- ghc-commits - Haskell.org

[Git][ghc/ghc][wip/spj-try-opt-coercion] Add changelog
by Simon Peyton Jones (＠simonpj) 04 May '26

04 May '26

Simon Peyton Jones pushed to branch wip/spj-try-opt-coercion at Glasgow Haskell Compiler / GHC Commits: 6efc02e1 by Simon Peyton Jones at 2026-05-04T23:12:27+01:00 Add changelog - - - - - 1 changed file: - + changelog.d/opt-coercion Changes: ===================================== changelog.d/opt-coercion ===================================== @@ -0,0 +1,18 @@ +section: compiler +synopsis: Substantially reduce coercion optimisation +issues: #26679 +mrs: !15175 + +description: { +The coercion optimiser can gobble up a lot of compile time when there +are lots of coercions. Its goal is to shrink coercions in order to reduce +compile time, so it's bad that it sometimes increases it. + +This change introduces a simpler and faster coercion optimser, called +`optCoRefl`, which is used frequently, relegating the big-hammer coercion +optimiser to a separate pass of its own. + +A new flag `-fopt-refl-coercion` (on by default) controls the new `optCoRefl` +optimiser. +} + View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/6efc02e154b973d0aaab0748aff2c33… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/6efc02e154b973d0aaab0748aff2c33… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][wip/spj-try-opt-coercion] Fix warning
by Simon Peyton Jones (＠simonpj) 04 May '26

04 May '26

Simon Peyton Jones pushed to branch wip/spj-try-opt-coercion at Glasgow Haskell Compiler / GHC Commits: 6ed99365 by Simon Peyton Jones at 2026-05-04T23:04:17+01:00 Fix warning - - - - - 1 changed file: - compiler/GHC/Core/Coercion/Opt.hs Changes: ===================================== compiler/GHC/Core/Coercion/Opt.hs ===================================== @@ -354,10 +354,13 @@ opt_co_refl subst co -- See Note [Optimising TransCo sequences] go co@(TransCo {}) = gobble gs0 out_cos where - out_cos :: [OutCoercion] -- None of these are TransCos - out_cos@(out_co1:_) = get_in co [] - gs0 = GS (mkReflCo role lk) tm0 - tm0 = insertTM lk gs0 emptyTM + out_cos :: [OutCoercion] -- None of these are TransCos, and the list non-empty + out_cos = get_in co [] + out_co1 = case out_cos of + (co1:_) -> co1 + [] -> pprPanic "gobble" (ppr co) + gs0 = GS (mkReflCo role lk) tm0 + tm0 = insertTM lk gs0 emptyTM lk = coercionLKind out_co1 role = coercionRole out_co1 @@ -396,18 +399,17 @@ When composing co_i, if coercionRKind( co_i ) is the same as any earlier type in the chain, we want to discard it and simply revert to the GobbleState of that earlier point. -So a GobbleState has +A GobbleState gs_k consists of: * The coercion so far: a coercion equivalent to (co1;co2;cok) * A map from types to all earlier GobbleStates, thus tyR_i :-> gs_i - where (INVARIANT) tyR_i is the coercionRKind of gs_i + where (INVARIANT) tyR_i is the coercionRKind of co_i stored in gs_i Wrinkles: * Wart: the TypeMap is (currently) multiplicity-sensitive so we might miss a possible TransCo optimisation. - -} optUnivCo :: UnivCoProvenance -> [Coercion] View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/6ed9936560e28687abddbc292a26536… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/6ed9936560e28687abddbc292a26536… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][wip/spj-reinstallable-base2] 13 commits: unpackCString, unpackCStringUtf8; DROP MonadThings
by Rodrigo Mesquita (＠alt-romes) 04 May '26

