Half a Page of Scribbled Linux

This shouldn’t have been so hard to find that I had to post this, but it was. After downloading the Fedora Core 5 DVD distribution from Bittorrent, there are two files, FC-5-i386-DVD.iso and SHA1SUM.

First, check that the ISO is OK:

sha1sum -c SHA1SUM

There will be a few errors about the CD ISO files you don’t have. As long as the FC-5-i386-DVD.iso is OK, you’re in business.

Next, pop a blank DVD+R in the DVD burner and run (as root):

growisofs -Z /dev/dvd=FC-5-i386-DVD.iso

do not run growisofs -Z /dev/dvd FC-5-i386-DVD.iso like I did the first time or you’ll just get a disc with the iso file on it. 🙁

Friends and relatives have home DVDs they wanted to copy for other friends and relatives. Turns out this is rather easy. Use k3b:

On the menu, choose Tools/Copy DVD. Don’t forget there are three tabs worth of data to check.

Recently I got a request to combine movies on two discs. Now it’s getting interesting. This is a lot like the camcorder to DVD project I did except it involves starting with DVDs as a source instead of miniDV tapes. A little research turned up ripping instructions. Between these two references, I had enough to do the job.

The first step was to use k3b to rip the discs into two ISO files, movie1.iso and movie2.iso. Assuming movie1.iso was in /mnt/bigfiles, you could check it with xine dvd:///mnt/bigfiles/movie1.iso.

If they’re OK, mount them on the loopback device:

mkdir /mnt/iso
mount movie1.iso /mnt/iso -t udf -o loop=/dev/loop3

Then copy the files from /mnt/iso/VIDEO_TS that are named VTS*.VOB. These are the actual movie parts. Rename them from VTS_01_1.VOB to movie1.vob so you can tell them apart from the second movie.

Unmount the ISO with umount /mnt/iso and repeat with the second movie.

Next, get convert them to mpeg2 files (from here):

tcextract -i movie1.vob -t vob -x mpeg2 > movie1.m2v
tcextract -i movie1.vob -a 0 -x ac3 -t vob > movie1.ac3
tcextract -i movie2.vob -t vob -x mpeg2 > movie2.m2v
tcextract -i movie2.vob -a 0 -x ac3 -t vob > movie2.ac3

mplex -f 8 -o movie1.mpg movie1.m2v movie1.ac3
mplex -f 8 -o movie2.mpg movie2.m2v movie2.ac3

Next, create the XML file both_movies.xml:

<?xml version="1.0"?>
<dvdauthor>
<vmgm />
<titleset>
<titles>
<audio lang="en"/>
<pgc>
<vob file="movie1.vob" />
<vob file="movie2.vob" />
</pgc>
</titles>
</titleset>
</dvdauthor>

mkdir both_movies-dir
dvdauthor -o both_movies-dir/ -x both_movies.xml

Test with xine: xine dvd:///`pwd`/both_movies-dir

You can also create menus with dvdwizard (German site) or dvdwizard dvdwizard 0.4.2c (English site). It’s all scripts, so extract it in /usr/local/src and run it with a script like this:
#!/bin/sh
export PATH=${PATH}:/usr/local/src/dvdwizard-0.4.2c
dvdwizard -l dvdwizard.log -T "Big Combined Movie" -N NTSC -A en -t "First Movie" movie1.mpg -t "Second Movie" movie2.mpg


This will author a DVD in the directory ./dvd. Test it with xine dvd:///`pwd`/dvd.

If it looks OK, create an ISO: mkisofs -dvd-video -udf -o both_movies.iso dvd/. Then burn the ISO with k3b. Use the Tools/Burn DVD ISO Image menu option.

Since installing the HP Photosmart drivers, I’ve noticed that duplexed printing no longer works on the HP Deskjet 970Cxi.

In my old installation notes, I wrote that the printer was installed in March 2001 with the DeskJet 880 driver for Linux PrintTool.

Linux Printing.org shows the driver to use with this printer as the hpijs and has instructions for CUPS.

