jonathan at jonmasters.org
Mon Oct 22 00:32:01 UTC 2007
On Mon, 2007-10-22 at 20:14 +1000, Dave Airlie wrote:
> On Sun, 2007-10-21 at 20:11 -0400, Jon Masters wrote:
> > On Sun, 2007-10-21 at 22:18 +0100, Ian Chapman wrote:
> > > Hans de Goede wrote:
> > >
> > > > The current multilib solution in rpm is far from pretty, it works well,
> > > > but definitively has downsides. I think as is its a reasonable
> > > > comprimise, lets not add bandaids and patches to it for issues which
> > > > should be solved elsewhere, I feel the pain of maintainers getting these
> > > > bugs (I got 15 of them), but they are fixable without requiring the
> > > > addition of yet another multilib kludge to rpm.
> > >
> > > Well the question is still really where should these issues ultimately
> > > be solved? Is kludging the rpms any more elegant than patching rpm? I
> > > must admit I have no idea how other distro's deal with these kind of
> > > issues.
> > Without ranting aimlessly, IMO the only real solution is to stop
> > kludging rpm, yum, etc. and split out multilib libraries properly - and
> > if needed, seek and get approval for a bin64/bin32 with alternatives
> > system. Hacking RPM to simply ignore the fact that two packages provide
> > the same file is not the solution.
> I have to agree, I didn't know about the stupid bin behaviour and
> recently started flushing 64-bit packages from my ppc64 machine, and
> ended up removing a fair few files from /usr/bin that were still being
> provided by the 32-bit packages.
> it just screamed kludge to me..
It obviously was a kludge. And a great one - Jeremy, Bill, and other
*smart* people (so don't get me wrong guys!) had to come up with a
solution at some point when this came up, but it's long since time to
take a big breath and think about a longer term fix.
Can we agree that needs to happen? If so, I'll happily actually get
involved in implementing a fix, as I've said to these folks in face to
face conversations about multilib before!
More information about the fedora-devel-list