jaspervdj
We dropped out of Stackage nightly
@jaspervdj : It happened. We dropped out of Stackage nightly because of Diff:
- https://github.com/commercialhaskell/stackage/issues/7559
I am fixing this:
- https://hackage.haskell.org/package/goldplate-0.2.2.1/revisions/
- https://github.com/commercialhaskell/stackage/pull/7616
However, I think maintaining precise upper bounds for the dependencies isn't sustainable.
You have to eagerly revise bounds whenever a dependency releases a new major version. In practice, however, the probability of breakage is low, so there is a lot of busy work for nothing.
I favour the approach of not having upper bounds in the latest release of a package, and if breakage occurs, install a upper bound on all releases (that miss one) using tools like hackage-cli.
This way, one only does work when it is actually needed.
Why do I care about Stackage? With matrix.hackage gone, there is no QA for Hackage anymore. QA is now only provided by Stackage. I made it my policy to only depend on packages that are part of Stackage. I haven't seen an alternative to this yet.
Ok, next time there is a problem with an upper bound being too restrictive, we remove it!
Oops we did it again, because of optparse-applicative:
- #45
- https://github.com/commercialhaskell/stackage/pull/7804