On this day, I own a buttload of Apple gear: iPhone, iPod, iPad, MacBook Pro, Apple TV, Apple TV 2, Time Capsule, Mac Mini… I’m planning on getting an iPhone 4S as soon as possible.

I was (and am) a big open source dork at heart, but Apple’s simplicity and ease of use, especially the past 7 years really spoke to the pragmatic side of me.  It turned me into a fan boy, I guess.  It didn’t hurt that OS X is NeXTStep is UNIX… I touch OS X and Linux every day.

One of Dinah’s first words was iPod which she taught herself so she could ask me to turn on music for her.

In college, I loved using the NeXT computers we had. My first class in college used scheme on the NeXT’s. The NeXT is one of the first places I played Doom.  I took calculus using Mathematica on the Mac in college.

Steve also touched my kids’ lives via Pixar. (Okay, my life too, I saw Toy Story in the theater first run, long before I had kids.) Also through their first computer, a Mac Mini.

There were too other public figures who’s lives and deaths touched me the way Steve’s passing is touching me today: Jim Henson and Fred Rodgers.

In any case, this point was to share that something that shouldn’t at all be fun is, but also to test to see if the blog migrated okay.   If you notice anything weird, please let me know.  But if its broken, you probably can’t let me know, and i’ll live in blissful ignorance.

• Duration is no longer hard-coded to 32:00 and actually reflects the duration of the show
• The script will attempt to find the TiVo via Bonjour/mDNS/ZeroConf/DNS-SD/whatever unless passed a -t flag with the TiVo’s IP address. If you have more than one TiVo, it will go with the first one it finds.
• Moved the stuff in lib to lib/tivo so the package is more easier sucked in by something like encap or stow

Download: tivoscripts-20100314.tar.gz

When I get some motivation later in the week, I’ll put the git archive online, incase anyone wants to clone it and do some development on it.

After two years of saying I was going to fully automate the process of downloading and transcoding shows for my iPhone, I finally got off my ass and did it.  The script is called TiVo2Podcast and it not only does the downloading and transcoding, but it stuffs the resultant video into a an RSS feed for easy consumption/playback by a podcatcher such as iTunes. I’m now automatically getting the shows off my TiVo and onto my iPhone for easy commute-time consumption. (I commute by train, I do not recommend commute-time consumption if you are driving.)

The ruby script wraps tivodecode, HandbrakeCLI, and AtomicParsley and is intended to be run from cron.  I’ve tested this on Linux, but it should run on any UNIX-alike, but it won’t run on windows since I make liberal use of the system() call. Also, this is intended for PERSONAL USE ONLY, do not set up podcast feeds and violate the ethics (and also the laws) of copyright left and right.

This is a very early version and can certainly use some tweaks and enhancements, primarily in configuring the shows you want to capture.  Right now, configuration is in the form of doing INSERT statements in SQLite.  Not very friendly, but it gets the job done until I can make a quick and dirty question based TUI. Here’s an example of setting up getting the best fucking news team on the planet:

insert into configs (config_name, show_name, rss_filename, rss_link,
                     rss_baseurl, rss_ownername, rss_owneremail, ep_to_keep, encode_decomb)
            values ('tds', 'The Daily Show', 'tds.xml', 'http://www.thedailyshow.com/', 
                    'http://example.com/podcasts/', 'Keith T. Garner', 'kgarner@example.com', 4, 1);

Download tivoscripts-20100304.tar.gz and let me know what you think. Make sure you read the README!

[Update 3/5: Forgot to add that all the code I wrote is under the Simplified BSD License, so have at it.]

It also makes me want to link to Anne’s idea of the best to learn Linux.

This reminds me of a post on usenet my friend Anne made in 1998.  To quote:

So I am left not bitter just tired of this whole meeting people, emotional ties, finding a compatible person and then having them yank away.

It is just an emptiness right now.

From what I have observed from my male friends, though, this is exactly the climate required to learn Linux. Without a full and happy lovelife or distraction of soft lips and a reason to kiss them, there is enough room to grasp the intricacies and nuances of such a fine operating system.

It has already begun to happen. As I walk down the street I am not thinking of emptiness, kising, nathan or any other previous SO’s, I am thinking of penguins, rm -rf / and lilo.

I am already convinved that linux will dull the pain better than heroin.

I still loves me some linux hardcore, though.

I’ve had a patch from MARK NOTARUS to make the menu have some more options and I’m using Console::ProgressBar from facets now.

It works well enough for my needs, but let me know if you hit any roadblocks. Just as a reminder, my target was to download stuff off my tivo on one of my headless Linux boxes for batch encoding for my iPhone and/or PSP.

