[Buildroot] [PATCH 3/5] python-cffi: add host variant

[Yegor Yefremov]

Hello Oli,

On Thu, Dec 31, 2015 at 8:20 AM, Oli Vogt <oli.vogt.pub01 at gmail.com> wrote:
>
> On 27 December 2015 at 11:54, Yegor Yefremov <yegorslists at googlemail.com>
> wrote:
>>
>> On Fri, Dec 25, 2015 at 10:44 PM, Yann E. MORIN <yann.morin.1998 at free.fr>
>> wrote:
>> > Yegor, All,
>> >
>> > On 2015-12-25 20:58 +0100, Yegor Yefremov spake thusly:
>> >> On Fri, Dec 25, 2015 at 8:12 PM, Yann E. MORIN
>> >> <yann.morin.1998 at free.fr> wrote:
>> >> > On 2015-12-25 20:00 +0100, Thomas Petazzoni spake thusly:
>> >> >> On Thu, 24 Dec 2015 23:12:49 +0100, Yegor Yefremov wrote:
>> >> >> > Target variant doesn't need pycparser. cffi's setup.py seems to
>> >> >> > distinguish, if we install real package or cross-compiled one:
>> >> >> >
>> >> >> >         install_requires=[
>> >> >> >             'pycparser',
>> >> >> >         ] if cpython else [],
>> >> >>
>> >> >> I don't understand how this "if cpython" is relevant to make the
>> >> >> different between the native (host) variant or the cross-compiled
>> >> >> (target) variant. cpython is defined as is:
>> >> >>
>> >> >>     cpython = ('_cffi_backend' not in sys.builtin_module_names)
>> >> >>
>> >> >> Can you give some more details?
>> >> >
>> >> > Yegor and I briefly discussed this on IRC yesterday, but I'm not
>> >> > really
>> >> > convinced by this reply either. I forgot to reply yesterday...
>> >> >
>> >> > What I understand, however, is that they want to differentiate
>> >> > between
>> >> > cpython (the 'ofiicial' native Python interpreter), and PyPy (the
>> >> > Python
>> >> > interpreter written in Python).
>> >> >
>> >> > Of course, I don't see how this all works. It seems in
>> >> > cross-compilation. there differentiation criterai beliveves it is
>> >> > running under PyPy amd thus they do not need pycparser.
>> >> >
>> >> > But thtat's just incorrect, because we may have a host-python with
>> >> > cffi_backend at one point, and thus the check above will eventually
>> >> > differentiate toward cpython, and we'd nreak again.
>> >> >
>> >> > So, two things:
>> >> >   - their check is wrong,
>> >> >   - we probably need host-pycparser
>> >> >
>> >> > At least, that's what I understood...
>> >>
>> >> We have host-pycparser: http://patchwork.ozlabs.org/patch/560970/
>> >
>> > Well, I meant that we should need host-pycparser as a dependency of
>> > python-cffi.
>> >
>> > But anyway, there is more to this: we really need to understand what is
>> > going on...
>>
>> I still don't really understand the difference between host and target
>> cffi requirements, but at least I can now shed some light on cpython
>> definition. Below is sys.builtin_module_names() description:
>>
>> "A tuple of strings giving the names of all modules that are compiled
>> into this Python interpreter."
>>
>> So our host-python version won't show '_cffi_backend' even if it is
>> installed on the system, because it is not built-in. But PyPy
>> (http://pypy.org/) comes with this backend out of the box.
>>
>> Will debug further.
>>
>> @Oli: what are you using cffi for?
>>
>> Yegor
>
> Hello Yegor, All
>
>  It's been some time since I worked with buildroot. At the time I was trying
> to include smbus-cffi <https://pypi.python.org/pypi/smbus-cffi> into
> buildroot's python modules and this needs cffi. I wanted to try an
> alternative module to pysmbus which did not integrate right away.
>  Meanwhile the module pysmbus <https://pypi.python.org/pypi/pysmbus/0.1> has
> been included in buildroot, thank you.

Thanks for clarification. I'll take a look at smbus-cffi.

Happy New Year to all Buildrooters!

Yegor