easing the pain of OpenEmbedded

Hi,

I'm in a process of setting up the OE environment for beagleboard.
I'm using now the stable 2009 branch as advised elsewhere in
discussion.
So far I am not able to actually finished the process due to a number
of errors.
My question is:
Would it be possible to put somewhere a .tar of a complete initial oe
environment for download?
It would be easier after that to customize it.

Jan

hi.

the process of setting up OE environment describe very well on wiki,
it's very good wiki, that maybe help you a lot.

Hi,

You didn't understand me.
The pocess is very well documented indeed and I don't have a problem
with it at all.
What I,m saying is that it requires a very fast computer and a very
fast internet and as such
it is limited to people who have that type of resources.
I have a 2 year old PC (3GHz and 1MB of RAM) and a high speed (cable)
internet.
With this it takes hours and days to bitbake a console image.
This is a warning for newbies to embedded linux like me.
You need to have a good PC and a bandwith if you're serious do do any
development.
(Similary the myth of Linux running well on an old PC should be
busted)
I have already asked for a better PC at work, but it will take time.
There may be more people like me.
Looks like I have to stay at work tonight, monitor the process and
once I have an error, restart it.
Currently most error are due to wrong MD5 checksum. I delete the
reported files, so they can be reloaded and restart.
Thanks for reading and if you have any helpfull ideas, please share.

Regards

Jan

Wow, 1 Mb of RAM? I remember 286 PCs used to come with such an amount
of RAM :slight_smile: 1Gb is more likely, right?

By the way guys, Fedora and SuSe are providing cross-compiled, ARMv5
instruction set targeted, packages. Would it be possible to use the
omap kernel with a root filesystem image built from those packages,
rather than the recommended OE or Angstrom?

-Ilyes

Hi,

I'm in a process of setting up the OE environment for beagleboard.
I'm using now the stable 2009 branch as advised elsewhere in
discussion.
So far I am not able to actually finished the process due to a number
of errors.

Which errors? We, OE developers, have a problems with our magic glass
balls :frowning:

My question is:
Would it be possible to put somewhere a .tar of a complete initial oe
environment for download?

You only need stable/2009 metadata and 2 environment variables to set.
No need to tar it.

Regards,

What I,m saying is that it requires a very fast computer and a very
fast internet and as such
it is limited to people who have that type of resources.
I have a 2 year old PC (3GHz and 1MB of RAM) and a high speed (cable)
internet.
With this it takes hours and days to bitbake a console image.

That's strange: my main PC is even older, at 2GhZ and 1GB RAM, and
only takes a few hours to bitbake console image and about a day for a
full graphical environment; maybe you have some bottleneck somewhere,
e.g. with the disk

I've even used a low power PC with 256MB RAM and an external USB2 HD,
and even that only took about a day to bake console_image

that's behind a common 2Mbps ADSL internet connection

This is a warning for newbies to embedded linux like me.
You need to have a good PC and a bandwith if you're serious do do any
development.

You do need the bandwidth, but I don't think that any commonly
available PC will reduce building times to (tens of) minutes, it will
always (for a short value of always, of course :slight_smile: ) take hours

(Similary the myth of Linux running well on an old PC should be
busted)

(I don't agree, but the explanation why would be OT)

I have already asked for a better PC at work, but it will take time.
There may be more people like me.
Looks like I have to stay at work tonight, monitor the process and
once I have an error, restart it.

bitbake can be interrupted safely: if you ctrl-C once during a task it
will finish that task and then stop, so that it can be resumed later

Currently most error are due to wrong MD5 checksum. I delete the
reported files, so they can be reloaded and restart.

luckily after the first build most source packages will be available
locally (except for updates), and you won't have to go through this
anymore

Thanks for reading and if you have any helpfull ideas, please share.

Since it seems that your main problem is with the unreliable internet
connection, the first thing that comes to my mind would be to try to
get a full list of the packages you are going to need and "manually"
download them (with bitbake -c fetch recipename), so that the long
compilations can go on without intervention

I'm not sure if there is a way to automatize it (i.e. tell bitbake to
do the do_fetch task for all dependencies of a single recipe)

Sorry, my misprint. It is 1GB.
As Cortex-A8 is ARMv7, my guess is that running ARMv5 instructions
will reduce performance significantly.
For my project it is no good then.

Jan

I have build OE recently on a smaller system (I guess it was 2.4 Ghz
and 512M or maybe 1GB of RAM).
It takes a while (I think a bitbake of virtual/kernel including all
the cross tools etc took most of the day).
Just switch it on in the evening and by the next morning it will be done :slight_smile:

I did not bump into MD5 errors. (and if desired you can specify in
your local.conf to ignore them).
Either your network connection is bad, or you are very unlucky....

BTW: linux runs very well on small machines. I've had it running on
NSLU2 (133 Mhz ARM4 with 32 MB of RAM) without problems.
Reason why things take so long is because there is a lot of stuff to
be build (and maybe because gcc is not blindingly fast and definitely
a memory hog).