Download tivo-ruby-0.2.tar.gz

On its own, the TiVo library stuff I whipped up will just download the content from the TiVo still locked up in TiVo’s DRM.  If you pair it with tivodecode (like the sample script does) you can get the video as an MPEG-2 stream.  Beware, HD content is HUGE.

I’ve only tested with my Series 3, and I’ve run it under both OS X and Linux, but it should work on any .  The UI on the sample script is pretty bad, but its an early version.  I’m interested in an feedback, patches, etc anyone might have.  I still need to do some rdoc as well.

This works for me, but I’d like it to be more useful for more people.   In any case, feel free to checkout tivo-ruby-0.1.tar.gz.

Some pros for cmake that I’ve found so far (compared to what we’ve been doing):

• support a big number of build environments on the different platforms. On windows it sports 11 different build environments, OS X 3, and Linux 2. For OS X and Linux, you only really need those two or three, but on windows it supports 4 different versions of visual studio as well as Borland, Watcom, and gcc.
• Takes care of the flags needed to build executables and libraries on those supported platforms.
• Does out of source builds on windows.
• Tracks dependences on all platforms without an external application
• Does search and replacing on things like .in files without having to call out to external applications

Some cons I’ve found so far:

• The documentation on the web page is pretty horrid. The book is pretty bad too, especially when compared to other technical books I’ve read recently, but its much better than the website. When combined with the book and experimentation, the FAQ is helpful.
• Doesn’t really have the concept of convenience libraries. This will result in common files being built multiple times. I don’t like this, but its not fatal.
• The CMakeCache is getting in my way more than being a help, but that might be the side effect of my learning process right now.
• I haven’t yet figured out how to make it query the person compiling the app if it can’t find something. This may not be possible. At the very least I want to make it bitch and bomb out if a required dependency isn’t there. I just haven’t found it yet, I’m thinking.

This isn’t an exhaustive review yet, but I wanted to get down what was in my mind before I forgot. I had just found the convenience library thing and that’s what inspired the post. My next step is to move a full existing project over to being built with cmake. This is a library that depends on expat, boost, curl, antlr, and (optionally) swig. Should be a good challenge.

[Update 11:58: I found an answer to my "bomb out if the dependencies are missing" question. Thanks, devchannel!]

[Update 2:51: No this isn't here just for g0ff. Turns out the latest cmake has modules to Find Java, Doxygen, Boost, Curl, Expat, and Swig already. It looks like just custom items for antlr and cppunit will be needed. Also, it only ever wants to link against dynamic libraries, not static ones. That's a PITA.]

[Update 5:52: Okay, the convenience library thing is upsetting.  The output of what I was working on is a static library and there are same example command line tools that link against it.  From my reading of the cmake stuff, I should just include the library source files to the target for the executables being created.  The problem with this is for n example programs, I'm compiling librets n times.  This doesn't seem very optimal.]

I’ve been playing with OpenVPN to get access to my network at home.  Since I have a MacBook Pro from work, that’s been my end point client.  I’ve been using Tunnelblick as my OpenVPN client to connect to OpenVPN server on my linux box (installed via DAG’s RPM repository.)  One thing that bugged me was how to get DNS so I can see my internal home DNS without breaking access to work’s internal DNS.  If I was using a linux laptop, I think my solution would have to do with running a local instance of named with some wacked out config to do caching only and refer to different DNS servers.  Hardly dynamic and a giant PITA to get going.

I was curious about how to make this go though, and what general solutions people had when I came across a post by Mike Erdely titled OpenVPN + DNS + OS X.  That is exactly what I wanted to do!  As a bonus he’s even using Tunnelblick.

Mike shows how OS X’s DNS resolver uses an /etc/resolver directory to get additional per-domain configuration information, as opposed to the blanket /etc/resolve.conf that unix users have come to know.  To get the mac to resolve kgarner.com using my doman’s internal DNS server I just need to create /etc/resolver/kgarner.com and put nameserver 192.168.1.10 inside of it.  This directs OS X’s resolver to ask 192.168.1.10 for any kgarner.com query.  He also shows how to flush OS X’s DNS cache via lookupd so if I had hit any of my public kgarner.com IPs the resolver will send me to the private ip instead of the public one i’ve already hit.  There’s also two simple scripts that you can integrate with OpenVPN to add and remove the /etc/resolver entry as needed.

The fact that OS X’s resolver will check for entries /etc/resolver first is the type of smart unix extentions I’d like to see more of.  There’s no reason Linux’s resolver can’t be doing something like this.  It would make VPNs easier to implement, and doesn’t seem to be that hard to add to the resolver code.

