[OE-core] [PATCH 2/2] [meta] atk: Specify gettext-native dependency

[Burton, Ross]

What was missing all along is that we're talking specifically about
atk-native here, which disables NLS explicitly.  As Meson doesn't yet
expose a way to turn off i18n we need to stop that.  For now, adding
USE_NLS=yes is a suitable workaround (and what glib already does).
Marko, can you send this patch?

Arguably, meson.bbclass could do this, but that would be forcing a
gettext dependency on everything using Meson.

Ross



On 1 July 2018 at 19:20, Marko Lindqvist <cazfi74 at gmail.com> wrote:
>  3) Adding 'USE_NLS = "yes"' to atk recipe fixes the problem for me.
>
>
>  - ML
>
>
> On 1 July 2018 at 21:03, Marko Lindqvist <cazfi74 at gmail.com> wrote:
>>  I think the problem is not with INHIBIT_DEFAULT_DEPS, but with
>> USE_NLS on the next couple of lines. With USE_NLS disabled, the
>> automatic dependency is gettext-minimal-native (not providing
>> xgettext, I think), and not gettext-native.
>>
>>  1) tmp-glibc/work/aarch64-oe-linux/atk/2.28.1-r0/recipe-sysroot-native/sysroot-providers/gettext-minimal-native
>> is there
>>  2) Adding gettext-minimal-native as atk dependency does not fix the
>> issue like adding gettext-native dependency does
>>
>>
>>  - ML
>>
>>
>> On 30 June 2018 at 21:43, Marko Lindqvist <cazfi74 at gmail.com> wrote:
>>>  This error happens to me with both qemux86 and qemuarm64. Haven't
>>> tested any others yet.
>>>
>>>
>>>  - ML
>>>
>>> On 15 June 2018 at 20:40, Alistair Francis <alistair23 at gmail.com> wrote:
>>>> On Mon, Jun 4, 2018 at 4:18 PM, Burton, Ross <ross.burton at intel.com> wrote:
>>>>> That implies that INHIBIT_DEFAULT_DEPS is also evaluating to true,
>>>>> which it shouldn't be in ATK.
>>>>>
>>>>> I'd definitely be checking if that is the case, and if you can
>>>>> replicate the failure with e.g. qemuarm.
>>>>
>>>> I can't replicate it with qemuarm and checking with bitbake -e I can't
>>>> see anywhere that the variable is set. Any ideas on what to try to
>>>> figure out how it is set?
>>>>
>>>> Alistair
>>>>
>>>>>
>>>>> Ross
>>>>>
>>>>> On 4 June 2018 at 19:46, Alistair Francis <alistair23 at gmail.com> wrote:
>>>>>> On Mon, Jun 4, 2018 at 10:20 AM, Alexander Kanavin
>>>>>> <alexander.kanavin at linux.intel.com> wrote:
>>>>>>> On 06/04/2018 08:17 PM, Alistair Francis wrote:
>>>>>>>>
>>>>>>>> On Mon, Jun 4, 2018 at 3:06 AM, Burton, Ross <ross.burton at intel.com>
>>>>>>>> wrote:
>>>>>>>>>
>>>>>>>>> So why isn't the inherit gettext sufficient?
>>>>>>>>
>>>>>>>>
>>>>>>>> I'm not sure. Here is the error I see when compiling for RISC-V before
>>>>>>>> this patch:
>>>>>>>
>>>>>>>
>>>>>>> So is the error specific to risc-v? You should dig deeper in that direction
>>>>>>> then.
>>>>>>
>>>>>> Ok, it looks like the line: if d.getVar('INHIBIT_DEFAULT_DEPS') and
>>>>>> not oe.utils.inherits(d, 'cross-canadian'): is evaluating as true in
>>>>>> gettext.bbclass so we don't depend on gettext-native.
>>>>>>
>>>>>> Do I need to ensure cross-canadian is set for RISC-V builds then?
>>>>>>
>>>>>> Alistair
>>>>>>
>>>>>>>
>>>>>>> Alex
>>>> --
>>>> _______________________________________________
>>>> Openembedded-core mailing list
>>>> Openembedded-core at lists.openembedded.org
>>>> http://lists.openembedded.org/mailman/listinfo/openembedded-core