Comment # 5 on bug 1183168 from Reinhard Max

I finally got around to look into this again. The problem with %stop_on_removal
was that it stops AND disables the service, but I need to split these two
actions. Stop has to be done in the postgresqlXX-server packages containing the
actual binaries, but disable has to happen in the postgresql-server (noarch)
package that contains the service file.

I think I can just directly call "systemctl stop" instead of using a here,
because everything else that the macro does is already being handled by my own
scripting.

Any objections, Franck?

BTW, I am also using %_restart_on_update in the postgresqlXX-server packages,
which still exists, but probably also isn't meant to be used directly. Are
there plans to remove that one as well?