Yes.
Yes. The history of this function is that is was removed altogether and later re-added. In the last step, apparently, the return fell over while it shouldn't.
With v2.84, we had the in-progress status because new queries for a domain+type that were already forwarded were just added to a list of recipients what would receive the reply once upstream returned. This had the side-effect that you could loose DNS resolution is some edge-cases where upstream replies just never arrive. In this case, no new forwarding ever happened and nobody ever got a reply.
Hence, the proposed behavior with v2.85 was to always re-trigger a submission upstream, even when domain+type was already forwarded. This obsoleted the in-progress status and, hence, the function was removed from FTL. Later on, before v2.85 was finalized, discussions on the mailing list lead to a modification that was a compromise I proposed: Don't re-trigger a forwarding for known domain+type immediately, but only after a short amount of time (I wanted to have this configurable, Simon hard-coded it to 2 seconds). This effectively re-introduced in-progress and I reverted the removal of the function from FTL's code. Apparently, this return slipped through.
So @tomporter518's analysis was spot on.
Fortunately, this is entirely harmless because the omitted return here always triggers the next fail check so the worst thing happening here is that there are these two error messages in the log.
Thank you. There are only three small things:
- The PR should have a meaningful title (currently it is just
Development). Something likeRe-add missing returnseems appropriate. - There is a complaint by DCO bot. It gives you instructions how to fix it when you follow the link.
- We should squash-merge this PR to avoid adding the merge commit to our repo (you don't have to worry about this, we'll do it)