Ubuntu Lucid on (X86) Mac Mini with EFI

I was gonna use my Mac Mini as replacement home server (uses less power than an idle Core 2 Duo + Geforce 8800GTS + 4 Harddisks), so I decided to put Ubuntu Lucid on it. Installation was really easy. I used the Lucid Lynx Beta 1 Server install CD, and installation went just fine. Press/hold c to boot from CD, be sure to create a seperate boot partition or keep some free space to create a seperate boot partition afterwards.

Just installing Ubuntu is no issue at all. You don't even need to create a seperate boot partition, because it'll boot just fine using the Mac's legacy booter. However if you're a Mac Mini owner and want to boot headless, there is only one solution (and a workaround that uses a 'dummy monitor dongle', which is not what I would want). This solution is making use of the EFI features of GRUB 2, which I'll detail in the rest of this post.

When the installation is done, install hfsplus and hfsprogs so you're able to create HFS+ volumes (you probably only need one of those, but it wasn't exactly clear for me which one to use and due to lack of time I haven't looked any further yet). Copy files from your boot partition to a temporary space, then unmount and format the boot partition as HFS+ (mine is /dev/sda2):

# mkfs.hfsplus -v boot /dev/sda2

Edit /etc/fstab to use the new boot partition as boot partition:

...
/dev/sda2 /boot hfsplus defaults 0 2
...

Unfortunately it seems UUID's can't be used for HFS+ volumes at the moment, so hardcode the device name in there.

After creating the volume mount it again and copy over all the files from your temporary space.

Now install grub-efi, this will automatically remove grub-pc. Generate a GRUB EFI executable using the following command:

# grub-mkimage -o /boot/grub/grub.efi -p /grub part_gpt hfsplus fat ext2 normal sh boot configfile linux

You actually don't even need the fat and ext2 modules and probably more can be stripped, but I haven't experimented with GRUB 2 a lot yet, that'll be for another day (and another post).

We're almost done, now it's on to getting the Mac to actually use our EFI enabled GRUB. First toggle the boot flag on your shiny HFS+ partition. Run parted (or any other GPT-aware partitioning tool) and type:

(parted) set 2 boot off

Parted will probably tell you the disk is in use and you need to reboot for the change to become effective, but it doesn't matter for us.

Now the final step is to tell the Mac (or actually, the filesystem) that our grub.efi is bootable so it'll show up in the Mac boot menu. There should be a utility call hfspbless, which allows you to do this from within Linux, however the first hit on Google doesn't seem to offer a quick guide, so I skipped this part. Instead, put in the Mac OS X install DVD. As soon as a menu bar shows up (I believe you have to click next at least once), fire up a terminal. In the terminal, enter the following:

# mkdir /Volumes/boot
# mount_hfs /dev/disk0s2 /Volumes/boot
# bless --folder=/Volumes/boot --file=/Volumes/boot/grub/grub.efi --label boot --setBoot

The bless command has now set some metadata on the HFS+ filesystem that the Mac uses to identify a native bootable image. I assumed label should set the label accordingly for the boot menu, however my entry showed up as 'EFI something' IIRC, but I can't care more or less since it's a server and I'll never see the menu anyway. Now reboot and enjoy!

This did the job for me, however there are a few issues I still have to take care of. The Mac created a fake MBR partition map for me, which I don't need and don't use. It now shows 'Windows' as option in the Mac boot menu, but luckily it starts Linux by default. Also, the MBR partition map is out of sync if you do stuff with the GPT partition table. I used rEFIt to resync the MBR table, but when I figure out how to remove the MBR that's what I'm going to do.

Also, there's a file called Volume Name Icon on my boot partition. I guess this is used for the Mac boot menu, so it probably can be changed easily too. However I have no clue what the format is, I'll have to look this up some day and change it for a genuine Tux!

Installing Alfresco on Ubuntu Jaunty (9.04)

Some of the information in here comes from http://wiki.alfresco.com/wiki/Installing_Alfresco_Community_WAR_on_Centos_5. Even though the guide is old, most of the information is still correct, albeit some file locations and names have changed and it's written for another distribution.

