if it's possible, i'd like to clarify and document the current state
of both toolchains and kernel source trees, completely independent of
their respective distros. more specifically, i'd like to clarify the
"best* possible combination of toolchain and source tree which can
build the "best" bootable kernel on a BB, totally irrespective of the
root filesystem to come after it.
(actually, i'm interested in the "best" possible kernel leaving out
the root filesystem entirely -- just let the kernel boot and fall over
when it tries to mount the rootfs and doesn't find one. who cares?
we just want to work the bugs out of the kernel. does that make
first, toolchains. IIRC, no one was terribly impressed with the
codesourcery products. has that changed at all recently? i thought
they'd had some new releases recently but i'll defer to the wisdom of
then there's the pre-built (android) toolchain downloadable from
embinux here: http://www.embinux.com/download_toolchain.php. that
would *seem* to work properly. then there's the toolchain you can
build from OE, and yann morin tells me that he created one using
crosstool-ng that would appear to work. so ... in a nutshell, what's
the best choice? in any event, it might be useful to put up a wiki
page explaining how many different choices there are for toolchains
for BB work and where you can get them, and i can do that.
then there are kernel source trees, with varying levels of
completeness. at a bare minimum, we can start with linus' git tree,
which will allow you to:
$ make ARCH=arm CROSS_COMPILE=<whatever> omap3_beagle_defconfig
$ make uImage
sure, you'll get a beagle-bootable kernel, but one that's lagging
behind the other trees in terms of features.
there's the (certainly more up-to-date) omap 2.6 git tree here:
and the android tree you can find here:
and i'm guessing angstrom comes with one as well, etc, etc.
and, in the end, in a perfect world, one would think that everyone
would eventually be able to converge to a common toolchain and kernel
source tree (the source tree being linus' tree once all the OMAP stuff
is pushed upstream), then diverge drastically in terms of kernel
boot-time parameters and root filesystems.
does all of this make sense? i'm going to test some of this by
building various kernels and test-booting them on my C2 rev BB, just
to see how stable they are.