Bitbaking the kernel, Angstrom and qt4-embedded all at once
This is what i should have done to begin with! The instructions are very similar to previous ones except this time we clone the right git repository!
On a fresh copy of Fedora 11 I updated the system and installed the packages reccommended by OE:
python-psyco perl texinfo texi2html diffstat openjade docbook-style-dsssl \
docbook-style-xsl docbook-dtds docbook-utils sed bison bc glibc-devel glibc-static \
gcc binutils pcre pcre-devel git quilt groff linuxdoc-tools patch linuxdoc-tools \
gcc-c++ help2man perl-ExtUtils-MakeMaker”
I then made a folder called OE in my home directory and git cloned the mini2440 repo:
mkdir OE
cd OE
git clone git://repo.or.cz/openembedded/mini2440.git openembedded
I then setup the source-me.txt similar to before:
I put the following in:
BBPATH=${OETREE}/:${OETREE}/openembedded/
echo Setting up dev env for Ångström
if [ -z ${ORG_PATH} ] ; then
ORG_PATH=${PATH}
export ORG_PATH
fi
if [ -z ${ORG_LD_LIBRARY_PATH} ] ; then
ORG_LD_LIBRARY_PATH=${LD_LIBRARY_PATH}
export ORG_LD_LIBRARY_PATH
fi
PATH=${OETREE}/openembedded/bitbake/bin:${ORG_PATH}
LD_LIBRARY_PATH=
export PATH LD_LIBRARY_PATH BBPATH
export LANG=C
export BB_ENV_EXTRAWHITE="MACHINE DISTRO OETREE ANGSTROM_MODE ANGSTROMLIBC LIBC"
su -c 'sysctl vm.mmap_min_addr=0'
echo "Altered environment for OE Development"
Now all you need to do is setup the local.conf. Simply edit the example and copy it to the right place:
gedit mini2440_local_conf_example.conf
Read this file and edit as appropriate. I have attached mine as an example. Now copy it to the conf/local.conf
If you want qt4-embedded or any other package included in the build simply edit the recipe in:
I added qt4-embedded:
IMAGE_PREPROCESS_COMMAND = "create_etc_timestamp"
ANGSTROM_EXTRA_INSTALL ?= ""
DEPENDS = "task-base-extended \
psplash-zap \
esekeyd u-boot-utils tslib \
i2c-tools i2c screen rsync nfs-utils \
directfb gdbserver directfb mtd-utils \
"
IMAGE_INSTALL = "task-base-extended \
${ANGSTROM_EXTRA_INSTALL} \
psplash-zap qt4-embedded\
esekeyd u-boot-utils tslib-calibrate tslib-tests \
i2c-tools i2c screen rsync nfs-utils-client \
directfb gdbserver directfb mtd-utils \
rsvg pango \
"
export IMAGE_BASENAME = "mini2440-image"
IMAGE_LINGUAS = ""
inherit image
Then all you have to do is source the script from wherever you put it:
and run bitbake:
from the ~/OE/openembedded/ directory..
If you get an error trying to build a package try cleaning it:
Then try and build the image again. Maybe you won’t have this problem. I guess it’s either because the checksum parser failed to build or my hdd is dodgy. It only happened once on binutils but worked fine after cleaning it.
A few hours later if all goes well you should have images in:
Obviously that directory depends on what you setup in the local.conf.
Good luck!
Hmm…ICBM? I need to think more about this, i’ve seen quite a few papers on using extended kalman filters to fuse sensor data from gyros, accelerometers and gps but i’ve only scratched the surface so far. It will be interesting to hear how you get on with all those sensors!
Well, cooking with gas now!
NOTE: Running task 718 of 2666 (ID: 1597, /home/charles/oe/openembedded/recipes/u-boot/u-boot-mkimage-openmoko-native_oe.bb, do_fetch)
git: error while loading shared libraries: libcrypto.so.0.9.7: cannot enable executable stack as shared object requires: Permission denied
Ends here. No output. u-boot must be the first thing built?
Have you got SELinux set to permissive?
If not, try a
su -c 'setenforce 0'to temporarily set to permissive until restart.Everything went great on the Linux box at work. Images of all sorts and lots to test. Very cool!
However, at home on the laptop with the same Fedora, I error out on the last task or second to last.
