[Debian-med-packaging] why x32?

[Adrian Bunk]

On Wed, Jun 07, 2017 at 08:01:44PM +0200, Adam Borowski wrote:
> > falcon is now only built for 64bit architectures.
> 
> Except that the package that has just been uploaded has:
> Architecture: alpha any-amd64 arm64 hppa mips64el ppc64 ppc64el s390x sparc64
> which includes x32 ("any-amd64" unintuitively means also x32).

Thanks for catching that.

> The package has built correctly, and seems to have passed the testsuite.
> 
> I don't know why falcon needs 64-bit, but x32 despite having 32-bit pointers
> has the full set of 64-bit integer registers, apparently that's enough.

Afif (maintainer of falcon) explained in #863929:
  It actually does not even make sense to have this package available for
  a 32-bit architecture since, for any actual use (de novo genome
  assembly), it will require more memory than would be addressable on a
  32-bit system.

At that point it was believed that the testsuite failures were 
32bit-only, but when I prepared the NMU today I got the same
test failure on amd64.

The testsuite also sometimes passed on i386 on the buildds.

My conclusion before the upload was that there is no strict relationship 
between the testsuite failures and 32bit in either direction, and these
two changes in my NMU are therefore not strictly related.

The problematic testsuite is now disabled on all architectures,
and #864376 submitted to track that this gets fixed.

> Thus, please explain -- if you don't, the next upload will reintroduce the
> x32 build which may or may not be what you want.

Since x32 is not a stretch release architecture it does at least not
matter right now, I'll submit a (non-RC) bug so that it won't be
forgotten.

> Meow!

Thanks!
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed