Regarding:
https://src.fedoraproject.org/rpms/gobject-introspection/pull-request/3 https://src.fedoraproject.org/rpms/dbus-glib/pull-request/1#
I'm in favor of tests. The thing is, isn't the right test for -devel packages whether *other existing packages* successfully build? For gobject-introspection, we could trigger a build of *tons* of existing packages, and IMO that's a far better test of whether things work than a completely artificial test case.
Also, why use "classic" for this instead of "container". Again in my opinion, *all* builds should be in a container. I don't have gcc on my host workstation: https://fedorapeople.org/~walters/2018.01-devconf-desktopcontainers/
But back to the bigger picture - I think this is a lack of integration between "CI" and "official builds". The *correct* test for whether a -devel package works is whether or not the other software we ship builds and works.
On Wed, Feb 07, 2018 at 02:18:15PM -0500, Colin Walters wrote:
But back to the bigger picture - I think this is a lack of integration between "CI" and "official builds". The *correct* test for whether a -devel package works is whether or not the other software we ship builds and works.
Yeah, I don't think it otherwise succeeds at point one of the Definition of CI in the Manefesto:
https://fedoraproject.org/wiki/CI/Manifesto