Sorry for the flood. I have tried reloading all the packages and a git pull and a cleanup and fsck and gc. Still get this deprecated message and error.
charles@fujitsu oe]$ bitbake mini2440-image
NOTE: Handling BitBake files: \ (6712/6712) [100 %]
NOTE: Parsing finished. 6410 cached, 0 parsed, 302 skipped, 0 masked.
NOTE: Cache is clean, not saving.
NOTE: Resolving any missing task queue dependencies
NOTE: Preparing runqueue
NOTE: Executing runqueue
NOTE: Running task 2665 of 2666 (ID: 10, /home/charles/oe/openembedded/recipes/images/mini2440-image.bb, do_rootfs)
ERROR: function do_rootfs failed
ERROR: log data follows (/home/charles/oe/oetmp/work/mini2440-angstrom-linux-gnueabi/mini2440-image-1.0-r0/temp/log.do_rootfs.30726)
| + rm -rf /home/charles/oe/oetmp/rootfs
| + mkdir -p /home/charles/oe/oetmp/rootfs
| + mkdir -p /home/charles/oe/oetmp/deploy/glibc/images/mini2440
| + ‘[' 0 '!=' 1 ']‘
| + for devtable in /home/charles/oe/openembedded/files/device_table-minimal.txt /home/charles/oe/openembedded/files/device_table_add-s3c_serial.txt
| + makedevs -r /home/charles/oe/oetmp/rootfs -D /home/charles/oe/openembedded/files/device_table-minimal.txt
| + for devtable in /home/charles/oe/openembedded/files/device_table-minimal.txt /home/charles/oe/openembedded/files/device_table_add-s3c_serial.txt
| + makedevs -r /home/charles/oe/oetmp/rootfs -D /home/charles/oe/openembedded/files/device_table_add-s3c_serial.txt
| + rootfs_ipk_do_rootfs
| + set -x
| + package_update_index_ipk
| + set -x
| + ipkgarchs=’all any noarch arm armv4 armv4t mini2440′
| + ‘[' '!' -z '' ']‘
| + touch /home/charles/oe/oetmp/deploy/glibc/ipk/Packages
| + ipkg-make-index -r /home/charles/oe/oetmp/deploy/glibc/ipk/Packages -p /home/charles/oe/oetmp/deploy/glibc/ipk/Packages -l /home/charles/oe/oetmp/deploy/glibc/ipk/Packages.filelist -m /home/charles/oe/oetmp/deploy/glibc/ipk
| /home/charles/oe/oetmp/staging/i686-linux/usr/bin/ipkg.py:39: DeprecationWarning: the md5 module is deprecated; use hashlib instead
| import md5
| Traceback (most recent call last):
| File “/home/charles/oe/oetmp/staging/i686-linux/usr/bin/ipkg-make-index”, line 7, in
| import ipkg
| File “/home/charles/oe/oetmp/staging/i686-linux/usr/bin/ipkg.py”, line 39, in
| import md5
| File “/home/charles/oe/oetmp/staging/i686-linux//usr/lib/python2.6/md5.py”, line 10, in
| from hashlib import md5
| File “/home/charles/oe/oetmp/staging/i686-linux//usr/lib/python2.6/hashlib.py”, line 136, in
| md5 = __get_builtin_constructor(‘md5′)
| File “/home/charles/oe/oetmp/staging/i686-linux//usr/lib/python2.6/hashlib.py”, line 63, in __get_builtin_constructor
| import _md5
| ImportError: No module named _md5
NOTE: Task failed: /home/charles/oe/oetmp/work/mini2440-angstrom-linux-gnueabi/mini2440-image-1.0-r0/temp/log.do_rootfs.30726
ERROR: TaskFailed event exception, aborting
The warhead problem is not so hard in itself. After all when it is “exoatmospheric” (I grind my teeth when they say that) You know it is on an ellipse centered near the center of the Earth. The Kalman is all about the sensor positioning errors – it is an IR telescope in a 747. (A girl who graduated with me in physics ran the instrumentation on the Keiper flying observatory out of NASA AMES for many years, the Kuiper flying observatory. (Its new replacement has about a 90 inch scope and flys with a door open for observation instead of an IR window.) The scope pointing software was similar. Plus there was the “endoatmospheric” stage where the IR signature can be used to tell a warhead from a decoy from the booster which is coming in right behind.