Other examples of OS X moving stuff forward is the init/cron/at all in one launchd.  I’m slowly starting to agree that init, cron, and at are all sides of the same coin.  Don’t get me wrong, launchd has some issues, but the idea is a step in the right direction, especially for machines that will sleep.  A lot of what OS X has done to make unix better is especially for mobile sleep-capable devices like laptops.

¹ Ask MARK for a laundry list of them…   

# grep –color=auto -i metadata todo.txt
Metadata Functions to Move:
   MetadataView…Make sure only our indexed items are passed up.

Its a very simple thing, but one of those that I’m surpised I haven’t been using.  I know have a shell aliases for that.  See the grep documentation for more information.

[Update 6/2: linux.com has a great article on GNU grep's new features which talks about the color.  One that I'm particularly expected about is the ability to use Perl-style regular expressions.]

a2ps -Eps -Afill -stumble 1.ps 2.ps 3.ps 4.ps

I tried this approach under OS X, but the problem is that the postscript that is generated on OS X is so detailed that it takes forever to process to print out, on the order of 2 minutes of processing per article.  Since PDF is the spooling format for printing in OS X (coming soon to linux) I thought I’d look to see if there was an easy way to concatinate PDF files so I could then have the regular printing interface (via Preview) handle the 2-up, double-sided, stapling goodness.

After much searching around I found this article and later this web page.  Combining a bit from both, I came up with following that works really well in my few days of testing.

texexec --pdf --paper=letter --pdfarrange --result all.pdf 1.pdf 2.pdf 3.pdf 4.pdf

It runs really quickly (especially in comparison to the a2ps method) and then I just open all.pdf and print from there.  It requires that you have teTeX installed.  On both Linux and OS X I had this installed as part of the prerequesets for docbook and doxygen.

rsync/unison/scp data off your co-loc to a local machine, and back it up again, just for good measure.

The new drive is the drive most likely to die first.

Losing /var really fucking sucks.

Recreating most of the information in ldap out of mail logs is cool, though.

Drinking doesn’t solve sysadmin nightmares, but it makes you feel good while you’re having the nightmare.

I’ve been pretty busy the past few months, so I was living without the RSS on my desktop since my move to FC4. I finally had some time to kill today, so I went about to get them up and running, and in RPM form. Since last time I installed them (version 0.7.4) Tugrul updated them to version 0.8.0. The spec file I used needed to be updated to match 0.8.0, to fix a small bug in the 0.8.0 build system, and account for differences in FC4. After a few moments of screwing around, I’ve got a spec file that works, a patch that works, and a built rpm.

This spec file is by no means perfect, as I’ve been learning how to build spec files on the fly, but it’ll work in most cases. I haven’t loaded it up with BuildRequires, but if you’ve got a modern desktop like GNOME or KDE installed and their devel packages, you should be in good shape.

I have to give credit to who originally put the spec file together, but I forget who that was. I’m fairly certain I got it from Tugrul’s page, but I can’t seem to find it again.

• rss_glx.spec
• rss-glx-0.8.0-1.fc4.src.rpm

When FC3 first came out, it didn’t have support for Firewire built into its kernel, so I’ve been hand compiling a kernel since then so I could use my iPod. When I installed FC4, I just used my .config and built a kernel for FC4. This was the source of my undoing and of losing a few hours debugging.

The show stopper that almost sent me back to FC3 was that I couldn’t get my Firewire based iPod working with FC4. I couldn’t find any reports of this problem, so I figured it had to be something unique to me. I finally found this series of posts from the HAL mailing list. This pointed out that it was a kernel problem.

I rebooted and dropped back to the latest Fedora-supplied kernel and lo-and-behold, it worked. So, it turns out that the way Firewire reports that device type in the kernel changed for the iPod. Its become more specific, which is not a problem, but it introduced a new type that the current version of HAL that comes with FC4 was unaware of.

The kernel used to report the iPod as (emphasis mine):

Vendor: Apple     Model: iPod              Rev: 1.53
Type:   Direct-Access                      ANSI SCSI revision: 02

It now reports it as:

Vendor: Apple     Model: iPod              Rev: 1.53
Type:   Direct-Access-RBC                  ANSI SCSI revision: 02

To fix this, I took the patch from that first post I linked to above and the hal src rpm from FC4, and merged them together and rebuilt. Boom, it works as it should. I’ve put a diff of my changes to hal.spec up, in case anyone else wants to redo this.

