My main work computer is a Windows 7 PC but I also have a Mac Mini that I use for cross-platform testing and other various activities.  I have two monitors attached to my PC and one connected to my Mac.  Since I have two computers but I only want to use a single keyboard and mouse, I had two options: using a KVM switch with the V part disconnected (thereby only using the keyboard and mouse) or use Synergy.

I’ve used a KVM switch before without the video switching and it’s incredibly confusing.  I’ve also used a KVM switch with video switching and it’s pretty inconvenient.  So, I decided to use Synergy.  Read on for how I got it all set up.

For the uninitiated, Synergy is a program that allows you to use a single mouse and keyboard with a number of computers.  You choose one computer to be the Synergy server and you connect your mouse and keyboard to it.  All the other computers you want to use are clients and they receive mouse and keyboard input from the server over the network.  You do some configuration to tell Synergy about the size and placement of the screens, as well as some networking stuff, and you’re all set.  When you move your mouse off the side of the screen connected to one computer, it shows up on the screen of the adjacent computer.  Keyboard input is routed to the computer containing the mouse cursor.

All in all, it’s a great system.  It’s cross-platform, so it works on Windows, Linux, Mac, etc.  It’s much less confusing than using a KVM switch without video switching and it’s much faster and easier to use than a KVM switch with video switching, since all screens are viewable simultaneously.

However, it does have one drawback: all data transferred between linked machines is sent unencrypted.  So, if I try to log in to a website on one of the client machines, my password is sent in the clear over the network.  You can set up SSH tunneling if you want, and I’ve done this in the past, but it’s a pain.

This time around, since I’m using wireless network adapters to connect both my PC and Mac to the network, I have a spare ethernet port on both machines.  This made me think that maybe I could set up Synergy to work with a cross-over cable, meaning that all my Synergy data would travel over this cable and not the university network.  If the data isn’t traveling over the main network, there’s no need to encrypt it.

Of course, the same thing could be done with a cheap network switch, and this approach would be necessary if more than 2 computers were involved.  However, I don’t have a network switch or more than 2 computers and I do have a cross-over cable, so I thought I would give it a try.

It turned out to be easy. I just configured the wired ethernet adapters on both machines with local IP addresses.  I assigned 169.254.0.1 to the Synergy server and 169.254.0.2 to the client.  I gave them both a subnet mask of 255.255.255.0 and I didn’t enter any further information – no default gateway, no DNS servers, etc.

I tried pinging each machine from the other and they found each other with no difficulty.  Of course, in this case, I was using the IP addresses I had just assigned and not the hostname of the machine on the main network.  One big upside of this approach is that the ping times are tiny.  Since these machines are connected directly, there is no other network traffic to slow down the Synergy traffic.

With the machines connected and seeing each other, I just had to install and configure Synergy.  The Windows installer is pretty good and it automates the creation of the config file.  I basically just followed the directions in the configuration utility and I looked up some stuff in the online documentation, as needed.

I did have one difficulty.  Initially, I couldn’t get my Mac to connect to the PC.  In the config file on my Windows machine I had used the name I had given my Mac.  However, the Synergy client on my Mac kept telling me that the connection attempts were being refused by the Synergy server.   It turns out that, while I had given my Mac a name manually, because it was getting its IP address via DHCP, it was getting another hostname on the network. So, I changed the hostname for the Mac in the configuration file on my Windows machine to reflect the hostname the Mac received from the DHCP server.  This fixed the problem.

With the Synergy server and client talking to each other, one thing that I needed to deal with was  the possibility that Synergy traffic would be sent over the main network despite the presence of the cross-over cable.  This can be addressed in two ways.  1) When invoking the Synergy client, you supply the hostname or IP address of the server.  So, you just make sure that you use the local (non-globally routable) IP address of the ethernet adapter on the server.  The client won’t be able to reach a local IP address over a network populated with globally routable IP addresses.  2) For an extra level of security, you can specify that the Synergy server only listens on a particular network adapter. So, I told the server to only listen for connections on the ethernet port with the local address.  Thus, it will ignore any traffic that comes in over the wireless adapter.

To verify that these measures worked, I turned off the wireless adapter on the server and tested whether the setup still worked.  It did!  As a final step, I configured the links between the monitors to deal better with the difference in size between them.  On my windows machine I have two 1680×1050 monitors in portrait orientation.  My Mac is connected to a 1280×1024 monitor in landscape orientation.  So, at the place where one of my Windows monitors meets the Mac monitor, I have 1680 pixels meeting 1024 pixels.  I just played around with the settings until it looked good.  Useful hint: 0% means the top of the screen; 100% means the bottom.