I went to the hpijs page and generated a PPD file for the printer. I saved the file in /usr/share/cups/model. Then restarting the services with service cups restart and service cups-config-daemon restart allowed the driver “HP DeskJet 970C Foomatic/hpijs (recommended)” to show up when modifying the printer in the CUPS Manage Printers page.

Now that I’ve got the HP Photosmart 8450 connected to the network, it’s time to set up Flounder (the Thinkpad 2611-472) to print to it. There is no HPLIP RPM for Red Hat 9, so I downloaded the source code for version 0.9.8. Configure ran but complained that it “cannot find net-snmp support (or –disable-network-build)”. It was looking for snmp_timeout in the netsnmp library. I ran ar t /usr/lib/libnetsnmp.a | grep time and didn’t find it. Oddly, it’s not in the library on FC4 either. I tried the older version of HPLIP, 0.9.7, that was in the FC4 RPM. My hope was that it didn’t have this dependency, but it does. So I downloaded the Net-SNMP 5.3.0.1 source. Configure asks a few questions about your set-up. I gave it all the defaults except for the administrative email address on that box.

The HPLIP configuration now works fine. After installing, running service hplip start failed with the error that it couldn’t find libnetsnmp.so.10. This library was in /usr/local/lib. I updated /etc/ld.so.conf to include this directory and ran ldconfig. Another service hplip start and service cups start and all’s well. See the prior post for setting up the rest.

I picked up an HP Photosmart 8450 on Woot. I was originally going to install it using USB, but thought better of it and set it up as a networked printer.

First, the CUPS drivers need to be installed:
yum install hplip

Get the PPD file and put it in /usr/share/cups/model. Set the owner and group to root and permissions to u=rw,g=r,o=r. Then run service cups restart.

Next, check the status with hp-makeuri. I got this error:
[ERROR]: Unable to connect to hpiod.

The hplip service doesn’t get started by the RPM installation, though it does get set up on chkconfig. BTW, doing service --status-all will show you hpiod and hpssd are not running. But the service name is hplib, not either of those. So what we need to do to get it going is service hplip start.

So, running hp-makeuri again gave me [ERROR]: You must specify IPs and/or DEVNODEs on the command line.

Adding the IP address of the printer (hp-makeuri 192.168.0.9) worked and gave me CUPS URI: hp:/net/Photosmart_8400_series?ip=192.168.0.9

Next, get the CUPS administration screen in your browser. It’s at http://localhost:631. Select Administration. Under Printers, choose Add Printer. The name can’t have any spaces, so I used hp8450. The location is the IP address, e.g. 192.168.0.9. Lastly enter a text description. I used “HP Photosmart 8450”.

At the device prompt, select IPP. At the Device URI prompt enter the URI from hp-makeuri, hp:/net/Photosmart_8400_series?ip=192.168.0.9. Enter HP for the Make. Select the HP 8400 driver.

Select Printers and print a test page.

I’ve got a SCSI scanner attached to /dev/sg0. Under previous versions of RedHat, I set the group to disk, added read and write permissions for the group and added myself to the group. Since upgrading to a 2.6 kernel which has the UDEV filesystem, every time I reboot, I have to do this over again. Until I found out how to make it stick. These directions are for Ubuntu, but they were close enough. Here’s what I had to do differently for Fedora Core 4:

Fedora doesn’t have /etc/udev/udev.rules, rather a directory called /etc/udev/rules.d. This is specified in the top-level configuration file /etc/udev/udev.conf. On my box, rules.d has one file called 50-udev.rules. I edited this file. On the line you add, I changed the group to disk instead of scanner. Once done, as root, do rmmod sg, then modprobe sg to reload the driver and check your work. One bit of goofiness I saw was that adding the SYMLINK parameter caused my permissions to be ignored. So I left it off. I can live with just /dev/sg0.

Copy TTF file to a directory, say /usr/X11R6/lib/X11/fonts/TTF.
Edit /etc/fonts/local.conf and make sure that directory is in a line like this:
<dir>/usr/X11R6/lib/X11/fonts/TTF</dir>
inside the <fontconfig> section

Restart X and you’re in business.

I’ve been running SETI@Home on a Windows box for years. Recently, they’ve switched from a dedicated processing program to the BOINC project, which lets you volunteer CPU cycles to projects other than SETI.

