, no doubt due to some legal issues with Adobe), the flash.repo file looks like: 
[macromedia]
name=Macromedia for i386 Linux
baseurl=http://macromedia.rediris.es/rpm/
enabled=1
gpgcheck=1
gpgkey=http://macromedia.mplug.org/FEDORA-GPG-KEY

Given this, just do: 

yum install flash-plugin

Best of all it works (in my case with firefox).

Java

And then there are some websites that use Java.

Go to the Sun Microsystems website and download jre-1_5_0_08-linux-i586.bin (JRE 5.0 Update 8).

What I do then is as follows: 

su
cp jre-1_5_0_08-linux-i586.bin /opt
cd /opt
chmod a+x jre-1_5_0_08-linux-i586.bin
./jre-1_5_0_08-linux-i586.bin
rm *.bin
ln -s /opt/jre1.5.0_08/plugin/i386/ns7/libjavaplugin_oji.so /usr/lib/mozilla/plugins/libjavaplugin_oji.so

It spits out a big license agreement, you answer yes, and it installs a bunch of files and you are good to go with firefox after making the link above.

VMware

I use vmware to run some windows stuff that I just cannot escape from.
See my vmware notes (which I updated after some fc5 fussing) at the prior link.
Have any comments? Questions? Drop me a line!

Adventures in Computing / tom@mmto.org