On 2012-05-16 06:41, Thilo Fromm wrote:
Hello Gary,
Khem Raj (3):
eglibc-2.12: Remove
binutils-2.20: Remove
gcc-4.5: Remove
Removing gcc-4.5 would kill Texas Instruments DaVinci (dm814x, dm816x)
support. You only get a heavily patched 2.6.37 kernel from TI for
these (linux-ti81xx-psp_2.6.37). Unfortunately gcc>= 4.6 is unable to
build a kernel which boots for these platforms.
I build with GCC 4.6.3 from that tree for the DM8148 all the time and
it works fine. Maybe there is some option/selection in your
configuration
that's not in mine? Can you send it& I'll give it a compare?
Sure! Our project is on github, and so is our defconfig:
<https://github.com/DFE/HidaV/blob/master/meta-hidav/recipes-kernel-bootloader/linux/files/defconfig>
We added aufs (Another Union File system) and the blockrom MTD FTL to
the kernel, so there are additional options in our configuration.
The only real differences (modulo the hardware/peripherals in use)
are some of the kernel debugging settings. Perhaps they are causing
some problems with the newer compiler. The diffs between our configurations
are attached.
Thanks, but I don't think I will disable most of the debugging
features for now. We have e.g. memory leak reports by the kernel
memchecker in ext4. Given the pitiful state of linux-ti81xx-psp_2.6.37
in general I'm not willing to give up the little debug info I get.
Fair enough, I was just pointing out a possible source of your pain :-)
BTW, do you know if TI ever plans on updating the kernel? 2.6.37 is so
painfully old...
As Martin points out, the newer compilers can have problems with unaligned
access. I've also [sometimes] had better success if I disable
OPTIMIZE_FOR_SIZE
The only other thing I see is the kernel version& recipe used. I am using
the stock one from meta-ti + some platform patches. Perhaps there's
something
else to look at.
meta-ti/recipes-kernel/linux/linux-ti81xx-psp_2.6.37.bb
We're almost using this recipe too. We just appended AUFS integration
and custom package management installer scripts for the kernel. I'll
look into -mno-unaligned-access w/ gcc-4.7 as Gary suggested as soon
as I am finished moving gcc-4.5 into HidaV so we can actually build
firmware again :)
Good luck
--
------------------------------------------------------------
Gary Thomas | Consulting for the
MLB Associates | Embedded world
------------------------------------------------------------
_______________________________________________
Openembedded-devel mailing list
Openembedded-devel@lists.openembedded.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