One final tip is that Synergy appears to not need to know anything about multi-monitor setups.  As far as Synergy is concerned, my Windows computer has a single monitor that is 2100 pixels wide and 1680 pixels high.  Initially, I was concerned that dealing with the two monitors on my Windows machine was going to make things hard, but it turned out to be a non-issue.  However, if, for whatever reason, you wanted to interleave monitors like (from left to right) monitor 1 from computer A, monitor 1 from computer B, monitor 2 from computer A, monitor 2 from computer B, I have no idea how to do that, if it’s even possible.

To sum up, I’m very pleased with my new set up.  It was easy to configure and it works great.  If you routinely work with more than computer, I recommend you give Synergy a try.  If your computers are at home and all the machines have local IP addresses, you probably don’t need to worry about the traffic being unencrypted.  However, if you’re at work in a large company or university like I am, it’s probably not the greatest idea to send your Synergy data over that network in the clear.  So, if you have spare network adapters and want to avoid tunneling over SSH, the approach described here might work well for you.

This bibliography is available as an EndNote library file and as a Rich Text Format (rtf) document.  Apparently, if you have a recent version of EndNote installed you can use its search features to sift through the data.  However, the file won’t open in EndNote 6, which is what I have on my laptop.  Zotero wouldn’t import it either.  I tried using the built-in search capabilities in Word and even jEdit on the rtf but nothing could provide me with what I wanted.  Basically, I wanted the ability to export entries that matched a given search criterion to a separate file.  Presumably, there are programs that can do this for you, but I don’t know of them and don’t have them installed.

In the end, I decided to convert the rtf to a simple txt file.  This put each entry on its own line.  With each entry occupying a single line in the text file, I just needed some way to search for a given term and then output each line that contains that term.  I have used sed and awk a little bit in the past and I knew that there must be some way to do that with either or both of those, so I looked into their syntax online.  I found this awk tutorial and, using the examples there, I was able to put together a command that does exactly what I needed:

awk '/term/ {print $0}' < bibliography.txt > term.txt

where “term” is replaced by whatever term you want to match.  You can further automate this by putting a bunch of these commands into a shell script or writing a little Perl program that will take a command line argument and insert it as “term” in the command.

Now I have a list of papers related to all the terms I’m interested in and I’ve got a fast way to search for further terms in the future if I need to.  The approach is a little “awk”ward (groan!) because I have to run it in Linux and I use Windows most of the time.  I no longer have Linux installed as a virtual machine on my laptop and I don’t even have Cygwin installed anymore.  So, I had to upload my text file to one my research group’s Linux servers, run the scripts, and download the results back to my laptop.  Once I figured out what I needed to do it took me less than half an hour to do it, though, so even if it’s kludgy, it’s still a lot faster than reading through the bibliography manually.

UPDATE: I just realized that all I did was replicate the functionality of grep using awk.  That is, I could achieve the same result with the following code:

grep term < bibliography.txt > term.txt

Additionally, it turns out that you can produce this functionality with sed as well, using the following code:

sed -n 's/term/&/p' < bibliography.txt > term.txt

I guess I missed the fact that I could use grep because I started thinking about using sed or awk before I converted the file to plain text.  Each entry was spread over multiple lines so I was thinking about needing something fairly sophisticated.  I know that grep can do regular expressions but my first thought went to sed and awk, which are like one logical unit in my brain because of the O’Reilly books that cover both.

However, upon rebooting, I was greeted with a message that Windows Explorer had crashed.  I tried restarting Explorer but it crashed again.  I tried rebooting a couple more times but to no avail.

So far, each crash has been one of two types:

Problem Event Name: APPCRASH
Application Name: Explorer.EXE
Application Version: 6.1.7600.16404
Application Timestamp: 4a765771
Fault Module Name: ntdll.dll
Fault Module Version: 6.1.7600.16385
Fault Module Timestamp: 4a5be02b
Exception Code: c000041d
Exception Offset: 000000000003d8db
OS Version: 6.1.7600.2.0.0.256.4
Locale ID: 1033

or

Problem Event Name: InPageCoFire
Error Status Code: c000009c
Faulting Media Type: 00000003
Damaged File Name: SyncCenter.dll
OS Version: 6.1.7600.2.0.0.256.4
Locale ID: 1033

The second one has been more prevalent by far.  Generally, if the first one is observed, it occurs first thing after a reboot.  All subsequent crashes are of the second type.  So, I searched for this information online.

