We've got requests such as this one (upgrade lablgtk to 2.10.0 in Fedora 8):
https://bugzilla.redhat.com/show_bug.cgi?id=424821 and a related one from Peter Lemenkov to support mldonkey.
As I commented in that BZ, if we do that upgrade then at least 4 dependent packages need to be rebuilt as well. Furthermore anyone writing their own software on F8 which used any of these packages would need to at least recompile.
On the other hand I took the opportunity this week to upgrade several packages in Rawhide (including lablgtk 2.10.0). That shouldn't be a problem because we expect Rawhide to break things.
So I wonder if we need a policy that once a version of Fedora has been released, we don't upgrade packages on a whim, but only if there is a serious need (eg. security or some otherwise unresolvable bug).
What do people think? I don't want to be too restrictive. If people don't mind recompiling that's another matter. Perhaps we should only have this policy for the base OCaml package and some other "vital" packages (eg. findlib) and leave the decision on a case-by-case basis for other packages?
Rich.
2008/2/16, Richard W.M. Jones rjones@redhat.com:
We've got requests such as this one (upgrade lablgtk to 2.10.0 in Fedora 8):
https://bugzilla.redhat.com/show_bug.cgi?id=424821 and a related one from Peter Lemenkov to support mldonkey.
MLDonkey is a killerapp for OCaml. I think satisfying its requirements must be high priority task. I don't see many problems in rebuilding few applications with latest ocaml libraries.
Let's create exact list of obstacles which prevents from updating ocaml for mldonkey because all I've heard so far are speculative estimates "that something will be wrong", OK there is something wrong right now - the latest version (with a couple of critical bugs closed) most popular ocaml application, e.g. mldonkey can't be built with currently shipped ocaml libraries w/o breaking its requirements.
ocaml-devel@lists.fedoraproject.org
-
Peter Lemenkov -
Richard W.M. Jones