GitLab

Otherwise, the sub-projects have both hostlabels as constraints. Was not
enforced like this before 2021-10-29. But now something triggered this,
possibly it was always ment to be like this? Only thing that happened
today was an OOM initiated kill of bs_sched on obs-backend, that maybe
re-read information that was stale for a long time.

Oh well...

This makes it easier in the spec files, e.g. in conditionals

Also, colorful diffs

It was incorrectly set also set for the "System" repository there, which
caused software there to be built incorrectly (with the micro arch
flags) and in the wrong location

Repos were missing reference to the home: 'System' repo

I suspect the dependency solver copes better that way. Not sure though..

The previous fix only worked in the 'gcc_6' repo itself and not in
dependent repositories such as cuda_10_0_gcc_6.

Apparently something has changed upstream and openSUSE has not yet
tracked this in their prjconf, added here meanwhile