Today, Dave couldn’t figure out why he was unable to launch an X application from a machine with both use. We both started looking into it, and it looked like xauth wasn’t being called to update the .Xauthority file. We spent a good half hour or more looking around trying to figure out if it was a bug in OpenSSH on his mac, or one on the Linux server, if xauth was wonky, or what other small differences there were between his server side environment and mine.

During the search I found this post on a Debian mailing list. It was a red herring as it had us investigating a few dead ends. However, it did point out to my mind the existence of the ~/.ssh/rc file. Up until this point, I didn’t know of this files existence. Anyway, while looking in ~dave/.ssh/ I saw he had such a file.

To quote from the man page:

$HOME/.ssh/rc
Commands in this file are executed by ssh when the user logs in
just before the user’s shell (or command) is started. See the
sshd(8) manual page for more information.

There was an emacs backup file (rc~) there, which I looked into. At one time Dave used it to set a umask for all his connections that came in via ssh. For whatever reason, he must have decided that was not doing what he wanted, so he removed the umask line, but didn’t remove the file. Because the file existed, ssh was trying to execute the commands in it, and since there was nothing in it, ssh did nothing and dumped him to a shell.

From the behavior of ssh, it appears there is a “default” rc that happens if you don’t have one or one doesn’t exist in /etc/ssh. One of the tasks of this default includes calling xauth if you’re doing X11 forwarding. By having an empty file there, Dave was bypassing all of it. I haven’t taken the time to see what other side effects came about from that, but there must not have been much, as Dave hadn’t noticed it since last April (at least according to the mod time on the rc file.)

Dave just IMed me and told me to look at the sshd man page and see the following:

When a user successfully logs in, sshd does the following:
[snip]
8. If $HOME/.ssh/rc exists, runs it; else if /etc/ssh/sshrc
exists, runs it; otherwise runs xauth. The “rc” files are
given the X11 authentication protocol and cookie in standard
input.

One other thing we learned in this is that xauth is dumb dumb stupid. xauth won’t create a .Xauthority file if there is nothing to put into it, such as when you call “xauth list” when a file doesn’t exist. However, if you do an “xauth list” and you don’t have a .Xauthority file, xauth spits out a diagnostic message saying its creating it. In reality, it doesn’t really create the file. Bad coding on someone’s part. This wasted us time, as we though it was xauth that was broken, not Dave’s ssh environment. One could argue that xauth IS broken by demonstrating this behavior, but that’s a different rant.

However, at work I’ve switched over to using Thunderbird. I like the offline folder stuff that Thunderbird does, and I recently found a silly extension that just cracks me up: the Display Mail User Agent extension. In general, Thunderbird does what I want, but I find myself missing a few features from mutt that I really wish was there.

• The ability to arbitrarily attach other mail messages. In mutt, I can tag a bunch of messages and do a tag-forward and they are all attached to the forwarded mail. Also, from the compose menu in mutt you can do a shift-A aka attach-message navigate to a mailbox, tag messages, and quit back to the compose mail, and they are attached. For the life of me, in thunderbird, I cannot figure out how to attach a message, much less a whole thread, to an arbitrary mail. I can forward a piece of mail as an attachment, but I cannot forward n>1 messages. I often send interesting threads from mailing lists to friends and co-workers and this crimps my style.
• The ability to do a reply to multiple messages and have them all quoted. Similar to the above situation, I can tag multiple messages in mutt and do a tag-reply or tag-group-reply. This feature quotes all the tagged messages as well as add all the appropriate people to the recipient list, for tag-reply this is the senders of the original mail, for tag-group-reply this is everyone who was listed as a recipient in any of the mails as well as the senders. While I don’t often use it to hit people scatter shot, I do often tag-reply to one person, when appropriate, if they have send me multiple messages in the time its taken me to get back to them.
• Custom e-mail headers. This one is purely for amusement value and is nothing that would keep me on mutt over anything else. But it is fun, and I do have a lot defined. In any typical mail you might see something like this if you look at the headers:
  

  X-If-I-Knew-Better-I-Would-Not-Be-Running: Linux 2.6.12 i686
  X-Time-Married: 3 years, 11 months, 14 days, 4 hours, 21 minutes, 42 seconds
  X-Dinah-Lives: 1 year, 3 months, 21 hours, 38 minutes, 42 seconds
  X-Uptime: 20:48:42 up 222 days, 28 min, 5 users, load average: 0.01, 0.07, 0.07
  X-The-Amount-Of-Stuff-In-My-Inbox: 92

  They are all silly, but they keep me entertained, and no one really sees them. However, every once in a blue moon I really give someone a chuckle.

Another feature that would be nice to see in thunderbird doesn’t come from mutt but from Apple Mail in OS X. This isn’t a make or break thing since I’m not used to it, but it is a nifty feature I admire.

