persistent clutter and ghc breakage (was Re: rawhide report: 20090803 changes)

[Peter Robinson]

> There appear to be a boatload of broken deps for packages based on
> clutter (presumably because of API bump from 0.9 to 1.0) that have
> been listed here for over a week and no movement to fix most of them.

It was actually broken on the move from 0.8 to 0.9 (and there were a
couple of smaller api breakages during the 0.9 process).

> Have the other apps been ported to the new API so they just need to be
> built, or have they not yet been ported?  (I tried to find an update
> of cluttermm to 1.0 but only 0.9 versions are on the gnome site).

No, they haven't in most cases been ported yet. All the pyclutter*
stuff and cluttermm are still to be ported to the new API. The 0.9
versions if they are available might/should compile against 1.0.

> If the latter, it would be nice the maintainer of clutter did an
> inventory of dependent-packages and co-ordinate with downstream
> consumers of clutter to make sure that there was a plan to update
> those packages, or built a compat- package if necessary.  Otherwise
> users of rawhide are left with huge swathe of breakage:

[snip]

> I must be a broken record on this, but it would be nice if maintainers
> could make announcements to all concerned (e.g. on
> fedora-devel-announce) about API/soname etc. bumps in major
> libraries/frameworks *before* they are put into motion so that the
> dependent packages can be rebuilt in a timely manner.  Or if that
> causes too much breakage where, for example, many apps aren't ready to
> be ported to the new version, then consider delaying the update for a
> while or introducing compat- packages.

No idea about the ghc stuff but until recently there wasn't really a
owner to the clutter* stuff.

> Otherwise breakage goes on for weeks and necessitating a rush to
> rebuild the software by provenpackagers just before a freeze which
> leaves an inadequate time for user/developer testing on rawhide.

Peter