fix: address Copilot review comments on PR #118

- Clarify dependency explanation in GHC.Hs.Instances.Parsed comment
- Add DEPRECATED pragma to empty GHC.Hs.Instances.Transitions module
- Clarify toRegNo helper comment in Dwarf/Constants.hs as a fallback
- Log a message when initializePlugins is skipped (no interpreter)
- Fix wrong function name in panic: runHscPhase -> runHscBackendPhase

Author: angerman

compiler/GHC/CmmToAsm/Dwarf/Constants.hs

@@ -18,8 +18,10 @@ import GHC.CmmToAsm.X86.Regs
 
 import Data.Word
 
--- | Extract register number from a real register
--- This is a local helper to avoid depending on any specific NCG's Regs module
+-- | Extract register number from a real register.
+-- This is a local, generic fallback used when a platform-specific NCG Regs
+-- module is not available (e.g. when that NCG is compiled out). It is not
+-- intended to replace backend-specific definitions elsewhere.
 toRegNo :: Reg -> RegNo
 toRegNo (RegReal (RealRegSingle n)) = n
 toRegNo _                           = panic "toRegNo: unsupported register"

compiler/GHC/Driver/Main.hs

@@ -328,9 +328,16 @@ import qualified GHC.Unit.Home.Graph as HUG
 import GHC.Unit.Home.PackageTable
 
 #if !defined(HAVE_INTERPRETER)
--- Stub when interpreter not available
+-- Stub when interpreter not available: plugin initialization requires the
+-- interpreter to load compiled plugin code, so we skip it in minimal builds.
+-- If plugins are specified in the flags but the interpreter is unavailable,
+-- they will simply be ignored.
 initializePlugins :: HscEnv -> IO HscEnv
-initializePlugins = return
+initializePlugins hsc_env = do
+  let logger = hsc_logger hsc_env
+  logMsg logger MCInfo noSrcSpan $
+    text "initializePlugins: interpreter not available; skipping plugin initialization"
+  return hsc_env
 #endif
 
 #if !defined(HAVE_JS_BACKEND)

compiler/GHC/Driver/Pipeline/Execute.hs

@@ -636,7 +636,7 @@ runHscBackendPhase pipe_env hsc_env mod_name src_flavour location result = do
               bc <- generateFreshByteCode hsc_env mod_name (mkCgInteractiveGuts cgguts) mod_location
               return ([], final_iface, emptyHomeModInfoLinkable { homeMod_bytecode = Just bc } , panic "interpreter")
 #else
-            panic "GHC.Driver.Pipeline.Execute.runHscPhase: bytecode generation not supported (HAVE_INTERPRETER not defined)"
+            panic "GHC.Driver.Pipeline.Execute.runHscBackendPhase: bytecode generation not supported (HAVE_INTERPRETER not defined)"
 #endif
 
 

compiler/GHC/Hs/Instances/Parsed.hs

@@ -10,8 +10,10 @@
 -- NO transitive dependencies on HsExpr, HsBindLR, HsLocalBinds, HsUntypedSplice,
 -- or HsType (which contains HsSpliceTy -> HsUntypedSplice -> HsExpr).
 --
--- Most GhcPs instances are in Renamed.hs because of the hard-coded [LSig GhcRn]
--- in NHsValBindsLR which creates: HsExpr -> HsLocalBinds -> NHsValBindsLR -> [LSig GhcRn]
+-- Most GhcPs instances are in Renamed.hs because NHsValBindsLR (used by
+-- HsLocalBinds inside HsExpr) has a field of type [LSig GhcRn].  Any Data
+-- instance for NHsValBindsLR GhcPs would therefore mention GhcRn, so those
+-- instances must live in Renamed.hs to avoid cyclic module dependencies.
 --
 -- Additionally, HsType GhcPs contains HsSpliceTy which uses HsUntypedSplice GhcPs,
 -- and HsUntypedSplice contains HsExpr, so HsType and all types containing it must

compiler/GHC/Hs/Instances/Transitions.hs

@@ -16,9 +16,13 @@
 -- parameter which determines which phase-specific types they need.
 --
 -- This module is kept for backwards compatibility and re-exports nothing.
+-- It should not be imported in new code; it exists only so that existing
+-- imports do not break. Consider removing it once all downstream consumers
+-- have been updated.
 --
 -- See #9557 and #18254 for why we use -O0.
 {-# OPTIONS_GHC -O0 #-}
+{-# DEPRECATED "This module is empty; import the phase-specific GHC.Hs.Instances.* modules instead." #-}
 
 module GHC.Hs.Instances.Transitions where