• The ability to multiple select mailboxes and have them mingled and threaded properly in message header/index pane. Apple Mail does this, and its very slick. Its another one of those features you don’t use all the time, but it would be often enough to make it worth while. Being able to select INBOX, Sent, and a mailing list would be nice to see the some threads properly threaded all the way through when searching for a message.

Hopefully someone has already written extensions to do this and I just need to be pointed in the right direction. If not, I’ve considered writing some of them myself, but I lack the knowledge of XUL and other technologies I would need. Anyone know any good HOWTOs to get me started? (Oh, and a device to stop time so I have time to actually work on it?)

[Update 8/16: This is a mad knowledge bomb from Mark via IRC:

15:06 <spruance> hey, keith
15:06 <spruance> thunderbird can already forward an arbitrary number of messages
15:06 <spruance> drag those messages to the "send" area of your compose window and they'll be attached
15:06 <spruance> grab a thread handle to take the whole thread
15:07 <spruance> Similarly, you can highlight more than one message and right click and choose "foward as attachment"

Well, the arbitrary forwarding works, I just must never have multiply selected images. However, under Fedora Core 3, with Thunderbird 1.0.6-1.1.fc3 (20050720) the dragging the message into the compose/address aread doesn't work. The attachment box opens, but no messages are attached. Mark reports this works on Mac and Windows. Damn Linux.]

I’m really excited about our beta release of ezRETS an ODBC driver for RETS data sources. This has been one of the coolest projects I’ve worked on for a long time.

We just put out the windows binary today, but the source is available via subversion and if you can build it, it’ll run under Linux or Windows. OS X to follow. We will probably get out a source release next week, we just didn’t have time this week.

—

One of the other things that drew us to the museum was the Game On exhibit. Unfortunately, you couldn’t take pictures with a flash or tripod, so my images from it are a bit blurry.

In any case, it was a nice walk down memory lane, and it had a few things in it I’ve read about and never seen. They had an original Pong machine. You could play Pong, just not on the original cabinet. They had recreated/reconstituted guts set up in a different case. I assume it was to protect the original machine.

The coolest piece of history was a PDP-1. Of course, it wasn’t running, but it was sitting there and they had some information on Spacewar. If you don’t know about Spacewar you can read The origin of Spacewar on-line or Hackers: Heroes of the Computer Revolution by Steven Levy

They also had a MAME machine set up with controls a good distance from the display projected on the wall. It was a pretty sweet setup, I may have to get one for home. The MAME machine was supposed to be running Dave’s Game Launcher front end for emulators. Unfortunately, the controls weren’t responding, so I couldn’t verify. Dave said other people he knows have gone there and verified its Game Launcher. I’m sorry I missed it.

There was a small section on console companies that released products that let you write your own games. Nintendo apparently released such a kit in Japan. However, as part of this display was the Linux for the Playstation2 kit that Sony put out. As a Linux geek, I was surprised to see it. It was also fun to see something I own in a museum.

There was a round table with portable games arranged in chronological order. I was inspired to take the picture above where the newest kid on the block was meeting the old timer. I’m really way too amused by it.

Unfortunately, many of the games and consoles were in bad shape. They’ve taken a lot of abuse by all the visitors to the exhibit. But there was enough playable to keep you busy for hours. There’s also many games that were a joy to see and play for a short bit again, but I won’t mention them here. I’ll leave some mystery for your visit.

I’m just glad I got to play two of my favorites from back in the day: Discs of Tron and the old sit-down vector Star Wars game.

Anyway, I just wanted to drop a note that I figured out last night how to specify any arbitrary mount option. I suppose it was possible all along, but last time I went down this trail, my brain was getting fried and thought only the mount options I saw in examples were possible. It turns out you can do any mount option as long as you specifiy it as a bool and set it true or false. You can also turn off default ones.

For some reason, for vfat filesystems, hald‘s default .fdis are set up to mount fat as UTF-8 by using the mount option of iocharset=utf8. This was causing some problems with gtkpod, so I wanted to remove that. Looking at some examples, I deceided to finally try specifing this in my iPod.fdi:

<merge key="volume.policy.mount_option.iocharset=utf8" type="bool">false</merge>

This worked, so I thought I’d try some other mount options I would have liked, but couldn’t figure out how to get going my last go around with the iPod.

<merge key="volume.policy.mount_option.shortname=win95" type="bool">true</merge>
<merge key="volume.policy.mount_option.noatime" type="bool">true</merge>

Those also worked. It looks like you can put anything after volume.policy.mount_option. as long as you specify the full option and set it to true.