Unfortunately, I haven’t been able to locate any information on this particular type of error.  A search of “Windows 7″ + InPageCoFire + SyncCenter.dll returns no hits on Google.  However, I did find some general advice, suggesting that reverting to an older known-good restore point could help.

So, I got out the install DVD and used it to go back to an earlier restore point.  I noticed that there was a restore point from early this morning, just prior to a critical update having been installed. I figured this must of have caused the problem so surely, reverting to the restore point prior to the update would fix it.  Unfortunately, it did not.

I tried reverting to an even earlier restore point.  That still didn’t help.  So, I tried running the built-in memory diagnostics.  That didn’t appear to produce any errors but I can’t be sure because it’s supposed to give you a report once you restart the machine and log in to Windows.  I can never get far enough to see the report.

I just downloaded memtest86+ and I’m running that right now.  My current expectation, however, is that there are no problems with the memory.  Rather, I expect that somehow SyncCenter.dll (and possibly ntdll.dll)  got corrupted.  I don’t know how to fix that.  I had a similar problem back in Windows 95 with some corrupted DLLs.  I ended up solving the problem by copying the affected DLLs from a friend’s computer.  However, I don’t know if I’ll be able to take that approach this time because I can’t even use Windows Explorer.  Perhaps I could use a Linux Live CD and the necessary DLLs on a USB drive to make the switch.

Or, I could just reinstall the OS.  When I installed the OS, I partitioned the 640GB drive into two partitions: 150GB for the OS and programs (on C:) and about 450GB for my data (on D:).  This would enable me to wipe the C drive and reinstall without requiring me to recopy all my data.  We’ll see.  If I end up having to spend much more time on this, I’m just going to wipe the drive and start over.

If anyone reading this has any suggestions on things I could try, please let me know in the comments.  Thanks!

UPDATE 1: According to this site, SyncCenter.dll is statically linked to ntdll.dll and won’t run properly if ntdll.dll is corrupt or missing.  So, maybe SyncCenter.dll is OK and the real problem is ntdll.dll.  I’ll have to see if I can find a way to replace this file.  Also, memtest86+ completed with no errors, so the problem doesn’t appear to be RAM-related.

UPDATE 2: After a bit more digging online, I found some references to problems with Windows Explorer crashing because of a weird interaction with Firefox.  I tried to uninstall Firefox but Windows Explorer crashed too rapidly for me to even open the Uninstall Programs window.  So, I booted into Safe Mode.  I was curious to see whether I still got the crashes.  I did.  Fortunately, in Safe Mode I managed to uninstall Firefox.  I rebooted into regular Windows, but much to my chagrin, Windows Explorer crashed again.

Earlier in the evening I had posted my problem on a Microsoft forum but I subsequently decided that all this troubleshooting was for the birds and that I should just go ahead and reinstall the OS.  The last time around, I had used my university’s site license to install Windows 7 Enterprise.  However, with the possibility that my problem was due to the corruption of a file, I wondered about how reliable it was to download a nearly 3GB ISO file via wireless, burn it on a DVD and then install it.  Fortunately, I had also bought a copy of Windows 7 Professional from Win741.com for $30 earlier this fall.  I paid a bit extra and had them send me a disc.  I decided I would use this disc to reinstall Windows 7 since it was a pressed DVD and is probably more reliable from a data integrity standpoint than a downloaded 3GB ISO image burned on a DVD.

The installation went fine.  A clean install seems to generally take 20-30 minutes, which I think is pretty reasonable for an operating system.  I used Ninite to install a bunch of software we use.  Importing all my music into iTunes was easy – I had spent some time cleaning up my iTunes library with my last install and since I had put all that data on my data partition, it was easy to import again.

Importing photos into Picasa looks like it’ll take a bit more work.  The photos themselves are no problem; the problem lies with the face recognition.  When I copied all our pictures over from the old hard drive, the face “groups” were preserved (in the .picasa.ini files in each folder containing photos) but the names associated with those groups were not.  So, I didn’t have to do any obnoxious manual sorting like I did when I first installed Picasa on our old machine but I did have to input names to go with those face groups or at least connect those face groups to contacts from my wife’s set of Gmail contacts.

