Description of problem: By making decisions based on FUD, you have introduced a severe regression in Fedora 19 that influences all packages that require MySQL compatibility. As an example look at your own Open Stack installer. It appears that the open stack installer no longer works because of the alias introduced for the package named "mysql-installer". Since the database you have attempted to replace mysql-server with has diverged, I can only imagine how annoying it is going to be to debug this longterm. I.E. This is just the icing on the cake. How reproducible: Start with your own open stack installer. Steps to Reproduce: 1. Download Fedora 19. 2. Use http://openstack.redhat.com/Quickstart 3. (repeat) Additional info: How about restoring the packaging to what users expect and request? If end users want to react on FUD, let them make that decision. If Redhat wants to play the FUD game, then don't screw the end user. Do it by introducing a QA/CI process where you relink and test everything. This slight of hand, "prey and hope it works", method of engineering doesn't make for a happy end user experience. Each database is its own special little thing,... you can't just slap a different one in and expect it to work.
Created attachment 768351 [details] Screenshot of your open stack installer blowing chunks.
I also see this bug...
Thank you for reporting. As for the problem in OpenStack, reading discussion on [1], it is caused by the fact that puppet can't handle situation where two packages provide the same symbol. In this particular situation puppet see mariadb-server and community-mysql-server, both providing mysql-server symbol. AFAICT, a quick fix should be (as proposed in [1]) to replace mysql-server by mariadb-server and mysql by mariadb for keys $client_package_name and $server_package_name in /usr/lib/python2.7/site-packages/packstack/puppet/modules/mysql/manifests/params.pp: -$client_package_name = 'mysql' -$server_package_name = 'mysql-server' +$client_package_name = 'mariadb' +$server_package_name = 'mariadb-server' So I believe this should actually be fixed in openstack-packstack package, thus re-assigning. However, this can happen any time in future again, since having the same symbol provided by more packages in one repository is not forbidden in Fedora. So IMHO puppet shouldn't panic in such situation, but should rather have some correctly defined algorithm to pick one. [1] http://openstack.redhat.com/forum/discussion/124/fedora-19-issues#Comment_783 As for decisions made by FUD -- there are different requirements that sound like a must for someone and like a madness for someone else. We often need to find a compromise. Brian, if you want to talk about decisions about including packages in Fedora, let's move that discussion to fedora devel-list, so other users/maintainers can be involved.
*** This bug has been marked as a duplicate of bug 981116 ***