The Inav of the missile on the other hand must have a detailed gravity map of the launch region. If you use inertial nav over much range, as in an aircraft, local crust density and mountains and hills will throw you off. Granite floats on basalt so if you travel from basalt to a granite dome, g changes enough to affect your nav system.
The coolest part was that while various IV League faculty “proved” in Scientific American that you couldn’t use space based LASERs to hit atmospheric missiles, the researchers developed the phase conjugate mirror, which made it impossible to miss.
By the way, this is why submarines had to launch from specific sites until GPS was finished.
Hmm.. i’d guess you have a problem with python module md5.
I’d try running the yum command again at the top of the page. It’s strange it’s worked for you on the other box and not the laptop.
If that doesn’t help i’d try importing md5 in python. Type ‘python’. Then type ‘import md5′ in the python shell. If it doesn’t import you’ve got a clue!
Normally when i get an error it can be resolved by cleaning the package and starting again. Here i’d try:
bitbake -c clean mini2440-image
And then build it again. I’m not sure why i get these seemingly random errors sometimes, it seems like on occasion things get corrupted, but this normally resolves the problem!
Until now i’d never heard of a phase conjugate mirror! After work i’ll have a look!
Hi Doug
I followed your instructions but I am having the error below. What is the problem?
By the way, I have Ubuntu8.1.
…..
yes
| checking for setlocale… yes
| checking for putc_unlocked… if [ x"" != x ]; then \
| gcc -c -DHAVE_CONFIG_H -isystem/home/sgoren/MINI2440/OE/oetmp/staging/i686-linux/usr/include -O2 -I. -I/home/sgoren/MINI2440/OE/oetmp/work/armv4t-angstrom-linux-gnueabi/gcc-cross-initial-4.2.4-r5/gcc-4.2.4/libiberty/../include -W -Wall -pedantic -Wwrite-strings -Wstrict-prototypes -Wc++-compat /home/sgoren/MINI2440/OE/oetmp/work/armv4t-angstrom-linux-gnueabi/gcc-cross-initial-4.2.4-r5/gcc-4.2.4/libiberty/pex-one.c -o pic/pex-one.o; \
| else true; fi
| cc1: error: unrecognized command line option “-Wno-overlength-strings”
| make[2]: *** [fixincl.o] Error 1
I got the mini2440 from FriendlyARM reseller to write a program for it. Since I wanted to develop in X11/GTK, after little research I found that I needed GPE. Further research revealed that there is an image available with the instructions at: http://narnia.cs.ttu.edu/drupal/node/131.
However, trying to drop the gpe image on the device won’t work. I followed an instructions, but when I reset the device, it didn’t reboot and went to the error message on my HyperTerminal. Then I threw the original kernel image that came with the device. It didn’t boot again, but now it didn’t boot at all.
Here is some info:
MINI2440 # nand info
Device 0: NAND 256MiB 3,3V 8-bit, page size 2048, sector size 128 KiB
I would like to have a working X11/GPE mini2440 device to start debugging my program.
Is there a working solution somewhere either locally or on the Web?
Thank you.
@sgoren
I haven’t experienced this problem but Isn’t this a bug in the older versions of gcc when using 64bit processors? Have you got the latest gcc? Maybe it is time to upgrade your distro
@Igor
You can build a GPE image based on angstrom using Open Embedded. Try using the command:
bitbake gpe-image.You will also want to follow these articles:
http://www.electronics.diycinema.co.uk/booting-using-nfs-and-tftp/
http://www.electronics.diycinema.co.uk/compiling-the-mini2440-kernel/
http://code.google.com/p/mini2440/wiki/MiniBringup
Which will build you a working boot loader (uboot), and explain how to flash it onto the mini2440.
Hi, Will,
I understand.
From my experience working with Linux and from what I see is happening here with mini2440, it’s either GPE on that instruction was provided for 64M “hard drive” and when I try to load it to 256M hard drive there is an unknown/unidentified space for the kernel and it fails to boot.
I got this from the situation where I threw the original bootloader, original kernel image and the GPE image from that link.
However, what I don’t understand is that why the boot didn’t even started. All it did was saying:
”
Setup Linux Parameters 0×30000100
linux command line is: “noinitrd root=/dev/mtdblock3 init=/linuxrc console=ttySAC0″
MACH_TYPE=1999
Now, Booting Linux….