I installed the Linux version 5.2.13 of BOINC per the instructions.

The BOINC client failed to download any work units until I opened port 123 UPD in and out on my firewall.

Then the client was running, but the BOINC manager (GUI) wouldn’t connect to it. I kept getting a dialog box that said The password you have provided is incorrect, please try again. The solution that I found was this:

su -
su boinc
cd /var/lib/boinc
/usr/local/BOINC/boincmgr


I have BOINC installed in /usr/local. Switching to root first is necessary because the boinc user has no password.

-or-

(as root)
$ cd /var/lib/boinc
$ echo somepassword > gui_rpc_auth.cfg
$ service boinc restart

(as any user)
$ cd
$ echo somepassword > gui_rpc_auth.cfg
$ /var/lib/boinc/boincmgr

On my RedHat 9 laptop, I’ve started using Fluxbox. I’m using GKrellM to monitor the system (CPU usage, disk space, available RAM, etc.) and wanted to use a WiFi monitor to see how good my connection is in different places. In KDE, I used kwirelessmonitor. GKrellM has a plugin for this, but compiling it caused the make file to complain that I didn’t have gtk+-2.0. There isn’t any such named package for RH9, but what did the trick was apt-get install gtk2-devel.

Then the make worked. From there, copying gkrellm-wifi.so to /usr/lib/gkrellm2/plugins and restarting gkrellm was all that was needed.

There’s another plugin, gkrellmwireless that displays noise level and bit rate, but not as a graph. I use that only for the bit rate, which gkrellm-wifi doesn’t show.

[This post is a combination of activity over the past year. –Z]

I’ve been using Mondo Rescue to back up Hoover, my main PC running RedHat 9. It took me a while to get comfortable with how it worked during restore, but I finally figured it all out. But Mondo is a combination of programs and this is what caused my problem.

I upgraded Hoover to Fedora Core 4 and ran into a trap that I had laid for myself two years ago. At that time, I had installed a new 200 GB drive and reloaded RedHat 9 and my older RH 7.0 and 7.3 partitions on it. That kernel version (<2.4.19) didn't support drives larger than 137 GB and the partition table got a bit bent. After figuring out the problem, I upgraded the kernel, but didn't wipe the disk clean. Parted and fdisk had some complaints with the partition table, but things worked. When I put FC4 on, I decided I didn’t need the old RH7 anymore. I mounted it and started deleting files I knew I didn’t need just to go through it all and make sure I didn’t miss anything. During this process all hell broke loose. X garbled and crashed, then I got a kernel panic. My guess is that the bent partition table or the RedHat 7 inodes were crosslinked. Deleting things from the RH7 file system caused files in my RH9 file system to go kablooey. So I started from scratch and loaded FC4. Then I used the Mondo Rescue disks to restore the older partitions. This all worked fine until I went to backup from FC4. The disks didn’t get written. I traced it down to incorrect options being passed to mkisofs. Here’s the error message in the log: Executing 'mkisofs -J -no-emul-boot -boot-load-size 4 -b isolinux.bin --boot-info-table -c boot.cat -boot-load-size 4 -r -p MondoRescue -P www.mondorescue.org -A Mondo_Rescue_GPL_Version -V 1 -v . | builtin_dd of=/dev/dvd obs=32k seek=0'
mkisofs: The option '-P' is reserved by POSIX.1-2001.
mkisofs: The option '-P' means 'do not follow symbolic links'.
mkisofs: Mkisofs-2.02 will introduce POSIX semantics for '-P'.
mkisofs: Use -publisher in future to get old mkisofs behavior.

I couldn’t find anything about this on the Mondo site or with Google. This forced me to start looking for other backup solutions. My primary criteria was that the program support backup to DVD. (Did I mention all the failed restores I had with Travan TR-4 tapes?)

One candidate was RAB. This program hasn’t been updated in almost three years. Not an option. Then I looked at Bacula. DVD backup is supported in the beta version, but the program seems to have a large user community. Although I was unable to get the DVD backup to work, I did create a full backup to file, then split the file into smaller chunks and copied those to DVD with growisofs.