Bug#677778: Conversion of libbonobo to multiarch

[Goswin von Brederlow]

On Sun, Nov 25, 2012 at 10:30:12PM +0100, Josselin Mouette wrote:
> Hi,
> 
> Le dimanche 25 novembre 2012 à 20:22 +0100, Goswin von Brederlow a
> écrit : 
> > this bug is still blocking the ia32-libs-gtk multiarch transition that
> > really needs to go into wheezy. The only issue with the patch from the
> > bugreport seems to be a question about making it "m-a: same".
> > 
> > Joss: "m-a: same" allows future packages to specify a dependency of
> > "bonobo-activation-server:any" for cases where the architecture of
> > the activation server is irelevant to the depending package. This
> > feature won't be usable in wheezy. Also, worst case, if it is wrong
> > then future packages must not use the "*:any" syntax. So this is
> > benign. Packages must test this works before setting "*:any".
> 
> What is not clear in the ???I object to this patch as is??? sentence?

Maybe the fact that you didn't say "I object to this patch as is" [1].

What you did was raise a question about bonobo-activation-server being
m-a: same (as opposed to not being multiarch at all), which I tried to
answere.

Any other concerns?
 
> I fail to see the point in migrating bonobo to multi-arch if the
> resulting packages only work for one architecture. You are going to
> achieve zilch.
> 
> I agree it would be nice to have it in wheezy but it is non-trivial work
> and low on my to-do list compared to polishing GNOME 3.4.
> 
> Cheers,

The point would be to make libbonobo:amd64 and libbonobo:i386 coinstallable
and therefore support 32bit packages that (directly or indirectly) depend
on libbonobo without the need to remove all 64bit packages depending on it.

For the past ~10 years ia32-libs(-gtk) has had libbonobo:i386
installed with biarch hacks and duct tape. In all that time there
has been no complaints that I'm aware of about the
bonobo-activation-server starting things for the wrong architecture.
So fully multiarchifying that component indeed has no high priority.
It either already works or isn't used by relevant 32bit software.

But making libbonobo:i386 installable has real benefits to users.
As for bonobo-activation-server, worst case, as I see it, will be that
the user will have the choice to make it work in 32bit or in 64bit
but not both. Unless you have concerns that the patch will break bonobo
in a purely mono-arch environment ...

MfG
	Goswin

--
[1] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=677778