Uncompressing Linux………………………………………………. done, booting the kernel
”
and that’s it.
It should at least start booting and throw the kernel panic on the HyperTerminal here.
Again this is done using the original bootloader/kernel and the GPE image from the instructions of Mr Bao.
Am I right in this assumption? Or it’s something else going on here?
Thank you.
Hi, did anyone had problem with 3.5″ screen and touchscreen?
I bitbaked image with tslib and qt-embedded,
but when running examples, axis on touchsreen are inverted!
tslib calibration and tests are working fine
display type is T35 by chinesse images,
for angstrom passing lcd type ‘3′ from loader
How to correct touchcreen?
Doug,
I was able to complete the bitbake of x11-image (as per your recomendation for X11).
It flashed OK on my mini2440 and bootsup fine but the LCD Screen is blank. I am able to login thru serial console.
Do you have any clues . I tried using mini2440=0tb and 1tb. I have a 3.5 Display but when I set to
1tb I see the linux penguin during start of boot and then all blank.
If you haven’t seen this, Can you point me to somewhere I can read about it.
Any help welcome
@suni
You may need to use
mini2440=3tbif you have a newer screen.Will
Thanks for the response. I tried with mini2440=3tb in bootargs(setenv bootargs ….)
I get a white screen. Then I tried mini2440=2tb and I see portion of a flickering Penguin at the
bottom Left corner.
I have a board from Developmentboard.net that I bought in end of Jan.
I cleanedup x11-image and bitbaked qtopia-core-gui-image as well. No success.
Should I try any other recipe. I tried init 3 and then init 5. It doesn’t seem to have any effect atall on the
LCD Screen.
My bootargs is
root=/dev/mtdblock3 rootfstype=jffs2 console=ttySAC0,115200 mini2440=1tb
I build on Ubuntu 9.1
Any clues/ suggestions/pointers please…
Hi Suni, i’m not sure what’s going wrong. To clear up the confusion with the boot args have a look at the source code for the machine conf:
http://repo.or.cz/w/linux-2.6/mini2440.git/blob/HEAD:/arch/arm/mach-s3c2440/mach-mini2440.c
Here you can see exactly what the options you are specifying are doing. As Will points out it looks like you need either 0 or 3 unless they have changed the screen again?
Hi Doug,
I have tried all the options with mini2440. I see that when the kernel starts coming up , a picture of
Linux Penguin is displayed. When I remove console=ttySAC0,115200, I see the boot message on the LCD screen.
Once the boot process is completed, the LCD screen becomes blank with a # at the bottom Left of screen.
init 3 ;init 5 did not make any change. How do I debug this
I was expection to see a GUI with some buttons so that I can write an App to display something.
Is there any other way of compiling to get this. Any other options. I am interested in a Linux system with GUI, that’s all.
Hi Doug,
I have been trying to follow this tutorial on my Ubuntu machine, but each time I get to the: bitbake mini2440-image part, I get the following message:
”
NOTE: Psyco JIT Compiler (http://psyco.sf.net) not available. Install it to increase performance.
ERROR: no files to build.
NOTE: Parsing finished. 0 cached, 0 parsed, 0 skipped, 0 masked.
ERROR: Nothing PROVIDES ‘mini2440-image’ (but ‘[]‘ DEPENDS on or otherwise requires it)
”
I have limited Linux experience, but it looks like an environment problem? I have gone over the tutorial several times but can’t see what I have done wrong. I have changed the source-me.txt and local.conf files to match my directory structure… any ideas/help appreciated.
Hi Matt,
The first warning is probably because you are missing the python psyco package. I don’t use Ubuntu but i think this should work:
apt-get python-psyco
..But before you try that have you read this page?
http://wiki.openembedded.net/index.php/OEandYourDistro
If you haven’t i’d work your way through it so you know you have all the right packages installed.
Secondly the mini2440-image.bb doesn’t exist in all repos. It’s just the base-image plus a few extra packages you may or may not need. Have a look in …OE/openembedded/recipes/images… If you can’t see the recipe in there then you can copy it over from BusError’s repo, but to be honest you might as well try building the base-image first and start worrying about adding extra packages when that works. You can do that by typing:
bitbake base-image
Hope that helps,
Doug
Ok,
Since my last post I have got things moving.. No doubt these are obvious to a Linux pro, but to a novice such as me I have to write down every step.
