Bug#733578: hwinfo/21.0-1 needs updates / FTBFS of libx86emu

[Tomasz Buchert <tomasz.buchert@inria.fr>]

Hi all,

[...]

> > Hum, and it's not a missing dependency to libc6-dev ? (I'm maybe wrong)
> 
> Yeah, seems to me as well. I'll try to test it out.

We were both wrong - sys/io.h does not exist on the architectures in question,
for example: http://packages.debian.org/wheezy/mips/libc6-dev-mips64/filelist

The calls used are 'inb', 'outb' and friends (see 'man inb') and
are architecture-specific therefore I think we should restrict
build to only architectures that provide io.h (as Johann proposed).
There may be a slight chance to build them
(see https://www.linux-mips.org/archives/linux-mips/2002-04/msg00085.html),
but I wouldn't bother unless somebody files a bug :).

Now, that poses a question whether we should restrict hwinfo in a similar way.
The old hwinfo is BD-Uninstallable on non-Linux archs (because it depends on
linux-kernel-headers, see https://buildd.debian.org/status/package.php?p=hwinfo)
and therefore it won't build on hurd, kfreebsd, etc. My proposal is to:

  1) release a new version of libx86emu 1.4-2 with Sebastien's fixes
     and restricted architectures (btw. Seb, can you update changelog with
     git-dch --auto ? I tried to do it myself, but I didn't know how
     to describe some of your changes)

  2) release hwinfo with explicit dependency on the above so that
     it will be automatically BD-Uninstallable for architectures that
     were restricted by libx86emu package

It's not perfect, but it will let us finally release hwinfo. Moreover, I have
doubts that anybody will miss hwinfo on the restricted architectures. If so,
a bug will be created and we'll start to worry, right?

I pushed the restricted version of libx86emu to the git repository.

Both libx86emu and hwinfo build lintian-clean in my jessie pbuilder (well,
excepting debian-watch-may-check-gpg-signature, to be fair). Hwinfo debian/control
must be updated if you agree what I said above.

Cheers,
Tomasz