Bug#198479: general: package coming to sarge before their dependencies : breaks them.

[Sven Luther <sven.luther@wanadoo.fr>]

On Mon, Jun 23, 2003 at 10:14:05AM +0100, Colin Watson wrote:
> On Mon, Jun 23, 2003 at 11:03:42AM +0200, Erwan David wrote:
> > Le Mon 23/06/2003, Colin Watson disait
> > > On Mon, Jun 23, 2003 at 10:36:27AM +0200, Erwan David wrote:
> > > > Several packages like html2ps or apt-file are broken in sarge because
> > > > they were put from sid before their dependencies. The coming to sarge
> > > > of a package should not be done if it makes it uninstallable.
> > > 
> > > This is generally the rule. Sometimes bad things have to happen to a
> > > couple of packages in order to benefit a large number of packages
> > > elsewhere, though.
> > 
> > For html2ps and apt-file, they have been broken in sarge for
> > weeks...
> 
> I haven't looked at them in detail. But:
> 
> html2ps is broken due to perlmagick, which is still at a perl 5.6
> version in testing. This was temporarily necessary because getting perl
> 5.8 was more important than waiting for all of perlmagick's
> dependencies, which remain very messy and complicated; my notes say that
> imagemagick needs the lcms dependency chain, which needs the gdbm
> dependency chain, which needs the libsigc++ dependency chain, which
> needs the libgc dependency chain. Only the last of those is close to
> being ready for testing yet.
> 
> apt-file is broken due to libapt-pkg-perl, which is still at a perl 5.6
> version in testing. Again, this was temporarily necessary because perl
> 5.8 was more important than waiting for all of libapt-pkg-perl's
> dependencies. Right now, apt's release-critical bugs need to be fixed
> before new versions of it and libapt-pkg-perl can move into testing.

Would the propper solution to this not be to remove the broken packages
from testing, until a fixed version is ready to enter testing again from
unstable ? This is what we did for the two ocaml packages which did
bloc the ocaml transition to testing, and they have now entered testing
again without further effort.

Friendly,

Sven Luther