Adam Williamson wrote:
Sigh. This shows that once again a purported replacement for a working piece
of software was deployed before it was able to perform the allegedly
replaced tool's most important task, even though the problem was known to
the replacement's developers. We really should not accept this kind of known
regressions.
I'm sort-of volunteered to write the approach I suggested in a
comment
as a new test, but it's going to have to wait until at least post-f21.
Your approach indeed makes sense. It will not cause issues caused by added
Conflicts or the like, but at least it catches the common case. (Just make
sure you also consider Obsoleted packages as "the old package" whose
Provides will no longer be available.)
Kevin Kofler