Alberto Embedded & Open experience

Notes on my experience on Open Source Embedded Systems

How to dock OpenOffice Navigator and Style windows when Graphics effects are enabled in Ubuntu

leave a comment »

Ubuntu is a great distribution, but not all is perfect (as all in the world)

No docking feedback

Moving on the left the Navigator window, the docking grey lines are not showed anymore

When the composite manager is enabled (System -> Preferences -> Appearance -> Visual Effects set to Normal or Extra), moving the OpenOffice’s Navigator and Styles windows on the right or left do shows docking capabilities for these windows anymore.

Well there’s a simple fix about this which I’ve found here: Simple Ctrl + Double left click on the grey area of Navigator (or Styles) and the window will be anchored on the left.

Then you will be able to drag the docked component on the side you prefer.



Written by Alberto!

07/02/2011 at 11:20 am

Tweet from the linux shell on Ubuntu Karmic

with 4 comments

Twitter is something like a panel where you write all what you want in as short message style.

What makes Tweeter powerful (for me and for  IT developer) is the fact that nowadays IT companies look for job candidates also in these Social Networks. This is a way to find out the best candidates for a vacancy looking for already skilled people in what the vacancy involve.

Well, I am a kernel developer, so what is better than the ability to tweet my messages from bash???

This tutorial is for whom that want to do this also after the Tweeter strength of its authentication policy (now plain authentication is forbidden) with the OAuth mechanism.

The utility that can do this is called bti version 028 and upper, made by one of the most important member of the Linux development community: Greg Kroah Hartman.

Getting the source

So move into your preferred source directory, I will call it $BUILD and clone these repositories (it is needed a working installation of git and svn ):

$ cd $BUILD
$ git clone git://
$ svn co

all other packages needed can be found with apt-get (or whatever other debian package management you like).

Compiling the source

First we have to compile liboauth (it is needed a working installation of automake and autoconf):

$ cd $BUILD/liboauth
$ mkdir m4
$ aclocal
$ automake --add-missing
$ autoreconf
$ ./configure
$ make
$ sudo make install

The library will be installed to /usr/local/lib, if you want another destination dir, change the –prefix as you want in configure time.

Now bti:

$ cd $BUILD/bti
$ ./
$ # Because of liboauth did not wrote the package configuration info:
$ ./configure LIBOAUTH_LIBS=/usr/local/lib/liboauth.a LIBOAUTH_CFLAGS="-g -O2"
$ make
$ sudo make install

In the configuration process you may be prompted to install other packages. You can find them with apt-get.

Make it work

Now you have the command bti available on your shell and you have to complete the OAuth authorization process to be able to post on your twitter account.

First create your configuration file as:

# comments are allowed in the bti config file
# rename this to ~/.bti so that you do not need
# to constantly enter your account name and/or
# password on the command line every time you send
# a message.
account="your account"
password="your password"
# Example of a custom laconica installation
#user="your proxy user"
#proxy="http://your proxy address:port"

and put it in ~/.bti.

Consumer key and secret

Now the “cloudy” part for me. Because consumer key and secret should be application specific as I understood so these keys  should be part of the bti application (correct me otherwise).

@update from Amir: Read this post to understand why those informations are not part of the bti sources,-OAuth-and-bti.html.

Anyway you can register a new application to have access at your twitter following this link: Please give the correct info on bti website as linked from this post but choose a different application name because bti is already choosen.

The application type is client without callback, must to have write permission and do not use twitter for the authentication.

Twitter will give you the two keys you need and You will put these at the and of your .bti config file as:


Then try your first run:

$ bti

You will be prompted to browse a specific link and past back the pin you see. Then you will be asked to add other two rows in your configuration file and  when you have done it, well all the process is completed! with the command:

$ echo "tweet what you want" | bti

You will be able to tweet all you want from the shell!


Maybe for someone lazy it is too much complicated the form with echo.

So I suggest to add the following lines at the end of your ~/.bashrc:

# $ tweet "your message"
function tweet {
 if [ ! "$*" ]; then
   echo 'Nothing to tweet!'
   return 1
 echo "$*" | bti

The comment on top explain how to use this function from the bash shell.

Hope you enjoyed this tutorial and please, any corrections or suggestions are welcome!


Written by Alberto!

04/09/2010 at 10:26 am

Posted in linux

Tagged with , , ,

OpenOCD with Amontec JtagKey2 support

leave a comment »

Recently I bought an Amontec JtagKey-2 debug probe for my embedded projects. This Jtag probe have great features over a relative little buying cost: over other low cost Jtag probes, this one supports the RTCK signal and nominal speeds are really interesting.

What this post will describe is how to setup a useful JTAG debug environment:


The JtagKey2 is based on FTDI FT2232 usb interface. Drivers for this chip can be found here:

Choose your arch (for me Linux x86_64) and  download the archive in  $PRJROOT/openocd.

$ cd $PRJROOT/openocd
$ tar -xvf libftd2xx0.4.16_x86_64.tar.gz


OpenOCD can be downloaded from its public git repository:

$ git clone git:// openocd

Then, move to the opnocd directory and configure with:

$ cd openocd
$ ./bootstrap
$ # If you need, choose the right prefix with --prefix=DIR
$ ./configure --enable-maintainer-mode --enable-ft2232_ftd2xx \
$ # n: Numbers of cores in your machine
$ make -j n

At this point the OpenOCD build process fire a linker error, this because  libtool mess up the library order when linking main.c so, let we do the right command:

$ cd src/

$ gcc -std=gnu99 -g -O2 -I/media/dati/android/openocd/libftd2xx0.4.16_x86_64 \
  -Wall -Wstrict-prototypes -Wformat-security -Wextra -Wno-unused-parameter \
  -Wbad-function-cast -Wcast-align -Wredundant-decls -Werror \
  -o openocd main.o ./.libs/libopenocd.a \
  /media/dati/android/openocd/libftd2xx0.4.16_x86_64/static_lib/libftd2xx.a.0.4.16 \
  -ldl -lpthread
$ cd ..
$ make -j 6

And at the end, install:

$ sudo make install

Using OpenOCD

OpenOCD is completely configurable: the server must be started with a number of “-f ” options that points to the necessaries configuration files.
Normally must be given two configuration files: The interface config and a board config.
So OpenOCD can be started with:

$ sudo openocd -f interface/jtagkey2.cfg -f board/your_board.cfg

We have to use sudo because of permissions needs on using an usb device, otherwise an udev rule must be set.