This time around, it looks as though the face who were connected to Gmail contacts have had those connections preserved while it looks like I’ll need to reenter the names of people (mostly children) who weren’t in my wife’s contact list and whose names I had entered manually.  I’m going to let it crank over night and in the morning I’ll see what needs to be done.

In any case, I hope the curious case of the crashing Windows Explorer doesn’t rear its ugly head again.  Though, if it does, hopefully, by then I’ll have received a useful answer to the question I posted on the forum.

I had installed CUPS on imladris only a few months ago, so the process should still be fresh in my mind. However, imladris ran Ubuntu 6.06 LTS, whereas orthanc runs Ubuntu 9.04. I’m not exactly sure whether 3 years* has made much of a difference but I’m going to try to follow online explanations to get it working. As I work on this, I’m just going to give an account of what I’m doing, where I found instructions, etc. until I get it working. If you’re interested in this, read on. If you’re not, maybe you can find a more entertaining way to spend your time, like playing Canabalt.

Last night I disconnected the printer from imladris and connected it up to orthanc. I also installed CUPS using the following command:

$ sudo apt-get install cupsys cupsys-client

I don’t actually know what the cupsys-client package does or whether I need it at all. Looking around at the Ubuntu package repository, it appears that cupsys provides the daemon and cupsys-client provides client software for the local machine. Thus, since I don’t care about being able to print from orthanc itself, I probably don’t need cupsys-client. Oh well.

Ok, so I’ve got the printer connected and CUPS installed. What now? I recall from the last time I installed CUPS, using the web interface was pretty easy. So, let me try to get that set up. First, I tried to simply enter what I thought was the correct URL and I got nothing. So, I did a quick search online. This site suggests editing /etc/cups/cupsd.conf and replacing the following line:

Listen localhost:631

with

Listen 0.0.0.0:631

I made this change and restarted CUPS by typing:

$ sudo /etc/init.d/cups restart

When I did this, rather than getting nothing when I pointed my browser to that same URL, I got 403 Forbidden. So, that’s a start. At least the web server is returning something.

Up to this point I had been working remotely on my laptop using the command line via PuTTY. I was doing this because I don’t have a high resolution monitor attached to my MythTV machine; it’s only connected to my TV. However, I decided that using the browser locally might make things easier.

So, I shut down the MythTV frontend on orthanc and opened up Firefox. I pointed it to http://localhost:631, at which point I was greeted by the CUPS administration page. Hooray! Using the CUPS administration page I searched for detectable printers. It detected my printer (HP LaserJet 6MP – old school, I know) and allowed me to add it. I could then pick the driver I wanted. On imladris I had used the Gutenprint driver. However, this time around there was a postscript driver available, so I chose that. I don’t know if it’ll be better than the Gutenprint driver – only time will tell. I printed a test page from the adminstration site and it came out very nicely.

So, it appears to work locally. Now, I need to knuckle down and actually figure out how to access the CUPS server remotely. To figure this out, I consulted the official documentation for cupsd.conf.

Basically, I was able to figure out that I needed to change something in the following code block:

<Location />
Order allow,deny
</Location>

This governs access to the main part of the server. There are separate, more restrictive blocks for the admin and config sections of the server. For now, I’m not going to worry about being able to access the administrative or configuration functions remotely.

From the documentation, I found that Order allow,deny rejects requests by default then applies additional rules starting with allow lines first, followed by deny lines. The other order, namely Order deny,allow accepts requests by default then applies rules specified in any deny lines followed by rules in any allow lines. I don’t really understand this fully because I tried a few things that didn’t have the results I expected. However, for now, changing the original block of code above to the following got things working:

<Location />
Order deny,allow
</Location>

I think that with this configuration, anyone in the world could print on my printer. That’s probably not a great situation. However, I think it may have been the situation with my CUPS installation on imladris as well. I never had any problems with it, so hopefully, I won’t have any problems with this new installation.

Ok, so now I can access the CUPS server from my laptop and presumably any other computer. The last step is installing the printer on each computer I want to be able to print from. I started with my laptop. I did a quick search to see how I should do it. I came across this site and tried their recommendation, which is to install the printer as a local printer using a new local port. However, when I tried to enter the URL of my printer, it told me it was an invalid port name. So, I looked around a little more and found this site. They recommend setting it up as a network printer. I tried this and it worked! I chose the appropriate driver and printed out a test page. Sweet.