If you want to recreate what I’ve done, take my .fdi from the previous post add the above lines to it for the second parition on the iPod and you two can get fstab-sync to create an fstab entry that looks like this:

/dev/sda2 /media/iPod vfat pamconsole,exec,noauto,noatime,shortname=win95,sync,managed 0 0

With FC3 comes some new ways to get external devices going easier. udev, which “is an implementation of devfs in userspace using sysfs and /sbin/hotplug,” used to be an add on, but is now a central part of the distribution. udev is configured in such a way that it controls all of the /dev entries. Along with udev is the addition of freedesktop.org‘s Hardware Abstraction Layer software. This isn’t like the HAL in some other operating systems and based on the documentation that is not the goal. From the documentation:

HAL which is a piece of software that provides a view of the various hardware attached to a system. In addition to this, HAL keeps detailed metadata for each piece of hardware and provide hooks such that system- and desktop-level software can react to changes in the hardware configuration in order to maintain system policy.

Its not the world’s most useful thing yet, but considering FC3 ships with version 0.4.2 its doing a good job for its level of maturity.

One of the most useful things that HAL provides FC3 is the auto configuration of devices that would have removable media, such as cdrom drives, or is added via USB or firewire. In its default configuration is works pretty good: udev creates the /dev nodes and HAL, after scoping out and probing the hardware, creats mount points in /media and calls /usr/sbin/fstab-sync (which comes with HAL) to edit /etc/fstab with the entries. For my devices that have removable media, HAL has noticed my CD and DVD burners and has created /media/cdrecorder and /media/cdrecorder1. If I connect my iPod up via firewire, I see that /media/ieee1394disk is created. Putting a compact flash card in a slot of my 8-in-1 USB media reader, I see that /media/usbdisk is created.

As I said before, this is pretty good. However, there were a few things that bugged me that I wanted to change. The first problem I ran into was that hald, constantly probes attached media devices for media changes. This works great for my 8-in-1, as it correctly drops access to the device when I run /usr/bin/eject on the mount points. However, this fails dismally with my iPod.

In FC2, to get the iPod to go into the “OK to remove” mode, the eject command did the trick. In FC3, thanks to hald, it tries to spin up the iPod harddrive again. However, the iPod, after its in “OK to remove” mode will not respond to any commands from the computer until removed from the cradle. hald trying to contact it again causes the SCSI subsystem to go into a loop trying to reset the device and the SCSI bus, causes hald to hang, and generally really pisses off Linux. Since this whole ordeal ends with the SCSI subsystem kernel panicing and hald being stuck in a disk wait, the only thing to really get things back to normal that that point is a reboot. The reboot then has some issues thanks to hald being in disk wait, and Linux can’t cleanly unmount /usr.

So, obviously, I wanted to fix the “everything goes to shit when I want to remove my iPod” problem and I wanted to create a better mount point name for the iPod. Face it, /media/ieee1394disk is pretty weak and /media/iPod is much sexier.

I started digging around the hal package to see what docs came with it and what the config files were. The documentation that ships in the RPM is fairly non-existant except for some sample Device Information Files aka .fdi files. Next I looked in /etc/hald/hald.conf. That had a directive called storage_media_check_enabled which was set to true. Setting that to false and halfway fixed my iPod problem, except now two mount points were showing up in media. Also, setting it to false broke the 8-in-1 which was behaving perfectly. Not wanting to throw the baby out with the bathwater, I set it back to true. This setting, however, started to point me in the right direction.

Next I discovered the hal-gnome package by doing an apt-cache search hal. From this I found the misnamed /usr/bin/hal-device-manager. I say its misnamed as you get a Windows-style device manager view, but you can’t actually manage anything, its totally a read-only view. However, this let me see the properties that were being set for each device and its volumes along the way. Thanks to hal-device-manager I was starting to be able to read the .fdi files. However, my next question was “what are all the possible values of propery names that can be set?” For that I needed more documentation.

I visited the freedesktop.org hal site (as linked above) and it didn’t have anything really useful on it. I was getting motivated enough to just start reading C code, so I downloaded the hal-0.4.4 source from freedesktop.org. To my delight, there was actual useful documentation, albiet in raw docbook XML form. I almost cried, but I pulled myself together and typed make to build the documentation into a useful HTML file. Here was a list of all the possible properies, where they were good, and mostly good descriptions of what they did. I still didn’t have a full enough understanding, so it came time to start experimenting.