In Ubuntu I was missing two packages, svn and makeinfo, for these I had to install the following:
sudo apt-get install subversion texinfo
I then had to change the default dash to bash using the following:
sudo dpkg-reconfigure bash
Selecting NO from the menu options, now the bitbake seems to be underway, so I will get the kettle on!
Well, not surprisingly after several hours.. it failed
with the following error message:
make[4]: *** [file-ostream.lo] Error 1
make[4]: Leaving directory `/home/matt/OE/oetmp/work/armv4t-angstrom-linux-gnueabi/gettext-0.17-r2/gettext-0.17/gettext-tools/libgettextpo’
make[3]: *** [all] Error 2
make[3]: Leaving directory `/home/matt/OE/oetmp/work/armv4t-angstrom-linux-gnueabi/gettext-0.17-r2/gettext-0.17/gettext-tools/libgettextpo’
make[2]: *** [all-recursive] Error 1
make[2]: Leaving directory `/home/matt/OE/oetmp/work/armv4t-angstrom-linux-gnueabi/gettext-0.17-r2/gettext-0.17/gettext-tools’
make[1]: *** [all] Error 2
make[1]: Leaving directory `/home/matt/OE/oetmp/work/armv4t-angstrom-linux-gnueabi/gettext-0.17-r2/gettext-0.17/gettext-tools’
make: *** [all-recursive] Error 1
FATAL: oe_runmake failed
I’m going to check the OE link Doug sent earlier, clean up and try again!
Hi Matt. I’ve never come across it failing there before. I find about 50% of problems are fixable by cleaning the package and trying again. For example here i think it would be:
bitbake -c clean gettext
I’d try baking the individual package that failed until the problem is fixed and then go back to the image recipe.
Good luck!
Hi Doug,
Thanks for the tip! I have read through the OE page, and installed the suggested packages. I am just in the process of: bitbake gettext as you suggest. If this works, will I need to celan the mini2440-image prior to starting the whole thing again? Thanks for your help with this, appreciated!
No problem…. you don’t need to clean the mini2440 image, just the broken package. It should then just carry on baking the remaining packages for the image file.
So far I have made it to step 1284 of 2692, and it now looks like I’m going to run out of disk space.. I’m only on a dual boot system which has a 40G partition for Ubuntu. I have removed as many unwanted files and packages as I can to free up more space, but the bitbake just keeps eating it up! How much should the resulting OE folder take up?
Hmm…i would have thought you’d just about be ok with 40GB. I forget exactly how much space mine is taking up, i’ll have a look when i get home. If you need to save space you can add:
INHERIT += “rm_work”
to the local.conf to remove build directories after the package has been built. That should save ~50%. Saying that you’ll most likely want more space in the long run when you start compiling other things. This is where LVM becomes very useful!
Well I got to 7Gb and ran out of free space, Turns out that it is only an 18Gb partition!
I presume if I change the local.conf file I will then need to do a complete clean and revuild in order to take advantage of the INHERIT option?
I think that’s very likely, although i haven’t actually tried using that option yet. If I were you i’d increase the size of the partition!
Increased my partition by another 14Gb, and have re-started the bitbake, so far so good!
Doug, I’m currently looking for a Linux Expert to help with the robotics project I’m working on, I don’t suppose this would interest you?
http://www.hexapodrobot.com/forum/viewtopic.php?f=22&t=442
Hi
I am succesfully able to bikbake with the above procedure. I need to build a modified kernel. (I need to put in
a bootup logo and enable some features in kernel) . Looks like bitbake unpacks a kernel and builds it. How do I make it build my modified kernel sources.
Thanks
Hi
First of all: thanks for the tutorial.
Everything worked fine to me, except for one thing:
Whatever I try, the calibration won’t work in Qt, even tough it works out of Qt, while doing a test.
I tried many ways, many sites, but it still don’t work…
Can somebody help me plz? I’m really getting desperated…
Thanks
for the source source-me.txt to work the line
su -c ’sysctl vm.mmap_min_addr=0′ should be sudo sysctl vm.mmap_min_addr=0
Are you using Ubuntu?
Hello !
