Re: question on hurd-i386 Debian architecture

[Pjotr Kourzanov <peter.kourzanov@xs4all.nl>]

Martin Michlmayr wrote:

> * Marc 'HE' Brockschmidt <he@ftwca.de> [2006-03-13 00:04]:
>  
>
> > > I am adding some additional archs to my local installation like
> > > i386-uclibc, which makes hurd-i386 an exception to the rule of
> > > having the CPU arch first and the OS name the next.
> > >      
> > There's also kfreebsd-{i386,amd64}, so why don't you use uclibc-i386?
> >    
>
> Actually, I disagree.  To me it makes perfect sense the way it
> currently is, namely:
>  kernel-arch-libc
>
> kernel and libc can be empty when they're the default (Linux and
> glibc respectively).
>
> The uclibc port uses Linux so I think i386-uclibc is fine.  There
> could be kfreebsd-i386-uclibc in the future, I suppose, or something
> like that.
>  
Makes sense. I would prefer however to stick with gcc's convention
of having arch(-vendor)-kernel-libc, however, kernel-arch(-vendor)-libc 
is also
suitable.

Dpkg maintainer(s), what do you think is the correct procedure for
additing these things i.e., extra -vendor and -libc fields? I already
have a patch for dpkg package which adds-in uclibc variants...