[yocto] [meta-raspberrypi] du giving non-integer result?

[Trevor Woerner]

On 09/23/15 16:12, Andrei Gherzan wrote:
> On Thu, Sep 17, 2015 at 02:14:21PM +1000, Jonathan Liu wrote:
>> On 11 September 2015 at 11:37, Trevor Woerner <twoerner at gmail.com> wrote:
>>> From the feedback, it looks like I'm the only one seeing this issue.
>>> That's okay, but I'll update this issue with new information in case
>>> it's useful to others. Today I also had some other strangeness with the
>>> "make sd card image" portion of a build, this time for cubietruck
>>> (meta-sunxi):
>>>
>>>
>>>     ERROR: Error: The image creation script
>>>     '/z/tmp/build/tmp/work/cubietruck-poky-linux-gnueabi/core-image-minimal/1.0-r0/temp/create_image.sunxi-sdimg'
>>>     returned 1:
>>>     0+0 records in
>>>     0+0 records out
>>>     0 bytes (0 B) copied, 3.0606e-05 s, 0.0 kB/s
>>>     expr: non-integer argument
>>>     WARNING:
>>>     /z/tmp/build/tmp/work/cubietruck-poky-linux-gnueabi/core-image-minimal/1.0-r0/temp/create_image.sunxi-sdimg:1
>>>     exit 1 from
>>>       parted -s
>>>     /z/tmp/build/tmp/deploy/images/cubietruck/core-image-minimal-cubietruck-20150910182624.rootfs.sunxi-sdimg
>>>     unit KiB mkpart primary ext2 $(expr ${BOOT_SPACE_ALIGNED} \+ 2048)
>>>     $(expr ${BOOT_SPACE_ALIGNED} \+ 2048 \+ 10240.0)
>>>
>>>
>>>
>>> Not only am I getting the floating point again, but it looks like
>>> "BOOT_SPACE_ALIGNED" isn't getting expanded!
>>>
>>> The frustrating part is that simply rerunning "bitbake
>>> core-image-minimal" fixes the problem. The second time it completes
>>> without error.
>>> --
>> I sent a patch to openembedded-core mailing list for ROOTFS_SIZE not
>> being an integer.
>> It seems http://git.openembedded.org/openembedded-core/commit/?id=af37bb8ae71c4f932e2126bb620b3fb2b11cc466
>> results in ROOTFS_SIZE now being set as a bitbake variable.
>> meta-raspberrypi will need to be updated to not have a conflicting
>> ROOTFS_SIZE variable as ${ROOTFS_SIZE} is being substituted by bitbake
>> instead of the shell there.
> Hi guys,
>
> I'm testing and reviewing this as we speak. As well, it would be a good idea to drop
> the sd card generation class for good and replace it completely by wic support.
>
> Something to be done. Wnat to pick this up? I know that there were people
> around testing wic and we had configuration too.

Jonathan's patch does work for me, and it was premature of me to put
[meta-raspberrypi] in the subject line since I saw this same failure on
other BSPs (and other people reported even more failures on other BSPs).
But it wouldn't hurt to move to wic either. It's probably about time I
started using it anyway :-)