Frans

I don't think that ignoring MD5 errors is a good idea, expecially if
they are not caused by a missing or outdated entry, but by actual
download corruption, as that would probably lead to errors in the
unpack stage

Hi,

You didn't understand me.
The pocess is very well documented indeed and I don't have a problem
with it at all.
What I,m saying is that it requires a very fast computer and a very
fast internet and as such
it is limited to people who have that type of resources.
I have a 2 year old PC (3GHz and 1MB of RAM) and a high speed (cable)
internet.
With this it takes hours and days to bitbake a console image.
This is a warning for newbies to embedded linux like me.
You need to have a good PC and a bandwith if you're serious do do any
development.
(Similary the myth of Linux running well on an old PC should be
busted)
I have already asked for a better PC at work, but it will take time.
There may be more people like me.
Looks like I have to stay at work tonight, monitor the process and
once I have an error, restart it.
Currently most error are due to wrong MD5 checksum. I delete the
reported files, so they can be reloaded and restart.
Thanks for reading and if you have any helpfull ideas, please share.

Have you looked at the online image builder at:

http://amethyst.openembedded.net/~koen/narcissus/

Philip

From: beagleboard@googlegroups.com
[mailto:beagleboard@googlegroups.com] On Behalf Of Philip Balister
Sent: Wednesday, April 29, 2009 10:01 AM
To: beagleboard@googlegroups.com
Subject: [beagleboard] Re: easing the pain of OpenEmbedded

>
> Hi,
>
> You didn't understand me.
> The pocess is very well documented indeed and I don't have a problem
> with it at all.
> What I,m saying is that it requires a very fast computer and a very
> fast internet and as such
> it is limited to people who have that type of resources.
> I have a 2 year old PC (3GHz and 1MB of RAM) and a high speed (cable)
> internet.
> With this it takes hours and days to bitbake a console image.
> This is a warning for newbies to embedded linux like me.
> You need to have a good PC and a bandwith if you're serious do do any
> development.
> (Similary the myth of Linux running well on an old PC should be
> busted)
> I have already asked for a better PC at work, but it will take time.
> There may be more people like me.
> Looks like I have to stay at work tonight, monitor the process and
> once I have an error, restart it.
> Currently most error are due to wrong MD5 checksum. I delete the
> reported files, so they can be reloaded and restart.
> Thanks for reading and if you have any helpfull ideas, please share.

Have you looked at the online image builder at:

Openembedded-Hotels, Villen, Unterk├╝nfte in Leipzig

Philip

Wow!, nice tool I see a lot of potential.

Philip thanks for the link.

Leandro

Hi Philip,

Yes, I have used it and it works nice.
However I need sources (headers) to write my own application.
Thanks everyone for input.

Jan

Hi Philip,

Yes, I have used it and it works nice.
However I need sources (headers) to write my own application.
Thanks everyone for input.

Sounds like we should get narcissus to emit toolchains compatible with
the images :slight_smile:

Philip

Select the toolchain checkbox :slight_smile:

Op 30 apr 2009 om 03:33 heeft Philip Balister
<philip.balister@gmail.com> het volgende geschreven:\

As a conclusion
I was reading the other thread (CCS + Beagleboard problem)
http://groups.google.com/group/beagleboard/browse_thread/thread/4f8fae177ee57304/72aa1c1c1921b780#72aa1c1c1921b780
and my dissapointment grows even futher.
Looks like there is no easy way for software development on
beagleboard.
I have to fight with openembedded even that I don't like it, but there
is no working alternatives.

Jan

As a conclusion
I was reading the other thread (CCS + Beagleboard problem)
http://groups.google.com/group/beagleboard/browse_thread/thread/4f8fae177ee57304/72aa1c1c1921b780#72aa1c1c1921b780
and my dissapointment grows even futher.
Looks like there is no easy way for software development on
beagleboard.
I have to fight with openembedded even that I don't like it, but there
is no working alternatives.

You can set your path into tmp/cross/.. to pick up the compiler +
binutils and select libpath's to point into staging, There are lots of
ways to solve the problem.

But I agree, there needs to be a straight forward solution.

Philip

Another day, another experience.
Finally I was able to create an image /home/xxxxx/oe/tmp/deploy/glibc/
images/beagleboard/Angstrom-console-image-glibc-ipk-2009.X-
test-20090501-beagleboard.rootfs.tar.bz2 with size of 8.3 MB (8745426
bytes)
I did follow exactly the steps from the web tutorials and guess what?
It doesn't work.
It is a console stable 2009 Angstrom image.
I have a number of SD cards with different images from different
sources working with no problems, but this one
boots sort of OK and then freezes at this point
on serial terminal

Freeing init memory: 196K
EXT3 FS on mmcblk0p2, internal journal
eth0: link up, 100Mbps, full-duplex, lpa 0x45E1
eth0: link up, 100Mbps, full-duplex, lpa 0x45E1

on DVI screen
hwclock: can't open 'dev/misc/rtc': no such file or directory

which after few minutes goes blank.

Can anyone confirm that the image created by tutorial actually works
and it is just me?
Looks like a long weekend ahead.
Thank you for your patience.

Jan