Re: d-i, pulling in fdisk-udeb
On Wed, Oct 01, 2003 at 12:56:39PM +0200, Steinar H. Gunderson wrote:
> On Tue, Sep 30, 2003 at 07:28:37PM -0700, Matt Kraai wrote:
> > Because it can't use just any partioning program, it uses a
> > particular one (or, actually, a particular one, two, or three)
> > depending on the architecture. If we depend on a virtual
> > partitioning-program package, it could install parted-udeb only
> > even though that doesn't work on that particular architecture.
> Am I missing something on subarchitecture vs. architecture here? Once again:
> If parted is not (for our purposes) usable on hppa, we do not build it for
> hppa. (That should be quite easy to change.) Likewise, we only build the
> other udebs for the platform we intend it to be used on. Thus, there will be
> one (and only one) udeb that provides "partitioning-program" (or
> "preferred-di-partitioning-program" if you want :-) ) for each architecture,
> and anna can easily pick that one.
On New World PowerPC systems, we use mac-fdisk. On PREP PowerPC
systems, we use fdisk. Suppose that we implement your
suggestion. anna sees that fdisk-udeb provides
partitioning-program. It installs it. But wait, we are
installing on a New World system. So partitioning fails.
Easy to fix, right? Just don't make fdisk-udeb on PowerPC. So
now I am installing on my (hypothetical) PREP system. anna
installs mac-fdisk-udeb since that is the only package that
provides partitioning-program. So partitioning fails.
A virtual package dependency is not enough.