Existing code has a serious performance problem with package uninstallation on bigger installations (a lot of transactions or a lot of packages in the system) interfering with the system upgrade. Moreover, database scheme has significantly changed - that would require another, relatively complex transformation script and cause issues with rollback. In addition, SWDB API has changed again and a lot of stuff is being moved to C++.
I definitely don't agree with cherry-picking the old version, it would cause a big mess. Consider upstream SWDB version being a proof of concept - it works, but it has some issues to be fixed. On Thu, Jan 18, 2018 at 4:35 PM Neal Gompa <ngomp...@gmail.com> wrote: > On Thu, Jan 18, 2018 at 10:15 AM, Kevin Kofler <kevin.kof...@chello.at> > wrote: > > Pierre-Yves Chibon wrote: > >> Even if it does, it should have been advertised as a system-wide change > >> for F28 which is no longer possible. > > > > FESCo could approve an exception. > > > > If this really makes package updates made through PackageKit show up in > DNF > > history, IMHO, it would be worth considering at least. (But of course > not if > > it breaks things.) > > > > SWDB could be cherry-picked out into the current DNF stuff. There was > a point where it worked with the existing code before the rewrite of > libdnf in C++ started. > > The DNF team could also give a better idea of when the current rewrite > stuff will stabilize. > > > > -- > 真実はいつも一つ！/ Always, there's only one truth! >
_______________________________________________ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
