Hello, it’s another weekly log.
I have been away since last Wednesday, so this log actually covers the week before last.
In the “recently completed” category, we have items in three areas:
GSoC:
- Haskell projects were accepted! Thus I will be mentoring one contributor who will expand my CI auto-recovery tool.
Stackage:
- I finished reviewing the logic of all the background processes.
- I opened a few issues/questions that came up during the review.
GHC CI:
- Got rid of the “Signal 9” spurious errors with Patch 12507, closing a long-standing issue that affected multiple repositories.
Coming up, I will continue working on a Stackage presentation for the Ecosystem Workshop (hence the recently-completed process review). I will be creating or updating documentation along the way.
GHC issues reviewed one week ago
(I was away for the issue triage that happened yesterday, so these are a week late.)
-
#24711: Re-think how to supply wired-in Names to GHC
- The rationale is good, but everyone agrees we need to keep thinking about it
-
#24712: GHCi segfault with breakpoint
- Already fixed
- !10448 was also a fix for the same thing? What happened
-
#24713: Template Haskell is unhygienic. Can we change it?
- How would we fix this without breaking TH? Has the boat simply sailed?
-
#24714: EPA: Span for GADT decl incorrect if sig only
- Part of his ongoing work
-
#24716: be a bit more helpful when it comes to distinguishing different type-level constructs
- A request for improving error messages
-
#24718: GHC panics when compiling code generated from Agda with profiling enabled
- Another excellent reproducer from Krzysztof
- The only question is whether to backport it to 9.8, since it looks difficult to do. The fix was part of a large patch doing other things.
-
#24719: Haddock should use lucid2 instead of xhtml
- Some pushback on changing out one boot package for another
- But moving to HTML5 is a solid argument
-
#24720: show COMPLETE pragmas in rendered Haddocks
- haddock ticket transferred from github
-
#24721: Inaccurate
since base-4.22.0.0comments on HpcFlags- fixed
-
#24722: Typecheck TypeAbstractions lambda expressions without explicit type ascriptions
- dependent-type-oriented shenanigans already being discussed by the experts there
-
#24723: Improve sharing in
ModIfaceafter generation- some performance improvement to ghc compile-time memory usage
-
#24724: Testsuite: The req_ghc_with_threaded_rts conditional does not work for the unregisterised backend.
- Basically a user question for the testsuite driver
- Cheng already supplied the answer and the fix
-
#24725: Allow higher-order patterns to match on constructors
- […skipped over this one pretty fast, I missed it]
-
#24726: -drule-check causes panic
- Simon is fixing it
-
#24727: Check maximum Sum<n># arity at renamer
- Is this a problem?
-
#24728: Generate a per-module static placeholder CCS for cafs instead of a global one.
- Sounds like a good improvement over the status quo
- Ben says ‘why not go further’ and add more cost centers… maybe -fprof-cafs by default
- (what the hell is a CAF)
-
#24729: Revisit the UI layout of instance definitions
- look like good ideas
- #24730: Simplifier turns value into ok-for-spec thunk
-
#24731: Build iserv (external interpreter) program on demand
- Seems like a pretty good idea
- Allows rebuilding iserv against a new template-haskell, allowing for reinstalling that library
- Provides flexibility for the combinatorial explosion of rts ways / cross-compilations that may exist for GHC in use
-
#24372: Annoying but possibly harmless errors in test log
- Excellent bug. Already fixed now
-
#24733: Use wasm clz/ctz/popcnt in wasm NCG
- Performance improvement for the wasm backend.
- Newcomer-friendly!
-
#24735: can't compile with llvm 18 on ghc 9.10.1-rc1 when invoked from shell in Aquamacs on Apple Silicon
- A ticket for tracking future llvm 18 support
- (it’s documented as not supported yet, as the issue author notes)
- Yet the error reported is odd and could use som efollowup
-
#24736: EPA: Comments missing for PrefixCon
- Another one on Alan’s exact print quest
-
#24737: Master branch pipelines are failing with "unknown unit: template-haskell"
- Now fixed