[Yum-devel] [PATCH] Prospective fix for https://bugzilla.redhat.com/show_bug.cgi?id=519172

[Tim Lauridsen]

On 09/08/2009 09:14 PM, Seth Vidal wrote:
> If we are running compare_providers() and any of the providers are updates for something installed
> then score them up just slightly and make sure to not REMOVE the items which are NOT updates
> for something installed. There still should be an obsoletes catch in here - maybe even to make sure
> that an obsoleted pkg is never considered AT ALL in compare_providers()
>
> however this makes the bug work out correctly and still has all the unittests working
> ---
>   yum/depsolve.py |   11 ++++++++++-
>   1 files changed, 10 insertions(+), 1 deletions(-)
>
> diff --git a/yum/depsolve.py b/yum/depsolve.py
> index d6b1b02..467f657 100644
> --- a/yum/depsolve.py
> +++ b/yum/depsolve.py
> @@ -1083,7 +1083,16 @@ class Depsolve(object):
>                   # could play a part ... this probably needs a better fix.
>                   newest = sorted(rpmdbpkgs)[-1]
>                   if newest.verLT(pkg):
> -                    ipkgresults[pkg] = 0
> +                    # give pkgs which are updates just a SLIGHT edge
> +                    # we should also make sure that any pkg
> +                    # we are giving an edge to is not obsoleted by
> +                    # something else in the transaction. :(
> +                    # there are many ways I hate this - this is but one
> +                    ipkgresults[pkg] = 5
> +            else:
> +                # just b/c they're not installed pkgs doesn't mean they should
> +                # be ignored entirely. Just not preferred
> +                ipkgresults[pkg] = 0
>
>           #  This is probably only for "renames". What happens is that pkgA-1 gets
>           # obsoleted by pkgB but pkgB requires pkgA-2, now _if_ the pkgA txmbr

Look OK to me.

Tim