Comment 3 for bug 911199

I had a look over the debian patch and compared it to one I was already working on.

Attached is a new version, which I'll also add to the debian bug. It may not reach debian in time for us to sync for precise, if so could this one be applied to precise?

A couple of things:
 * I'm using $(DEB_HOST_MULTIARCH) in debian/control instead of just /*/ - I *think* that's what's recommended
 * Changed depends on cpp to cpp:any due to (discussion from #ubuntu-devel):

 <TREllis> I'm trying to multi-arch libidl0, it depends on cpp, which means the i386 package still won't install on amd64, needs cpp:i386... does that mean cpp needs multi-arching?
<cjwatson> TREllis: Multi-Arch: foreign on cpp *might* be the right thing if the version from another architecture will work; but I would have thought that cpp has architecture-specific defaults in it ...
<cjwatson> TREllis: we don't have provision for real multi-arch of binaries yet
<TREllis> cjwatson: right I thought that might be the case. That's one of my blockers at the moment whilst trying to fix a chain of multi-arch dependancies
...
<slangasek> cjwatson, TREllis: cpp is already marked Multi-Arch: allowed; packages that need to be able to depend on a foreign-arch version of cpp need to be updated to depend on cpp:any
<slangasek> TREllis: because it's not correct to say that a cpp package of any architecture will satisfy the dependency on cpp for a package of any other architecture, because of the per-arch defaults cjwatson mentions.