1. Start with updating your system

   
   # apt-get update




2. Install tomcat and mysql-connector

   
   # apt-get install tomcat6 libmysql-java




3. Edit tomcat startup settings

   
   

   # /etc/defaults/tomcat6
   
   #JAVA_OPTS="-Djava.awt.headless=true -Xmx128M"
   JAVA_OPTS="$JAVA_OPTS -Xms512m -Xmx512m"
   
   #TOMCAT6_SECURITY="yes"
   TOMCAT6_SECURITY="no"
   

   
   Take note of the fact I disabled security, otherwise you need to create a policy file with everything that is allowed in it. I guess it is not that hard, but I wanted to get Alfresco running at all first.




4. Create the Alfresco root

   
   # mkdir /opt/alfresco
# cd /opt/alfresco
# wget http://dl.alfresco.com/release/community/build-2440/alfresco-community-wcm-3.2r2.zip
# wget http://dl.alfresco.com/release/community/build-2440/alfresco-community-sample-extensions-3.2r2.zip
# wget http://dl.alfresco.com/release/community/build-2440/alfresco-community-war-3.2r2.zip
# unzip alfresco-community-war-3.2r2.zip

# mkdir wcm
# unzip http://dl.alfresco.com/release/community/build-2440/alfresco-community-wcm-3.2r2.zip -d wcm




5. Create Alfresco database and user

   
   $ mysql -u root -p < extras/databases/mysql/db_setup.sql




6. Create Alfresco and tomcat directories

   
   # mkdir -p /var/lib/alfresco/alf_data/
# mkdir -p /var/lib/tomcat6/shared/{classes,lib}




7. Add Alfresco and Alfresco share wars to tomcat

   
   # ln -s /opt/alfresco/alfresco.war /var/lib/tomcat6/webapps/
# ln -s /opt/alfresco/share.war /var/lib/tomcat6/webapps/




8. Add mysql connector to path where tomcat finds it

   
   # ln -s /usr/share/java/mysql-connector-java-1.5.6.jar /var/lib/tomcat6/shared/lib/mysql-connector-java.jar




9. Add extension sample files to tomcat

   
   # unzip alfresco-community-sample-extensions-3.2r2.zip -d /var/lib/tomcat6/shared/classes/




10. Setup Alfresco global settings

    
    cp /opt/alfresco/extensions/extension/alfresco-global.properties /var/lib/tomcat6/shared/classes/
    Edit the contents of the file:
    

    # /var/lib/tomcat6/shared/classes/alfresco-global.properties
    
    #dir.root=./alf_data
    dir.root=/var/lib/alfresco/alf_data





11. Add WCM bootstrap

    
    # cp wcm/wcm-bootstrap-context.xml /var/lib/tomcat6/shared/classes/alfresco/extension/




12. Setup catalina loader paths

    
    

    # /var/lib/tomcat6/conf/catalina.properties
    #shared.loader=
    shared.loader=${catalina.base}/shared/classes,${catalina.base}/shared/lib/*.jar





13. Fix permissions

    
    # chown -R tomcat6:tomcat6 /var/lib/{tomcat6,alfresco}




14. Do a first run so the wars get extracted

    
    # /etc/init.d/tomcat6 restart




15. Setup log file

    
    

     # /var/lib/tomcat6/webapps/alfresco/WEB-INF/classes/log4j.properties
    
    #log4j.appender.File.File=alfresco.log
    log4j.appender.File.File=/var/log/tomcat6/alfresco.log





16. Restart tomcat so log settings are re-read

    
    # /etc/init.d/tomcat6 restart

Now you should be able to reach Alfresco on [ip]:8080/alfresco and Alfresco Share on [ip]:8080/share.

Update:
Seems that Alfresco prefers to run on OpenJDK for me, getting out of memory errors when using Sun's JDK. However, if OpenJDK is installed on Jaunty, there's a symlink for rhino in /usr/lib/jvm/java-6-openjdk/jre/lib/rhino.jar, which prevents loading of rhino included in the Alfresco WAR and will result in errors in Alfresco Share (and probably other places too). A # rm /usr/lib/jvm/java-6-openjdk/jre/lib/rhino.jar fixes this.

Convert PDF to ... PDF

I keep running into people having issues with PDF files. Most of the times there's a very easy solution to fix their issues: convert the PDF to a PDF.

