[OE-core] [PATCH v2 00/10] wic: bugfixes & --fixed-size support, tests, selftest: minor fixes

[Maciej Borzęcki]

On Mon, Nov 14, 2016 at 11:16 AM, Burton, Ross <ross.burton at intel.com> wrote:
>
> On 14 November 2016 at 10:02, Maciej Borzęcki <maciej.borzecki at rndity.com>
> wrote:
>>
>> I see that the builds were running with qemux85-64 MACHINE, shouldn't
>> TARGET_ARCH be set to x86-64 then? There's a check that builds
>> syslinux(-native) for x86 archs only since it's not buildable for
>> anything else. Same for HDDIMG.
>
>
> Not quite:
>
> $ MACHINE=qemux86 bitbake core-image-minimal -e | grep ^TARGET_ARCH=
> TARGET_ARCH="i586"
>
> $ MACHINE=qemux86-64 bitbake core-image-minimal -e | grep ^TARGET_ARCH=
> TARGET_ARCH="x86_64"
>
> (had to patch bitbake.conf to export TARGET_ARCH)

Do you think it makes sense to replace TARGET_ARCH check with something
like this instead:

        arch = get_bb_var('HOST_ARCH', 'core-image-minimal')

        is_x86 = arch in ['i586', 'i686', 'x86-64']
        if is_x86:
            self.write_config('IMAGE_FSTYPES += " hddimg"\n'
                              'MACHINE_FEATURES_append = " efi"\n')

Basically, I do not want to build syslinux or set config options that
are usable on x86 target when building for other targets.


>
>>
>> One more question, the assert raised when comparing parted's output is
>> caused by different locale. Do you think it would be ok to just enforce
>> LC_ALL=C when running commands under wic?
>
>
> Yes, enforcing a  locale is essential when comparing output of commands.
> Bitbake switches to en_US.UTF8 on startup for the same reason, so feel free
> to use that or C.

I suppose I'll just enforce that in oe-selftest.



-- 
Maciej Borzecki
RnDity