Ian Bonham wrote:
> I think I found it. I did a find|grep spitz and looked for defconf
> That gave me a kernel config file, which I then started to mess about in
> with vim. I'm really hoping that the edits stay in place when I kick off a
> build in a bit.

Yes. But you will see conflict during the next mtn up. I did the same in
org.openebedded.dev.

> I must say thanks for the pointer tho! Gave me a clue to where to start
> looking!

You are welcome.

Hint:

If you seek for a feature and you know the project, which provides it
(in this case Linux kernel), a good start point is searching for some
smart string in the sources (and then oe.dev packages/ directory).

In this case you would find .config file in the tmp/work/*/linux-rp-*
build directory. Then you would probably find that recipe in
org.openembedded.dev/packages/linux overwrites it by device specific
defconfig file, so you have to edit oe.dev sources.

If I am totally helpless, I am trying to search for some smart string in
the whole /OE directory, oe.dev sources or tmp/work build directory. It
takes a bit long time, but it seems to be an effective method.

> I like hacking about, it's how I learned about Linux since I first *actually
> brought* SuSE 7.1 and started playing.

Well, playing with openSuSE is my profession.

> I just wish Angstrom/OE was easier to get into. As  I've messed about, I
> sort of wonder what the point of bitbake is? It's bloody awful to configure,
> and took me a day to get sorted. That was hacking about in files, and doing
> nasty things! Avg Joe would walk off in a secoond. Hence we are desperate
> for developers!

Bitbake is a compilation tool, which understands the oe.dev recipe set.
Thank to bitbake, adding a kernel module package means one line of
change in the defconfig file. Bitbake will care the rest - compile the
kernel, package the module as a separate package, put it to the
repository structure.

If you have ideas, how to simplify build setup, then you can join
openembedded-devel list and discuss it there.

> Has anyone had any success running Angstrom under qemu? I wonder if it could
> be easier to install Angstrom as a qemu graphical enviro on a faster
> machine? Then use point&shoot tools to compile in a 'native' envrio which
> looks like the machine that's ur target? Then just lob ova the files u
> actually need to the target machine?

I guess that Andrzej Zaborowski did it. He wrote most of the spitz
emulation code. But I am afraid that it was not properly merged into
upstream qemu-0.9.1.

I also seen several threads on the web trying it. I guess that you would
need full NAND + PROM images. With current kernel, you can read NAND,
but not PROM (nobody yet fixed the PROM partitioning - partition named
Boot surprisingly does not contain the PROM bootloader, but the EN-JP
dictionary database, which occupies the rest of the PROM). The fix would
need to fix few lines in the partitioning table hardwired in the
kernel.

But for example since qemu-0.8.x, emacs lisp compilation successfully
crashes qemu during oe.dev build.


________________________________________________________________________
Stanislav Brabec
http://www.penguin.cz/~utx/zaurus


_______________________________________________
Angstrom-distro-users mailing list
[email protected]
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/angstrom-distro-users

Reply via email to