Ubuntu: add personal path to the global path

My best choice when I want to extend the system path is to create a profile shell script in the /etc/profile.d/ directory with and extension of .sh. The content is something like this:

export PATH=$PATH:/home/ikon/bin

This way my ~/bin directory is added to the system path.  Check it with

$ echo $PATH

Problem installing debugger on 1.9.3 and RVM

This is a recent problem now, so if you are reading it years later try to adopt your numbers on these lines.

Version ruby 1.9.3-p327 using RVM, debugger and debugger-linecache restisted the bundle install command.
Workaround:

gem install debugger-linecache -v '1.1.2' -- --with-ruby-include=\$rvm_path/src/ruby-1.9.3-p374

gem install debugger -v '1.2.3' -- --with-ruby-include=\$rvm_path/src/ruby-1.9.3-p374

Vboxheadless and bridge nic trouble

I wanted to have my VM built on Windows to be hosted on an Ubuntu linux Virtualbox headless. The VM had a bridged interface and when I tried to fire it up after a plain import I got this:

Error: failed to start machine. Error message: Configuration error: Failed to get the "Trunk" value (VERR_CFGM_NOT_ENOUGH_SPACE).

Failed to attach the network LUN (VERR_CFGM_NOT_ENOUGH_SPACE)

The solution is simple:

1. check the host's main network interface name (or the one you want to bridge to), e.g. eth1
2. modify the VM's bridge interface to that with:

vboxmanage modifyvm vm_name --bridgeadapter1 eth1

After this adjustment is done and the bridge interface knows where to point to the VM will surely fire up.

Missing menubar in (K)Ubuntu 12.10

After updating to and installing 12.10 of Kubuntu my GTK apps are missing their main menubars. Somewhere I read a cool solution that seems to work right offthe bat:

sudo apt-get purge appmenu-gtk appmenu-gtk3 appmenu-qt

ORACLE JDK on Linux

This is a small summary for anyone who struggles often - on Ubuntu recently - to get the so called "Sun Java JDK", now belongs to Oracle onto their system. My strategy is to exclude any kind of Java packages from my installation and download only the self extracting JDK from the Oracle website. Once downloaded and extracted into my home directory I can easily  reference that directory in the various global variables that my softwares are searching for.
On Ubuntu I use a dumb-simple shell script that is put into the /etc/profile.d directory with the extension of .sh. The content of my version is like this:

export WEBIDE_JDK="/home/ikon/app/jdk"export RUBYMINE_JDK="/home/ikon/app/jdk"export SMARTGIT_JAVA_HOME="/home/ikon/app/jdk"export SMARTSYNCHRONIZE_JAVA_HOME="/home/ikon/app/jdk"

Here the /home/ikon/jdk is where I extracted the JDK to. The variable names can be looked up from the starter shell scripts of the softwares. This way I even made those softwares "portable" :)

Using the same icon theme for GTK under KDE

I wanted to use Clearlooks on both Kde and within Kde all the GTK apps. But GTK apps started to use the default, ugly gnome icon theme. So I made this change to the following file, called .gtkrc-2.0-kde4 and back to business, using Oxygen icon theme everywhere ^_^.

Create desktop launcher in Ubuntu 12.04

There is a certain flaw in usability for power users in the new Ubuntu 12.04 out-of-the-box. I cannot create a desktop shortcut as easy as it was used to be.
Instead, I have to install gnome-panels package and then use this command:

gnome-desktop-item-edit Desktop --create-new

to get a popup that was so natural it is a crime to miss it.

Find files and do something with them

• Find removed (deleted) files still in use via /proc:

# find -L /proc/*/fd -links 0 2 > /dev/null

• Remove .svn folders recursively. (I have seen solutions using the -exec switch, but it is too slow, so I use find's own delete option here)

$ find . -name ".svn" -type d -delete;

• Find and delete files with multiple choice of names:

$  find /path -name '.DS_Store' -or -name '._*' -delete

•  Finding file in .h with "raoul" in it ( the "-n" is for displaying line number ):

$  find /path -name "*.h" | grep -n "raoul"

• Find empty folders:

$  find /path -type d -empty

• Find empty folder and list:

$  find /path -type d -empty -ls

• Find empty folder and save result in temporary file:

$  find /path -type d -empty -fls /path/to/tempfile.txt

How to configure a local subdomain development with Linux

Note

In the following scenario everything is done within one machine. This setup can be recreated with more then one machine as well. Use your imagination!

Task

I want to have subdomains like static.mydomain or whatever.mydomain.

Ingredients

• any host OS that is capable of running Virtualbox, VMware or your favorite virtualisation software
• a linux distribution that uses/supplies a small kernel, like Ubuntu server, Fedora without X, Arch or any of your choice, preferably one that you are familiar with. Of course if you are not a linux guru you can  still use a gui, just give enough memory for your guest linux.

Setup process

1. have your virtualisation software set up on your host
2. install your linux distro of choice as a guest, it is best to use at least one "Bridged" (my choice) or "Host only" network adapters that every other guest machine can see too if you later need it. This is a preview of a setup choice.
   
3. set up your /etc/resolv.conf file to have your nameservers in the following order:
   (NOTE: only the first one has to be your localhost, the remaining two is your choice of providers)
   
   nameserver 127.0.0.1
   nameserver 8.8.8.8
   nameserver 8.8.4.4
   
4. install the package that has dnsmasq in it or compile it from source.
5. setup dnsmasq according to the example below:

dnsmasq configuration

Generally the config file is at /etc/dnsmasq.conf.

