Re: [master][krogoth][PATCH] python-numpy: fix build failure with python-matplotlib

[akuster808 <akuster808@gmail.com>]

On 05/16/2016 08:35 PM, Khem Raj wrote:
> 
>> On May 16, 2016, at 8:34 PM, akuster808 <akuster808@gmail.com> wrote:
>>
>> On 5/16/16 8:09 PM, Khem Raj wrote:
>>>
>>>> On May 16, 2016, at 7:26 PM, akuster808 <akuster808@gmail.com> wrote:
>>>>
>>>>
>>>>
>>>> On 05/16/2016 09:44 AM, Khem Raj wrote:
>>>>>
>>>>>> On May 16, 2016, at 9:17 AM, Armin Kuster <akuster808@gmail.com> wrote:
>>>>>>
>>>>>> From: Armin Kuster <akuster@mvista.com>
>>>>>>
>>>>>> numpy/core/include/numpy/npy_common.h:149:10: error:
>>>>>> |          #error Unsupported size for type off_t
>>>>>>
>>>>>> Signed-off-by: Armin Kuster <akuster@mvista.com>
>>>>>> ---
>>>>>> meta/recipes-devtools/python/python-numpy/aarch64/_numpyconfig.h | 1 +
>>>>>> 1 file changed, 1 insertion(+)
>>>>>>
>>>>>> diff --git a/meta/recipes-devtools/python/python-numpy/aarch64/_numpyconfig.h b/meta/recipes-devtools/python/python-numpy/aarch64/_numpyconfig.h
>>>>>> index 191e5a1..109deb0 100644
>>>>>> --- a/meta/recipes-devtools/python/python-numpy/aarch64/_numpyconfig.h
>>>>>> +++ b/meta/recipes-devtools/python/python-numpy/aarch64/_numpyconfig.h
>>>>>> @@ -11,6 +11,7 @@
>>>>>> #define NPY_SIZEOF_PY_INTPTR_T 8
>>>>>> #define NPY_SIZEOF_PY_LONG_LONG 8
>>>>>> #define NPY_SIZEOF_LONGLONG 8
>>>>>> +#define NPY_SIZEOF_OFF_T 8
>>>>>
>>>>> Why does the test to select size of off_t fails ?
>>>> core/include/numpy/npy_common.h
>>>>
>>>> #if NPY_SIZEOF_OFF_T == NPY_SIZEOF_SHORT
>>>>       #define NPY_OFF_T_PYFMT "h"
>>>>   #elif NPY_SIZEOF_OFF_T == NPY_SIZEOF_INT
>>>>       #define NPY_OFF_T_PYFMT "i"
>>>>   #elif NPY_SIZEOF_OFF_T == NPY_SIZEOF_LONG
>>>>       #define NPY_OFF_T_PYFMT "l"
>>>>   #elif NPY_SIZEOF_OFF_T == NPY_SIZEOF_LONGLONG
>>>>       #define NPY_OFF_T_PYFMT "L"
>>>>   #else
>>>>       #error Unsupported size for type off_t
>>>>   #endif
>>>>
>>>> its gotta match one of them. X86-64 has NPY_SIZEOF_OFF_T  value set.
>>>
>>> yes thats the use of it. i was asking for place where it is set, most probably during configure
>>>
>>
>> There is no configure for this python package. It uses hardcoded values
>> in config.h or _numpyconfig.h per arch. X86-64 has the NPY_SIZEOF_OFF_T
>> defined in _numpyconfig.h. To keep it consistent, I put the Aarch64
>> changes in the same place.
> 
> is this config.h or _numpyconfig.h generated too ?
> see if thats the case

Sorry, missed this email last night.

Not sure what you mean by generated. Both config and _numpyconfig.h are
files that reside under an "arch" dir and brought in when building. I
don't know why this needed to be done this way. Don't you recall, you
did a patch for this package too ; ) Maybe Ross remembers,  his
fingerprints are all over this package.

- armin

> 
>>
>> - armin
>>
>>>>
>>>>
>>>>> we might need similar fix for mips64 too.
>>>> yep, I just built mips64 and ppc64 and they both fail like aach64.
>>>>
>>>> will send v2 to fix those as well.
>>>
>>> OK
>>>
>>>>
>>>> -armin
>>>>
>>>>>
>>>>>> #define NPY_NO_SMP 0
>>>>>> #define NPY_HAVE_DECL_ISNAN
>>>>>> #define NPY_HAVE_DECL_ISINF
>>>>>> --
>>>>>> 2.3.5
>>>>>>
>>>>>> --
>>>>>> _______________________________________________
>>>>>> Openembedded-core mailing list
>>>>>> Openembedded-core@lists.openembedded.org
>>>>>> http://lists.openembedded.org/mailman/listinfo/openembedded-core
>