On Tue, 2010-02-16 at 16:17 -0500, Bill Nottingham wrote:
Jesse Keating (jkeating(a)redhat.com) said:
> On Tue, 2010-02-16 at 20:09 +0000, Richard W.M. Jones wrote:
> > On Tue, Feb 16, 2010 at 08:06:16PM +0000, Rawhide Report wrote:
> > > 1:libguestfs-1.0.84-1.fc13.i686 requires
/lib/libdevmapper-event-lvm2snapshot.so.2.02
> > > 1:libguestfs-1.0.84-1.fc13.i686 requires /lib/libntfs-3g.so.73
> > > 1:libguestfs-1.0.84-1.fc13.i686 requires
/lib/libdevmapper-event-lvm2mirror.so.2.02
> > > 1:libguestfs-1.0.84-1.fc13.i686 requires /lib/libntfs-3g.so.73.0.0
> >
> > Fixed already. Is there going to be a freeze on sonames at
> > some point?
>
> We don't necessarily have a soname freeze. I kind of wish we did, but
> I'm sure that'd be met with lots of loud yelling.
Looking at these requirements, your package is set up such that it will
fail even if the soname doesn't change (but the filename does).
That seems needlessly fragile.
libguestfs is using its own find-requires script, which is what appears
to be generating those. From a quick look it's not clear whether it
really needs exactly those files or whether just a soname match would be
good enough.
http://bugzilla.redhat.com/547496 seems to be the relevant
bug.
In general we should probably carp about file reqs that look like DSO
deps, especially since rpm has %{isa} now. But this case might be
special.
- ajax