Hi,
On Mon, 1 Feb 2021 at 01:32, Jens-Ulrik Petersen petersen@redhat.com wrote:
I have thought for quite while it would be better for Haskell packages to opt out of Koschei, since my impression is that for nearly all our Haskell packages Koschei doesn't really give us anything but a lot of failures, so it seems to me just to be wasting a lot of Koji cycles for us.
The only way I can think of to do that is to opt out of release-monitoring... which generally makes sense for all our Haskell libraries at least anyway since we are tracking specific Stackage LTS versions anyway.
Since we appear to only update most things with Stackage, it makes sense to not have release monitoring enabled. I've largely ignored bug reports from it, since that requires a large rebuild of things even to do.
However, koschei and release-monitoring are unrelated. The only real relation is that if we actually act on release-monitoring bugs, then koschei runs might have some use. But along those lines, if we stop release-monitoring, then there's not so much point to running koschei for most packages. It might still be useful when the package depends on non-Haskell things though.
Any thoughts on this, not sure if anything else could be done to improve the situation. I decided to open the discussion first here but before taking it up with Devel list. I would rather save some Koji electricity.
Thanks, Jens