Now, I just have to install this printer on our other two Windows machines and we’re all set. It actually didn’t take very long to get this all set up. Like I said, the CUPS configuration might not be the most robust but I think it’ll get the job done for now.

If anyone reading this has any suggestions for a more robust cupsd.conf setup, please leave me a note in the comments. Thanks!

________________________________________

*For the uninitiated, Canonical releases a new version of Ubuntu every 6 months. Recently, these versions have been released in April and October. The version numbers use the year and month of release. So, 6.06 means it was released in June of 2006, where 9.04 means it was release in April of 2009. Each release is also named after an animal preceded by an adjective beginning with the same letter as the animal’s name. These names have progressed in alphabetical order since 6.06 (Dapper Drake). 9.04 is called Jaunty Jackalope, while the upcoming version will be called Karmic Koala. The next long-term support (LTS) release will be 10.04, called Lucid Lynx.

I just checked my records and it shipped from Dell on September 5th, 2003.  We likely received it about a week later.  So, we’ve had it almost exactly 6 years.

Over the years, it has undergone quite a  few changes.  It’s a 2.4GHz Pentium 4 with a 800MHz front-side bus.  The processor, the motherboard, and the case are about the only components that are original but they have been a solid core around which to build.  A new graphics card required that I upgrade the power supply (now 500 W).  I expanded the RAM from 512MB to 1.5GB.  I’ve upgraded or swapped out hard drives a couple of times.  I’ve upgraded or swapped out optical drives a couple of times as well.  The floppy drive was replaced due to fatal dust intake but its replacement has also succumbed…

Despite the upgrades, the fact remains that it is a single core processor with an 800MHz FSB connecting to pretty slow RAM and it’s definitely showing its age in terms of performance.  When we finally retire it, it will likely find new life as a Linux machine for my daughter.  Until then, I expect it will continue to chug along like it has been doing almost uninterrupted for 6 years.  Thank you, Rechneer, you’ve served us well.

First the issues that motivated me to upgrade:

1) I wanted to use Netbeans 6.5.  This requires Subversion 1.5 or higher.  I had Ubuntu 7.10 installed and the package repository for that version contains only version 1.4 or something. I probably could have forced it to use a version from a different repository, but I didn’t know what the dependencies would be and this approach usually leads to a broken or less stable system.

2) We’ve been playing our music from iTunes on our Windows computer through our stereo via a USB FM transmitter.  The sound quality isn’t great and occasionally the reception is really dependent on where people are standing in the room – not ideal.  So, I copied all our music over to our MythTV machine with the idea that we could play all our music from it.  However, I couldn’t install the Songbird .deb file on 7.10 and I couldn’t get the regular version to play anything other than MP3s.  Since most of our tracks are (unencrypted) AAC, this is a problem.  By upgrading to a more recent version of Ubuntu, I hoped to solve this problem.

3) The fan on the graphics card I had been using (NVidia GeForce 5200) died due to excessive dust intake.  I wanted to replace it with a better card that has passive cooling (i.e. no fan but a good heat sink).

4) We started watching Hulu using this machine.  Under my old setup, the TV-out functionality had been handled by the MPEG-2 decoder on my Hauppauge PVR-350.  While this looks good for watching shows in MythTV, it’s pretty jerky if you try to watch other video.  So, I wanted to stop relying on my PVR-350 for TV-out and move it to the graphics card. I probably could have made this change without upgrading the software, but it’s just easier to let the installer auto-detect everything.  Also, some MythTV functions are not supported by the PVR-350 TV-out, so pressing certain buttons on the remote would cause the system to freeze or crash – not good.

The upgrade:

I ordered a new graphics card (NVidia GeForce 6200 4x AGP) and a new hard drive (320GB Western Digital EIDE) from GovConnection.  Let me say that I love GovConnection because I get a discount as well as free shipping due to my affiliation with MIT.  Plus, we have an account manager, so if I have any questions, I can send him an email or call him.  It’s great.  Anyway, I got a new hard drive because my preferred method of installing a new operating system is to take out the old drive and put it away for safe keeping, install the OS on the new drive, and then copy over any files I need.  This method works well if the installation takes too long and you need to use your computer again – just pop the old drive back in and you’re good to go.  This time around, I used an old drive I had put away for safe keeping from a previous OS installation (80GB Western Digital EIDE) as the boot drive and the new drive I ordered to store all my media.

