Santiago Vila:
> On Fri, Nov 27, 2015 at 11:12:50AM +0100, Jérémy Bobbio wrote:
> > I think the question can be phrased as: should architecture-independent
> > packages be buildable on all architectures?
> > 
> > My own answer would be: yes, as long as they don't mandate a particular
> > architecture in their Build-Depends-Indep.
> > 
> > Is there something wrong with this answer?
> My answer would be: Not necessarily.
> In theory, the build process could need to compile a program which is
> used to create some file which is later used in the binary-indep target.
> Such program does not need to be portable because, well, some programs
> are just not portable. Not in the sense that it's inherently non-portable
> but just in the sense that nobody had the time to "port" it yet.
> If we had a case like this, an architecture field in the source
> package saying "i386 all" would mean that we could do all this:
> "dpkg-buildpackage" under i386 to build the i386.deb and the all.deb.
> "dpkg-buildpackage -A" under i386 to build only the all.deb.
> "dpkg-buildpackage -B" under i386 to build only the i386.deb.
> and at the same time it would be possible that the source package is
> just not designed or ready to build the all.deb under, say, amd64.
> In most cases it will probably work, but I do not see it as a
> requirement that we should take for granted. For example, I would not
> submit it as a bug (unless I find a way to fix it first).

While I agree, I really think you've identified a hole in the policy
here. Either we need a formal agreement that "i386 all" = "Arch: all"
packages must be built on i386 or an extra field somewhere to indicate

I would suggest moving the discussion to a better suited communication
channel as this is a general issue. Maybe debian-dpkg@l.d.o?

Thanks for finding this. :)

Lunar                                .''`.                    : :Ⓐ  :  # apt-get install anarchism
                                    `. `'` 

Attachment: signature.asc
Description: Digital signature

Reproducible-builds mailing list

Reply via email to