On my F34 upgrade, I noticed an unexpected package, which I determined was due to Requires: bundled(...) that happened to be satisfied by something. This was of course an error, and should have been Provides: bundled(...).
Requires: bundled(...)
Provides: bundled(...)
AFAIK, there's not supposed to be any use for bundled Provides other than checks for security/bugfix propagation. Requiring one seems like it's always a bug. Can this be blocked somehow, with a brp, or gating test?
We would probably want to enable the RPM feature to run package checks post-build to trigger build failures for things that are absolutely out of policy. This is something that the Mandriva family has done for a long time with rpmlint, and we could do the same for a variety of things.
Sure, I guess implementation is of secondary concern here, and I probably should not have brought it up yet, because this is not explicitly forbidden, and the question is whether it should be?
It should be forbidden, yes. +1
This is already forbidden by policy, we just don't have a way to enforce it yet.
If rpm doesn't give is a way to hard-enforce this then I don't see why we wouldn't at least try to get a a gating test implemented. Of course, I have no idea at all how to actually do that.
It would also seem smart to have a rpmlint check implemented as well. That may well be easier, though I also don't know much about rpmlint development these days and whether this would be something added in the Fedora package or something sent upstream.
Login to comment on this ticket.