I opened up the case, vacuumed it out, installed the graphics card, installed the drives, cut my hand on a sharp edge of the case, and put the case back together.  I decided to do a clean install rather than a version upgrade because I would have needed to do 3 consecutive version upgrades to get to the most recent version (7.10 -> 8.04 -> 8.10 -> 9.04). Using the Mythbuntu 9.04 ISO I had burned earlier, I started up the installer.  Everything went smoothly until I got to the partitioner.  I learned to install Linux back in 2002-2003 on old versions of Debian so I’m most comfortable with the manual partitioner, however, I got momentarily tripped up while trying to partition the disk.  I made 2 swap partitions on the 80GB boot drive – one at the beginning and one at the end.  I made an ext3 partition in the middle with the remaining space.  Then, I formatted my 320GB drive using XFS (because it does really fast deletes on large files – necessary for a DVR).  I told the partitioner to mount this disk at /media, since it would be storing all my DVR recordings and music.  However, I got an error saying that I was trying to install a separate filesystem on /media and that it needed to be part of the root filesystem.  I couldn’t figure out what the problem was.  Finally, after maybe 20 minutes, I realized that Ubuntu uses the /media mount point to mount removable disks when they are plugged into the system.  So, I changed the mount point of my 320GB drive to /data and everything worked.  Lesson learned: Don’t try to create a partition and mount it at /media – that mount point is reserved by the operating system.

Once the partitioner has done its thing, the installer copies over all the files you need.  Mythbuntu installs XFCE by default since it is less resource intensive than KDE or Gnome.  It also basically takes care of almost everything you need that is related to MythTV.  In addition to installing and partially configuing MythTV, it installs MySQL and creates the appropriate user and database, it installs Apache2 and configures MythWeb to communicate with MySQL, it installs and configures LIRC with a little input from you.  You still have to do some configuration of your MythTV backend and frontend, though.  I had some recordings on my previous MythTV installation that I didn’t want to lose so I followed these instructions to migrate the recordings over into my new setup.  They worked perfectly!  Also, when the installer was setting up the remote control (LIRC), I didn’t see my particular remote in the list of options, so I picked the one that seemed closest.  I still had my config files from the old setup, so I figured I could just copy those over.  However, once I started up the MythTV frontend, I was shocked to find the remote already worked!  When I was setting up my tuner, the default was V4L something.  I picked that but it didn’t work.  So, I switched to the MPEG-2 ivtv tuner and that seems to work.  Lessonss learned: 1) if you have a PVR-350, don’t pick V4L as your tuner type.  2) if you have a PVR-350 do choose the generic Hauppauge remote when you’re asked to choose a remote – it’ll work without any further intervention.

The installer asks about how you want the video to be displayed.  I choose TV-out via S-Video.  After everything was setup and I rebooted the system, everything was displayed on my TV.  This was great but I also have an LCD monitor that I want to use for other tasks.  Unfortunately, I had a really hard time getting a picture back on my LCD monitor.  I didn’t want to edit any files manually because I didn’t know whether the settings were now automatically managed by some program since a lot has changed in Ubuntu since 2007.  So, I tried the display configuration utilities in XFCE – no dice.  I installed Gnome and tried its display configuration utilities – no dice.  I installed KDE and tried its display configuration utilities – no dice.  I tried the NVidia display utilities (in package “nvidia-settings”) – no dice.  In the end, I had to edit /etc/X11/xorg.conf manually and fiddle with the nvidia settings utility.  It turns out that you need to set up 2 devices in the config file.  Since both displays are running off the same card with a single bus id, you need to specify a screen in each entry as well. Initially this didn’t work because I had set the S-Video port as screen 0 and the VGA port as screen 1.  These need to be reversed.  Then, you have to go into the NVidia utilities and enable the 2nd display (the TV), which, for some reason is disabled by default.  After some tweaking I got it all to work.  Lessons learned: 1) don’t be afraid to manually edit your xorg.conf file. 2) separate ports on a graphics card are accessed using two separate device entries in xorg.conf – both have the same busid entry but different screen numbers.