In my experimenting, I focused totally on making the iPod work the way I wanted and figured I’d tackle the 8-in-1 later as it mostly works the way I want now. The existing .fdi files have a comment that says to create system local files in /usr/share/hal/fdi/95userpolicy as all the other files in the other directories could be overwriten when/if the vendor updates the hal package. I started to write an iPod.fdi file and based my <device> entry on some of the existing firewire disk entry and merged it with the sample firewire connected hard drive example from /usr/share/doc/hal-0.4.2/conf

What I ended up with after my experimenting is something that seems to work really well. It turns off the media check just for the iPod, ignores the first partition used by the Apple Firmware, and makes a mount point at /media/iPod for the second partition where all the music lives. I also was able to add the sync mount option to the automatically created fstab entry. Lots of iPod on Linux people suggest mounting the iPod in sync mode, but I forget exactly why at this moment as my brain is running out of juice as I’m ending of this post.

This was through a bunch of trial and error, and if someone has a better way to do it, please let me know.
Here is my /usr/share/hal/fdi/95userpolicy/iPod.fdi

<?xml version="1.0" encoding="ISO-8859-1"?> 
<deviceinfo version="0.2">
  <device>
    <match key="storage.vendor" string="Apple">
      <match key="storage.model" string="iPod">
        <merge key="storage.requires_eject" type="bool">true</merge>
        <merge key="storage.removable" type="bool">false</merge>
        <merge key="storage.media_check_enabled" type="bool">false</merge>
      </match>
    </match>
    <match key="@block.storage_device:storage.vendor" string="Apple">
      <match key="@block.storage_device:storage.model" string="iPod">
        <match key="block.is_volume" bool="true">
          <match key="volume.fsusage" string="filesystem">
            <match key="volume.partition.number" int="1">
              <merge key="volume.policy.should_mount" type="bool">false</merge>
            </match>
            <match key="volume.partition.number" int="2">
              <merge key="volume.policy.desired_mount_point" type="string">iPod</merge>
              <merge key="volume.policy.mount_option.sync" type="bool">true</merge>
            </match>
          </match>
	</match>
      </match>
    </match>
  </device>
</deviceinfo>

But the reality is, this case is about a contract dispute. Linux got dragged into it as hostage. SCO is holding Linux by the neck, pointing a gun at its head, and telling IBM, “Do what I tell you, or I’ll shoot your little friend.”

FC2 came with a udev rpm, and I’ve been meaning to play with it for awhile. udev is a userspace way to have a dynamically managed /dev directory. In theory, you’re supposed to have it manage your /dev directory. However, the default config from the FC2 rpm is to set up a seperate directory called /udev. I actually like this decision. You end up getting /dev with its everything-including-/dev/kitchen-sink directory for devies that never move, which works better for some drivers like the NVidia drivers, and /udev reflects only what’s on your system.

One of the other cool things you can do with udev is set up persistant names for dynamic devices. I’ve got it set up so my iPod always shows up as /dev/ipod and its partitions as /dev/ipod1 and /dev/ipod2. This has made it really easy to set up in fstab. I’ve also done similar with my 8-in-1 digital media reader. Compact flash, smartmedia, etc all show up with sane device names in the /udev tree.

In the future, I assume that a full ruleset for udev will be created, in the mean time, I had to create my own /etc/udev/rules.d/60-ktg.rules file that looks like this:

BUS="scsi", SYSFS{model}="iPod ", NAME="ipod%n"
BUS="scsi", SYSFS{model}="USB Storage-SMC ", NAME{all_partitions}="sm"
BUS="scsi", SYSFS{model}="USB Storage-CFC ", NAME{all_partitions}="cf"
BUS="scsi", SYSFS{model}="USB Storage-MMC ", NAME{all_partitions}="mmc"
BUS="scsi", SYSFS{model}="USB Storage-MSC ", NAME{all_partitions}="msc"

I don’t know if these rules are 100% correct, but they are working well enough for me now.

The other thing that really helps with these hotplug firewire/usb devices is that the SCSI layer in Linux Kernel 2.6 became hot pluggable, so I don’t need to run a shell script every time I add or remove one of these devices that act like a SCSI device.

All in all, once you get over the “broke XP” and “no firewire” problems, FC2 is a really solid distribution that I’m happy with. Of course, part of my happiness is due to the other apt/yum repositories I’m using, but FC2 plus those is right where I want my Linux experiance right now.

I started looking into how java finds the default information via java.util.TimeZone. It tries to use System.getProperty("user.timezone"), if that fails then it then calls a native method passing along user.country and user.home. So, that dead-ended me.

Seva found the only place that “America/Chicago” was being used was in /etc/sysconfig/clock which only appeared to be used in startup and shutdown routines when calling hwclock. He changed it so the correct value was being put in ZONE, and Java acted how it should.