Written by Alberto!

16/03/2010 at 12:11 pm

Posted in linux-arm

Tagged with ,

On Android smoothness

with 2 comments

Currently I am working on Android optimizations for Freescale i.MX SoCs.

I am frustrate because of a lot of CPU utilization in jobs where the load should not be as heavy as it looks.. But I am working on! And solutions are coming up.
I found by chance this  Freescale video, where one Freescale engineer present the opportunity to develop Android based embedded projects on an MPC8536 (Power III based SoC that works at 1.25 GHz) provided board.
He say “Optimized implementation of  Android operating system” but look at time 2:18! How smooth is this Optimized implementation 🙂


Written by Alberto!

25/02/2010 at 12:53 am

Posted in android

Tagged with , , , ,


leave a comment »

I'm going to FOSDEM, the Free and Open Source Software Developers' European Meeting

Too late, I know, but I’ve been there!! 🙂

It was the fist time and sure, I think it is normal, but I was too excited!! Lot of Open Source stuff: areas, projects, peoples, minds.. All was Open!! and lot of Embedded talks and stands. I was crying like a baby :’)
I had the chance to introduce me to two of mine bigger open source guru, and it was Great! they are people like me and you, with a lot of desire to do their work well!

If you are thinking on going or not the next year, CAME WITH ME!!!!!!


Written by Alberto!

09/02/2010 at 11:30 am

Posted in Meeting Events

Tagged with ,

Git clone with local references

leave a comment »

In this post I explain how to clone a remote git repository basing it on a local one.

If you work with the Linux kernel  there are so many different repositories for developing, so when you are asked to base your patch series on branch x of repo y, if you follow those rules you don’t have the need to re download the n-est Linux kernel repository!

First: You must have a clean linus repository!

Saying the project root dir for developing is $PRJROOT we can setup a clean linus repo with:

$ mkdir kernel-linus
$ cd kernel-linus
$ git clone git://

Second: clone a remote repository referencing the local linus one:

The parameter –reference in the git clone command make what we need

$ mkdir kernel-mfd
$ cd kernel-mfd
$ git clone git:// \
  --reference ../kernel-linus/linux-2.6/

Pay attention that the default behaviour for git in local referencing is to setup a .git/objects/info/alternates to share the objects with the source repository, resulting in much more space occupied but deleting old objects  in the new one, deletes the same in the old one that is not always a good thing.

So only go ahead in referenced local repositories 🙂



Written by Alberto!

11/12/2009 at 6:03 pm

Posted in Git

Tagged with , ,

NAND Flash Support

leave a comment »

In the CPU board there is a 2 Gb NAND Flash chip organized in :

1 Device = (2K+64)Bytes x 64Pages x 2048 Blocks = 2112Mbits

This chip is connected to the i.MX31 Application Processor through the Nand Flash Controller Interface with an 8 bit data bus. The in kernel driver “drivers/mtd/nand/mxc_nand.c” can easily support it registering the nand device in the machine code with the follow platform data:

static struct mxc_nand_platform_data imx31pdk_nand_flash_pdata = {
       .width          = 1, /* 1 Byte data bus */
       .hw_ecc         = 1, /* No suppress hardware ECC */

The mxc_nand.c driver is enabled by default with a $ make mx3_defconfig

What can be enabled is the MTD Partitioning support with support for RedBoot partition table parsing where the Location of RedBoot partition table is the number of the Flash page where it is stored the RedBoot “FIS directory” that for me start at 0x80000, so it is the fourth page (4).

The resulting patch:

Subject: [PATCH 2/2] MXC: imx31pdk: Add support for on board NAND Flash.

Signed-off-by: Alberto Panizzo <>
arch/arm/mach-mx3/mx31pdk.c |   10 ++++++++++
1 files changed, 10 insertions(+), 0 deletions(-)

diff --git a/arch/arm/mach-mx3/mx31pdk.c b/arch/arm/mach-mx3/mx31pdk.c
index 0f7a2f0..ca1fe5d 100644
--- a/arch/arm/mach-mx3/mx31pdk.c
+++ b/arch/arm/mach-mx3/mx31pdk.c
@@ -34,6 +34,7 @@
#include <mach/board-mx31pdk.h>
#include <mach/imx-uart.h>
#include <mach/iomux-mx3.h>
+#include <mach/mxc_nand.h>
#include "devices.h"

@@ -53,6 +54,14 @@ static int mx31pdk_pins[] = {

+ * NAND Flash
+ */
+static struct mxc_nand_platform_data imx31pdk_nand_flash_pdata = {
+    .width        = 1,
+    .hw_ecc        = 1,
static struct imxuart_platform_data uart_pdata = {
@@ -241,6 +250,7 @@ static void __init mxc_board_init(void)

mxc_register_device(&mxc_uart_device0, &uart_pdata);
+    mxc_register_device(&mxc_nand_device, &imx31pdk_nand_flash_pdata);

if (!mx31pdk_init_expio())

Written by Alberto!

04/12/2009 at 4:45 pm

Posted in imx31pdk

Tagged with , ,