Bug#746998: pyfftw: FTBFS on architectures without long double

[YunQiang Su]

On Thu, 22 May 2014 08:00:51 +0100 Ghislain Vaillant <ghisvail at gmail.com> wrote:
> Hi Aurelien,
>
> Thanks for reporting this issue.
>
> I'd have to check with upstream the best approach for dealing with this.
> I am not sure whether you can  enable / disable support for specific
> precisions in pyfftw. I think the author assumed that all these
> precisions were available.
>
> I could restrict the package build to selected architectures that
> provide all precisions for now, whilst upstream finds a solution to
> this. How does that sound ? In this case, I'd need to know which
> architectures are currently failing besides mips64.

This bug is about mipsel/mips aka 32bit mips, not about mips64(el).
You are confused due to he uses a 64bit kernel with 32bit userland.

I noticed that in fftw3, long double support is enabled on i386,
while the long double on i386 is 12Byte, quite strange,
as other ones are all 16Byte.

>
> Please let me know what you think,
>
> Cheers,
>
> Ghislain
>
>