+++ Wookey [2010-05-09 14:32 +0100]:
> +++ Wookey [2010-05-09 03:36 +0100]:
> > +++ Wookey [2010-05-09 03:06 +0100]:
> > > +++ Wookey [2010-05-04 11:49 +0100]:
> > > r895 gives a backtrace when it dies in the same place:
> > > (However I'm not convinced that was a clean build, so view that
> > > datapoint with some suspicion)
> >
> > Yes. In fact r895 is fine. And now it really _is_ bedtime.
>
> And now I have to catch a train, so have run out of time but it breaks
> between r976 (working) and r994(bust), which in kernel/2.6.29.1 (which
> I have been testing) is only a couple of changes - nominally nothing
> more than a file renaming.
OK. I've got to the bottom of the kernel-itis. Turns out that
CONFIG_AEABI was not getting set, so we were building OABI kernels with
an EABI compiler, which works in itself, but fails to boot an AEABI
rootfs.
I've moved the configs out of the quilt system which makes it easier
to change them and possible to manage the AEABI/OABI_COMPAT options
programatically.
I've only done a basic 'does it boot' test with 2.6.29.1 so far but
expect things to generally work now. I've set full builds in progress
and will do tests on Monday.
Wookey
--
Principal hats: iEndian - Balloonboard - Toby Churchill - Emdebian
http://wookware.org/
(text/plain)