Let's say you want to have a domain that is resolved by your machine -who take part in the development- as one of the dev machines address on the local network. For example I want to have the domain ikondev to be resolved to the ip address of VM1 (Virtual Machine 1 on the above image). This way we can have several subdomains for ikondev without updating /etc/hosts file each and every time we add a new subdomain to our webserver to listen to.

In your dnsmasq.conf set the following lines: (these lines has to be uncommented too)

• domain-needed
• bogus-priv
• strict-order (this will make this local DNS server to be the main nameserver)
• address=/ikondev/192.168.1.101 (this is the entry for the ikondev domain to point to the machine with 192.168.1.101. You can set up several more entries like this of course)
• interface=eth0 (eth0 has to be changed to your network interface which connects to the development network)
• no-dhcp-interface=eth0 (eth0 has to be the same as the one for the previous setting as we do not want to set up a dhcp server)

Final touches

Now all we have to do is restart the network and the dnsmasq services and set all the dev machines primary nameserver to be VM2's ip address as it will be serving our local and mainstream resolves.

Voila! all subdomains typed for ikondev will be resolved as VM1. :)

Access a remote service localy via ssh

This will be just a short reminder for myself not to forget how to forward a remote service's port to a local one through ssh. Here is the schema:

ssh -i .ssh/ssh_keys/root -L 55432:localhost:5432 root@thinks.iamallama.com

With -i we use a private key to the remote host's user named root at the end. Next we say we want to access a local port -55432- and send the data through that one to the remote port -5432-. We want to use our localhost to address the service which is hosted on the thinks.iamallama.com machine.

This wraps it up. ^_^

How to cut FLAC with CUE files

I have read the best and simplest procedure for Windows on this forum thread: http://www.head-fi.org/t/547712/cut-a-flac-ape-file-with-the-cue

Here is the recipe:

1. Get Flac, you just need flac.exe, put it in some directory;
2. Open the .cue file in Foobar;
3. Select all the tracks, right-click, choose 'convert', choose '...';
4. Click 'output format', choose FLAC, click 'edit' and set it to 8;
5. Click 'destination' and choose your options;
6. Click 'processing' and make sure 'enable decode...' is off, and 'active dsps' is empty;
7. Click convert, foobar will ask you were you put flac.exe, navigate to it and voila.

You can use the "bit compare" plugin to make sure the new converted tracks are bit identical to the original ones.

Directory tree in Bash

I was recently searching for a bash script that could be used to create the skeleton of a yaml file containing directory structure. The best material I found is here: http://systembash.com/content/one-line-linux-command-to-print-out-directory-tree-listing/

With the modification of this line:

find ./ -type d | sed -e 's/[^-][^\/]*\//--/g;s/--/ |-/'

I managed to produce the best skeleton:

find ./ | sed -e 's/[^\/]*\// /g'

Since I still have to shape this file up with more data I am not bothering with finding a method to put the ':' at the end of each directory as I can do it myself with little effort. I use Vim's visual mode to get rid of the extra leading spaces at each line.

Limiting Access with SFTP Jails on Ubuntu

This post is based on this article with it's comments.

In the /etc/ssh/sshd_config file edit the Subsystem like this:

Subsystem sftp internal-sftp

Add these to the end of the file:

Match group filetransfer
  ChrootDirectory %h
  X11Forwarding no
  AllowTcpForwarding no
  ForceCommand internal-sftp

Finally restart the OpenSSH server:

service ssh restart

Create the group for sftp access:

addgroup filetransfer

If you add this group to existing users, then they will be prevented to login via remote shell, though still be available through su.

usermod -G filetransfer username
chown root:root /home/username
chmod 755 /home/username

Now users cannot create anything in their jail, but it can be tuned with the last command. Though I personally use this technique to jail those users that are able to manage their own websites' content through sftp. I add those web directories through mount's bind option as the normal symlinks won't work. You can either add bindings via the /etc/fstab like this:

/existing/dir /dir/pointing/to/existing/one auto bind 0 0

or via the command line:

mount -o bind /existing/dir /dir/pointing/to/existing/one

Convert virtual machine disks VDI to VMDK

I recently wished to migrate my dev vm from VirtualBox To VMware. So the only thing I needed is to convert the VDI virtual disk into VMware's VMDK format as VirtualBox can read several disk types but VMware needs it's own child.
So as long as VirtualBox is still installed, I could use this simple command:

vboxmanage clonehd old_virtualbox_disk.vdi new_vmware_disk.vmdk –format VMDK –variant standard

Obviously, you have to find the vboxmanage executable yourself on your system. This command is crossplatform. :)

Sync MAC address of virtual machine

When I use VirtualBox or VMware I sometimes have to change the network adapters of the virtual quests and there comes the problem in case of my Ubuntu VMs that their ethX adapter's expected MAC address differs from the one they get from the host. Here is that single config file that you have to sync to contain the new address:

/etc/udev/rules.d/70-persistent-net.rules

The setting containing MAC address in the interface's line is:

ATTR{address}==""

 After reboot everything is ok :)

Faster transactions in PostgreSQL

First of all: BE WARNED! 

These settings are not recommended for all of you (if any) in production as they might cause you data corruption or loss in case of sudden hardware failure.
I would on suggest to have these settings used in a development environment, where you have to run many tests for your application which include database side create/drop transactions.

My case here was that I ran so much database inclusive tests, that I got fed up with PostgreSQL's ~6s database creation time. So I just Googled it a few days how this could be done in a dev env. So here it is, you just set these two variables in your postgresql.conf file and reastart the database.

fsync = off
full_page_writes = off