Welcome to the next weekly log.
The main theme of this last week has been finalizing the Stackage handover. At this point I thnk I’m actually done with the technical work. All that remains is to double check everything and write the official announcement.
- I cleaned up the Stackage data migration — it’s done now!
- I set up a new Hackage mirror so FP Complete can retire theirs eventually. I think it’s ready for final testing.
- I am investigating a performance issue with the new Stackage data bucket. It only affects curators, not regular Stackage users. I’ve been in touch with the service provider and I trust they’ll be able to fix it pretty quickly.
In the next week, I want to officially wrap up the handover 
and start on some important followup work. Among the followup tasks, I want to highlight two documentation projects: I want to document the history of my handover effort and the Stackage architecture itself. I am now intimately familiar with Stackage, and I want to share how it works, where it could be improved, and how it affects the stability and reliability of the Haskell ecosystem. For instance, did you know it’s valuable to get your package included in Stackage, even if you prefer to use cabal-install and believe—as I do—that conservative upper bounds on your package dependencies are a good thing™?
That’s all for the log, see you next time.
This week's GHC issue triage notes
(I had another obligation and only stayed for fifteen minutes of the triage session. The GHC team actually triaged a much larger number of tickets than what I’ve listed here.)
-
#24594: Consider adding a CI configuration that builds stage2 with -O2 for AArch64.
- The suggested fix is to run the “release” flavour on the nightly run. But the nightly run is not used (hardly at all) so I don’t know if that is a good idea. (To be fair, the nighlty is getting healthier.)
-
#24593:
putEnuminGHC.StgToJS.Objectcould be simpler and more general (and more correct)- task
-
#24592: Off by one error in seekBinNoExpand
- high prio compiler-crashing bug
-
#24591: Arity analysis fails badly
- SPJ talked about this in the call
- Missed opportunity at runtime performance improvement
-
#24590: Out-of-range literal warnings interact poorly with rebindable syntax
- Yeah good to fix that. It’s a bit complicated because of the interactions between a bunch of different syntax-modifying language extensions…
-
#24589:
inlinemay only inline the wrapper and not the worker- “Makes complete sense, I don’t think there’s very much we can do about it”
-
inlinejust isn’t a very useful function, maybe.
-
#24588: ghc-toolchain and configure disagree on whether windres exists
- high prio bug in the new ghc-toolchain utility
(Maybe I’ll rename it since it’s misleading.) Stackage continues to live at