The following ghostscript command will successfully convert a PDF to a new PDF without any restrictions, suitable for printing, editing with PDF editors or whatever you want:
gs -dSAFER -dBATCH -dNOPAUSE -sDEVICE=pdfwrite -dPassThroughJPEGImages=true -sOutputFile=out_file.pdf in_file.pdf.

When something's broken...

I've spent a large amount of today figuring out how to get some MP4 files playing on my TV (Samsung LE37B650T2, with DLNA support). Turned out that my TV doesn't like a MP4 file with the video track before the audio track, so I now add the audio tracks first. That at least got one movie playing, I guess I can get more to play this way.

Moving files in Banshee library

I moved most of my music to the localized music folder (yay for localized name fail). Of course Banshee didn't know I moved them, so I had to tell it they were moved. I knew Banshee uses sqlite, so I just had to find the db and do the replacement. Banshee's database can be found in ~/.config/banshee-1/banshee.db. The following block highlights all you need to do to change the path for your files:

$ sqlite3 ~/.config/banshee-1/banshee.db
SQLite version 3.6.10
Enter ".help" for instructions
Enter SQL statements terminated with a ";"
sqlite> UPDATE CoreTracks SET Uri = replace(Uri, '/home/mycroes/Music/', '/home/mycroes/Muziek/') WHERE Uri LIKE 'file:///home/mycroes/Music/%';
sqlite> 

Of course you need to pass in the correct arguments to 'replace', but that should be an easy one.

Watching TV with mplayer

Occassionally I want to watch something on my computer, and currently I'm using mplayer to do so. When I finally was able to play video and sound at the same time, I decided I wanted to have a bit easier way to watch tv then constantly searching for my mplayer command. So I made a small shell script, and while I was at it I added in the channel list so I can just start watching without searching frequencies too, so here it is:

#!/bin/bash

#OPTS="-fs"
FILTER="-vf pp=lb,crop=672:420"
OUTPUT="-vo gl -ao pulse"

CHANNELS="768000-Nederland_1,\
776000-Nederland_2,\
784000-Nederland_3,\
752000-RTL_4,\
744000-RTL_5,\
736000-SBS_6,\
728000-RTL_7,\
712000-Veronica,\
704000-Net_5,\
496000-Discovery_Channel"

TV="-tv outfmt=i420:chanlist=europe-west:width=720:height=576:amode=1:alsa:adevice=hw.2,0:forceaudio:immediatemode=0:channels=$CHANNELS"

mplayer $OPTS $FILTER $OUTPUT $TV tv://$1

As you can see I'm using the first argument as channel, so you can invoke this script with a channel number to immediately select a channel on startup, but leaving out the argument will just start mplayer on the first channel.

Last but not least, you can switch channels by binding keys to tv_step_channel 1 and tv_step_channel -1, in Ubuntu these are bound to h and l by default, and I suspect this to be the upstream default.

Magic comments

So you're writing some code, and there seems to be an error somewhere. As a real programmer, you're going to use printf or echo to debug your code. I happen to be writing a lot of PHP lately, so I also debug with var_dump. Not that it really matters, because after all it boils down to eliminating lines with errors, and as long as you use a programming language that supports // for single line comments and /* ... */ for multiline comments you can use magic comments.

Honestly, there's no magic involved, just a bit of logic. Let's say there's some code consisting of 3 blocks of lines:

[block 1]

[block 2]

[block 3]

If I want to comment one block, I could simply put /*, */ around it like this:

/*
[block 1]
*/
[block 2]

[block 3]

However, soon enough I figure I needed to comment block 2 for a while. Because I'm lazy, I try to type no more than needed, so it becomes this:

//*
[block 1]
/*/
[block 2]
*/
[block 3]

I'm almost happy now, but it seems I need to switch back to having block 1 commented again instead of block 2... Of course I can remove the added characters, or I can just add a few more:

/*/*
[block 1]
/*/
[block 2]
/*/
//*/
[block 3]

Now if I decide that I need to comment block 2 again, I only need to edit one character:

//*
[block 1]
/*/
[block 2]
/*/
//*/
[block 3]

And it's actually possible to chain these by using a /*/ between 2 blocks. If the block in front was commented, the next block won't be commented. If the block in front however was not commented, this block will be commented. And there you go, boolean logic with comments, it's almost magic.