Finally, I had things mostly working.  I could start MythTV up on my TV screen and watch TV or a recorded show.  However, when I exited out of watching either live TV or a show, MythTV crashed.  I looked around online and found a group of people complaining about the same problem.  Fortunately, they had a solution: remove pulseaudio.  This fixes the problem but limits what your system can do.  So, if you reinstall pulseaudio, also mark libavcodec and libavutil for reinstallation.  This fixed the problem for others and it worked for me as well.  Lesson learned: if you install Mythbuntu 9.04, reinstall pulseaudio, libavcodec and libavutil to fix problems with MythTV crashing.

I think the upgrade has already accomplished most of what I set out to accomplish when I undertook this project.  MythTV works, the remote works, the MythTV features that I couldn’t use before (like fast forward) now work, watching shows on Hulu is now much smoother, I have Subversion 1.5 installed along with Netbeans 6.5.  There are still a couple of things on my to do list, however.  I haven’t installed Songbird yet, so I don’t know if we’ll be able to stop using the FM transmitter just yet.  Also, I am still having a problem with the wireless networking.  When I reboot, my wireless card doesn’t work.  I have to type >sudo ifup wlan0 to get the network to work.  Perhaps I need to put that in a startup script or something.  On the other hand, I have the wireless card in the bottom-most PCI slot in my computer.  The ethernet card, which is in the slot above it, is enabled when the machines boots.  Maybe if I just swap the cards’ slots, the wireless card will be enabled and the ethernet card disabled.  I’ll have to give that a try.

As a last note, working with old hardware is a pain.  This computer was bought in September of 2002, so it’s nearly 7 years old.  Nowadays, graphics cards almost exclusively use the PCIe interface; my computer uses AGP (4x AGP at that!).  Nowadays, many computers use serial ATA connections exclusively to connect to hard drives and optical drives; my computer uses exclusively IDE.  What I found in pricing components is that antiquated technology is more expensive than current technology.  That is, low-end AGP graphics cards are more expensive than low-end PCIe cards.  Similarly, all the good deals on hard drives are on SATA drives, not IDE.  In fact, the selection of IDE drives available at all is very small.  Hopefully, though, I won’t have to deal with this old hardware much longer.  I look forward to ultimately replacing this machine either with a more powerful MythTV box or with a Windows Media Center machine hopefully with CableCard tuners.

At the very least, the URL will need to change. However, I’d like to stop having to do admin work on the computer, like dealing with upgrading to the latest version of the software or troubleshooting hardware problems. So, ideally, I’d like to migrate all my photos to an online photo sharing site. I’ve been looking at the various options but none has really jumped out as a the logical choice. I have a lot of pictures, so I’m resigned to paying for a “pro account”, as they’re called on many sites. Generally, this amounts to $20-25/year, so it’s no big deal. There are a bunch of criteria that I’m judging the sites on but I won’t get into details now – I’ll save that for a later post.

In the meantime, I’ll focus on one criterion: that a tool exists that would allow me to automatically migrate my existing albums to a new location. I was aware of a sourcforge project called gallery2flickr, which will migrate your Gallery v.2 albums to Flickr. However, it’s nice to have choices and I was afraid that I would be locked into Flickr because it was the only service with a migration tool. However, today, I discovered Migratr, which will convert between any of 10 different photo sites, including Gallery v.2, Flickr, Zooomr, and Picasa Web. This gives me a lot more options.

Unfortunately, as you may have noticed, all these migration tools work with Gallery v.2 while I run Gallery v.1. So, I’ll need to migrate my Gallery v.1 albums to Gallery v.2 albums. I’m not sure how difficult this will be. I’m pretty sure that it will involve at least two steps, though: 1) upgrade from my current Gallery version to the latest version in the v.1 line and then 2) convert that installation into Gallery 2. This could be complicated by the fact that Gallery 2 uses a MySQL database and I’m not sure if I remember the root password for the MySQL server on the machine, which I’ll need in order to create a new user, which, in turn, I’ll need to give Gallery 2 access to the database and the appropriate tables. If things don’t go smoothly, this could turn into a huge pain since I need my MySQL installation for my blog, which, incidentally, is something further that I will need to migrate before too long…

Upon inspection, I discovered that the power supply fan had died.  The computer was still working but the fan in the power supply had given up the ghost.  It wasn’t too surprising, actually, since it had been chirping on and off for probably 2 years.  In fact, I had bought 2 replacement power supplies in July of 2006 just in case the existing power supply would fail – an occurrence that actually took a lot longer than I expected.  In any case, I discovered this problem about 15 minutes before the taxi was supposed to pick me up, so I had to just shut everything down.  After returning late on Monday night, I spent some time on Tuesday trying to get things fixed.  However, things weren’t quite as easy as I expected.