I ran strings on $JAVA_HOME/jre/lib/i386/libjava.so and found references to /etc/sysconfig/clock as well as ZONE=". So, it appears, at least on Linux systems, Java will read /etc/sysconfig/clock and if that’s succesful, not even bother checking with /etc/localtime.

This, I feel, is totally incorrect behaviour and against what one would expect from a UNIX system. This should probably be changed, but most likely won’t be. I figured I’d drop a note about it here in the blog to help others in the future as we couldn’t easily find the defined behaviour on the web.

Seva complained that he would have found this much faster using strace, but on the RHEL3 box he was using, strace segfaulted when he run against java. Works fine on my FC1 box, and I could recreate it and see it. So I guess this is also Linux dumbness of the day.

So, the install went okay, its pretty much the same as FC1s. As the install went, I noticed a lot more java support, mostly with the base of gcj and stuff from the GNU Classpath project. (As a total aside, from the Classpath homepage I see there are a few open source JVMs I had no idea existed. I’ll have to check those out.) Of course, due to licensing, no official VM from Sun.

I hit a snag right away with FC2 when it came to video drivers. Since I have an Nvidia card, I generally use the proprietary drivers from Nvidia because I like eye candy like 3D screen savers (especially the Really Slick Screensavers) and things like Unreal Tournament. FC2′s default 2.6 kernel is compiled with the “4k stacks” option turned on. This totally braeks Nvidia’s current driver for linux, which appearently needs 8k stacks. Nvidia said its next driver release will play nice with 4k stacks, but they aren’t out yet.

The second snag is that there is no firewire support out of the box. It appears that the 2.6 kernel that FC2 ships with had really broken firewire support, so they turned it off. For most PC users, this probably isn’t a big deal. For me it is, as I have an iPod that I use under Linux with gtkpod. Currently, my firewire card exists just to talk to my iPod so this is disheartening. (In the near future, I’ll probably be getting a camcorder with firewire out, if only to make DVDs of my child.)

In either case, both of these problems can be overcome by recompiling the kernel by getting the latest clean/Linus version of the kernel. Unfortunately, for some users this is too difficult of a task. Since I’ve been building kernels since Jan 1994, I suppose I can live with it. It also appears that Fedora released a new kernel rpm today that addresses the firewire problem. However, the nvidia issue is still there. So it looks like I’m hand compiling.

Chris Stamborski reports that he couldn’t get pvmove (an LVM command) to work with FC2 due to something in the 2.6 kernel. However, earlier today he told me on IRC he found a fix for it. I’m bugging him to get a reference on it. I don’t need it currently, but it never hurts to have the info.

On the plus side, FC2 might have been worth upgrading if only for the ATAPI CD/DVD burner DMA support. I burned a few CDs last night at full 32x and my system didn’t even break a sweat. With a 2.4 kernel, it bogged down the system something fierce to where I couldn’t web surf. The whole PIO access to hard drive just kept the system too busy. It felt like when you’d format a floppy under Windows 3.1.

Also, it seems snappier on the same hardware. It probably has to do with scheduler and IO scheduler fixes.

To sum it up, I’m mostly happy with FC2, and I’m thinking about putting it on my laptop. I’ll be even happier after a kernel compile and I have firewire again.

To sum it up, I’m really impressed with EL3. It seems like a really well QAed version of RH9, which was a solid distribution.

To keep the maintenance on the box down and to make it easy for someone else to pick up should I fall off the wagon, I’ve made it a point to stick with RPMs and redhat tools as much as possible.

The RPMs for apache and php were compiled just how I wanted them to be. (I’m not sure if this means RedHat in general is getting better, or I’m just less picky.) I was able to get gallery and SSL working out of the box. Gallery was a good test app, since half of us using the box are moving our photos there.

This is my first experiance using up2date, and its not too bad. Its certainly not apt-rpm or yum but its not a complete pain to use.

The only things I’ve really had to do by hand were Tomcat and its apache connectors (because it just doesn’t ship with redhat) and Postfix because I wanted a more up to date version. Also, any webapps that would be shared by everyone such as Jamm and SquirrelMail. Okay, that was a bit more than I thought, but most of those are setup and forget, until a security exploit comes along.

Actually, I’ve added courier-imap by hand as well, but I grabbed a SRC RPM and used that. I’m not sure if that really counts as “by hand” but it certainly wasn’t a standard package from RH.

At work, our IT group is looking to move a bunch of stuff over to Linux. I feel comfortable advocating EL3 thanks to my experiance here. There’s really no other choice for an IT dept. like ours.