I’m new and not have experience with OE. I start work with angstrom and I can build examples ..but when I try to write specific reciepe .BB file I recieve this ERROR :
bitbake -b test_qt.bb
NOTE: Preparing runqueue
NOTE: Executing runqueue
NOTE: Running task 7 of 19 (ID: 10, /home/eukrea/ttk-fgnet-20100630/eukrea-project/recipes/test_qt/test_qt/test_qt.bb, do_compile)
NOTE: package test-qt-r0: task do_compile: Started
ERROR: Task 10 (/home/eukrea/ttk-fgnet-20100630/eukrea-project/recipes/test_qt/test_qt/test_qt.bb, do_compile) failed with 256
ERROR: ‘/home/eukrea/ttk-fgnet-20100630/eukrea-project/recipes/test_qt/test_qt/test_qt.bb’ failed
ERROR: Function do_compile failed
NOTE: Task failed: (‘function do_compile failed’, ‘/home/eukrea/ttk-fgnet-20100630/tmp_angstrom/work/armv6-angstrom-linux-gnueabi/test-qt-r0/temp/log.do_compile.2899′)
ERROR: Logfile of failure stored in: /home/eukrea/ttk-fgnet-20100630/tmp_angstrom/work/armv6-angstrom-linux-gnueabi/test-qt-r0/temp/log.do_compile.2899
Log data follows:
| /home/eukrea/ttk-fgnet-20100630/tmp_angstrom/sysroots/armv6-angstrom-linux-gnueabi/usr/lib/crt1.o: In function `_start’:
| init.c:(.text+0×30): undefined reference to `main’
| collect2: ld returned 1 exit status
NOTE: package test-qt-r0: task do_compile: Failed
ERROR: TaskFailed event exception, aborting
ERROR: Build of /home/eukrea/ttk-fgnet-20100630/eukrea-project/recipes/test_qt/test_qt/test_qt.bb do_compile failed
ERROR: Task 10 (/home/eukrea/ttk-fgnet-20100630/eukrea-project/recipes/test_qt/test_qt/test_qt.bb, do_compile) failed with 256
ERROR: ‘/home/eukrea/ttk-fgnet-20100630/eukrea-project/recipes/test_qt/test_qt/test_qt.bb’ failed
eukrea@eukrea-laptop:~/ttk-fgnet-20100630/eukrea-project/recipes/test_qt/test_qt$
Place help my !
My test_qt.bb file is:
# source : http://docs.openembedded.org/usermanual/html/recipes_examples.html
DESCRIPTION = “My first QT test”
PR = “r0″
SRC_URI = “file://main.cpp \
file://mainwindow.cpp \
file://welcom.cpp \
file://contact.cpp \
file://mainwindow.h \
file://welcom.h \
file://contact.h”
do_compile() {
${CC} ${CFLAGS} ${LDFLAGS}
${WORKDIR}/mainwindow.cpp -o mainwindow
${WORKDIR}/welcom.cpp -o welcom
${WORKDIR}/contact.cpp -o contact
}
#do_install() {
# install -m 0755 -d ${D}${bindir} ${D}${docdir}/test_qt
# install -m 0644 ${S}/test_qt ${D}${bindir}
# # install -m 0644 ${WORKDIR}/README.txt ${D}${docdir}/test_qt
#}
Thanks in advance !
Hi George,
Have a look at this link:
http://docs.openembedded.org/usermanual/html/ch05s08.html
Bitbake the Qt E SDK, and follow the example.
Hope that helps,
Doug
Thanks for the quick response!
I’m new.
I bitbake SDK now and stiil wait ..
I apologize in advance for stupid questions
What is the problem ? I can build examples and console applications but when I try build my custom project I recieve errors . I think my recipe file is wrog or …????
Thanks !
Hello again Doug !
I succeеs bitbake QT SDK . I build examples from http://docs.openembedded.org/usermanual/html/ch05s08.html. But I do not understand why my project created with QT Creator is not allowed to compile a qmake2 and make. I get the following error:
eukrea@eukrea-laptop:~/ttk-fgnet-20100630/eukrea-project/recipes/test_qt/test_qt$ qmake2
eukrea@eukrea-laptop:~/ttk-fgnet-20100630/eukrea-project/recipes/test_qt/test_qt$ make
make: Nothing to be done for `first’.
If You can place help me !
Thanks!