Replacing the power supply was easy since I had replacements on hand.  I also took the opportunity to clean the case out properly since it was full of dust.  I expected that after replacing the power supply and cleaning everything I would just power the machine on and everything would be back to normal.  Unfortunately, the network interface didn’t work.  I was getting some kind of IRQ conflict.  If I brought the interface up, I could ping it for awhile from a different machine and then it would stop responding.  I wasn’t able to figure it out yesterday so I just left it and worked on some other, more pressing things.  I took a look at the issue this morning again and this time, the network interface wasn’t showing up at all.  The IRQ conflicts were gone since the system was no longer seeing the network hardware.  So, I turned off the machine, opened it up, and put in an old ethernet card I had lying around.  This was recognized by the system as eth1 so I made a quick change to /etc/network/interfaces to move all of my previous settings from eth0 to eth1.  I restarted the machine and everything worked!

I think there are two lessons to be learned from this story.  One is that keeping spare computer components around pays off.  I fixed the problem with a spare power supply and ethernet card that I had on hand.  I didn’t need to go to any store or order anything online.  I probably saved myself a bunch of money and I definitely saved myself a bunch of time by hanging on to old stuff.  The second lesson is that this machine may go completely belly up before too long.  When motherboard components (like my ethernet port) start failing, the computer may not be too much longer for this world.  That said, as a 500MHz Pentium III from 1999, I’d say it’s had a good run.

In any case, I thought the problem might be related to our KVM switch. I have a Linux machine (hosting this blog) and a Windows machine in our computer desk and I use a KVM switch to share the keyboard, mouse, and monitor between the two computers. It’s really convenient most of the time, but every so often, it flakes out and has to be reset. So, I tried that (which involves unplugging it from both computers, since it draws power from the computers’ PS/2 ports), but it didn’t help. Then, I cut the KVM switch out of the equation and plugged the keyboard, mouse and monitor directly into the still incessantly beeping computer. This solved the problem of the blank monitor. Unfortunately, the monitor now showed me that my machine was hopelessly locked up.

Seeing the computer frozen was painful because I had been performing a 35GB online backup operation since 3AM last Thursday that was about 90% complete at the time of my rude awakening. With online backup services like Mozy (which I’m using), the first backup takes the longest (often about 1 day for every 3-5 GB being backed up) and I had read some reports online that when these backups are interrupted they often must be started again from the beginning. I was not excited about the prospect of losing 4 days’ progress. Not only that, we’re leaving for the holidays soon and I would have been unable to get things backed up before leaving. Unfortunately, I felt like I had no choice but to turn off the computer and hope for the best.

While the computer was loading the BIOS, the beeping began again, I got an onscreen message: “Keyboard Failure”, and then a blank screen plus lots of beeping. So, I shut down the computer again, swapped out the keyboard (which I had salvaged from the hallway near my office during a building cleanup campaign) and tried again with a different keyboard. I still got the beeping and the keyboard failure. I started thinking that perhaps the PS/2 ports on my computer had failed, so I planned to head into my office to borrow a USB keyboard and mouse. If the PS/2 ports had failed but the computer was otherwise fine, a USB keyboard and mouse would help me figure that out. Before heading into work, however, I took a shower.

I’m convinced that most of my big breakthroughs come in the shower and today was no exception. It occurred to me that perhaps the problem was related to the cheapo optical mouse we have. It flakes out probably twice a week but it behaves itself again if we unplug it and then plug it back in. So, I thought maybe it had finally gone belly up. I swapped it out with another PS/2 mouse I happened to have on hand and restarted the computer (after finishing my shower and getting dressed). It worked! No beeping, no blank screen! Windows booted up fine and I was able to resume the backup where things had left off, rather than having to start over from the beginning. The mouse promptly found its way to the trash and everyone lived happily ever after.

When I got into work, I was talking with a friend of mine about this problem and he mentioned that he thinks his keyboard is failing. Apparently, by pressing the Alt key a couple of times, he can get the X-server on his Linux machine to crash. I guess the moral of the story is that even the most unlikely parts of your system can fail and you need to be careful not to quickly assume that the problem is a part that has a bigger reputation for failure.