On 10/06/12 15:55, Wookey wrote:
> +++ Neil Williams [2012-06-06 08:04 +0100]:
>> On Wed, 6 Jun 2012 02:35:16 +0100
>> Wookey <wookey@wookware.org> wrote:
>>
>>> So I tried building trunk with defaults except setting 
>>> BR2_VARIANT_CUED=y
>>> BR2_BUILDROOT_NOV_2011=y
>>> BR2_OPENOCD=y
>>> # BR2_ROOTFS_YCT is not set
>>>
>>> i.e we also have: 
>>> BR2_KERNEL_2.6.37.6=y
>>> BR2_YAFFS_2011-02-09=y
>>>
>>> Should I expect this to work? 
>>
>> I haven't done any CUED builds for ages, also haven't been doing stuff
>> in trunk/ for almost as long.
>>
>> However, I just tried a similar CUED build in trunk and
>> configBalloon3-CUED-2010.11 does not exist anywhere in my tree, so that
>> code doesn't get called.
>>
>>> if [ -f configBalloon3-CUED-2011.11 ]; then \
>>> 		cp configBalloon3-CUED-2011.11 /home/wookey/projects/balloon/svn/trunk/build/buildroot/buildroot-2011.11/.config; \
>>> 	else \
>>> 		cp configBalloon3-2011.11 /home/wookey/projects/balloon/svn/trunk/build/buildroot/buildroot-2011.11/.config; \
>>> 	fi;
>>
>> That logic looks, umm, borked. If file exists, copy it - else copy it
>> anyway? That else can't be being called.
>>
>> FTR it's the same code in the menuconfig2 branch which I am building
>> quite often recently, without problems.
>>
>>> cp: cannot create regular file `/home/wookey/projects/balloon/svn/trunk/build/buildroot/buildroot-2011.11/.config': No such file or directory
> 
> OK, so the problem here is that
> /home/wookey/projects/balloon/svn/trunk/build/buildroot/buildroot-2011.11
> does not get created. Only
> /home/wookey/projects/balloon/svn/trunk/build/buildroot/buildroot-2010.11
> does, which seems wrong to me. The config seems to have downloaded the
> 2010.11 version of buildroot, and unpacked it, but then notices a bit
> later in the build that 2011.11 was configured and tries to use the
> config and dir for that. The config exists but the dir doesn't.
> 
> Discouraged by this, and having a loon that works fine until it tries
> to mount yaffs in the real kernel where it segfaults and dies in
> yaffscheckpoint, I tried a few other apporaches:
> a) mount YAFFS FS in initrd kernel, check it was OK and unmount, in
> case on-flash corruption was the issue. Didn't help
> b) set the rootfsflags=no-checkpoint-read option to skip that bit. No
> joy.
> c) load up the 1.0 release kernel from the website for a reality check. 
> This was fairly painless but it fails badly:
> 
> nand_get_flash_type: second ID read did not match 8d,8f against 8f,8f
> No NAND device found!!!
> quickly followed by:
> VFS: Cannot open root device "mtdblock1" or unknown-block(0,0)
> 
> This kernel
> (http://balloonboard.org/files/balloon3/distro/rel-v1.0/binaries/kernel/)
> is 2.6.29.1-pxa270 from May 2010. Seems odd that it can't recognise the
> flash on old CUED boards that earlyier kernels could. Looks like we
> screwed up there.
> 
> Any idea what's wrong there?
> 
There was a mapping change of the nand registers in the cpld at some
point. All to do with the failed attempt at vlio. This means one needs a
matched cpld/kernel pair. The CUED fpga variant may not have been
created/tested.

> So I'm back to trying to build current and see what happens...
> 
> Wookey