04 May '26

Rodrigo Mesquita pushed to branch wip/spj-reinstallable-base2 at Glasgow Haskell Compiler / GHC Commits: aad3f72a by Rodrigo Mesquita at 2026-05-04T15:33:00+01:00 unpackCString, unpackCStringUtf8; DROP MonadThings lookupKnownKeyGlobal to `getMkStringIds` in `Driver.Config.Tidy` - - - - - a47d441c by Rodrigo Mesquita at 2026-05-04T15:54:32+01:00 krepStar, krepStarArrStar, etc... I think this one might have an issue. When compiling the GHC.Internal.Types module itself, for which we'll generate these bindings, we look the Ids up. But how will the lookup find anything? We haven't defined the name at all, it won't be in scope. Now that we don't use special names for the kreps (krepStar rather than krep$* or what was it), can we define them in GHC.Internal.Types directly and get rid of all this business ExportedKindRepsTodo? TODO!!!!!!! ROMES:TODO!!!! - - - - - c4afbe4b by Rodrigo Mesquita at 2026-05-04T16:30:35+01:00 Tweaks and note to ExactRdrName Co-authored-by: SPJ - - - - - d2d221fc by Rodrigo Mesquita at 2026-05-04T16:47:01+01:00 ic_monad; ic_int_print; ExactRdrName; rmv printName - - - - - 0e36b376 by Rodrigo Mesquita at 2026-05-04T17:01:37+01:00 ioTyConName - - - - - 888e0a18 by Rodrigo Mesquita at 2026-05-04T17:23:19+01:00 ioDataConName - - - - - fad5c84d by Rodrigo Mesquita at 2026-05-04T17:33:52+01:00 ipClassName - - - - - 002f334b by Rodrigo Mesquita at 2026-05-04T17:37:54+01:00 hasFieldClassName - - - - - 9fa810c0 by Rodrigo Mesquita at 2026-05-04T17:47:42+01:00 ghciIoClassName, ghciStepIoMName - - - - - cf7ed9ec by Rodrigo Mesquita at 2026-05-04T17:51:13+01:00 toAnnotationWrapperName - - - - - 819dc248 by Rodrigo Mesquita at 2026-05-04T17:53:16+01:00 specTyConName - - - - - 9a4f6638 by Rodrigo Mesquita at 2026-05-04T18:06:08+01:00 srcLocDataConName - - - - - c8f3006f by Rodrigo Mesquita at 2026-05-04T18:10:00+01:00 pushCallStackName - - - - - 38 changed files: - compiler/GHC.hs - compiler/GHC/Builtin/KnownKeys.hs - compiler/GHC/Builtin/KnownOccs.hs - compiler/GHC/Core/Make.hs - compiler/GHC/Driver/Config/Tidy.hs - compiler/GHC/Driver/Env.hs - compiler/GHC/HsToCore/Foreign/Wasm.hs - compiler/GHC/HsToCore/Match/Literal.hs - compiler/GHC/HsToCore/Monad.hs - compiler/GHC/HsToCore/Quote.hs - compiler/GHC/HsToCore/Utils.hs - compiler/GHC/Iface/Load.hs - compiler/GHC/IfaceToCore.hs - compiler/GHC/Plugins.hs - compiler/GHC/Runtime/Context.hs - compiler/GHC/StgToCmm/Bind.hs - compiler/GHC/StgToJS/Apply.hs - compiler/GHC/StgToJS/Expr.hs - compiler/GHC/Tc/Errors.hs - compiler/GHC/Tc/Errors/Ppr.hs - compiler/GHC/Tc/Gen/Bind.hs - compiler/GHC/Tc/Gen/Expr.hs - compiler/GHC/Tc/Gen/HsType.hs - compiler/GHC/Tc/Gen/Splice.hs - compiler/GHC/Tc/Instance/Class.hs - compiler/GHC/Tc/Instance/Typeable.hs - compiler/GHC/Tc/Module.hs - compiler/GHC/Tc/Solver/Default.hs - compiler/GHC/Tc/Solver/Dict.hs - compiler/GHC/Tc/Solver/Monad.hs - compiler/GHC/Tc/Utils/Env.hs - compiler/GHC/Tc/Utils/Monad.hs - compiler/GHC/Types/Name/Reader.hs - compiler/GHC/Types/TyThing.hs - ghc/GHCi/UI.hs - libraries/base/src/GHC/Essentials.hs - libraries/ghc-internal/src/GHC/Internal/Classes/IP.hs - libraries/ghc-internal/src/GHC/Internal/Types.hs The diff was not included because it is too large. View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/29aaf0863f90ad7a61946f16af0697… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/29aaf0863f90ad7a61946f16af0697… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][wip/supersven/libDir-setting] WIP: relative LibDir
by Sven Tennie (＠supersven) 04 May '26

04 May '26

Sven Tennie pushed to branch wip/supersven/libDir-setting at Glasgow Haskell Compiler / GHC Commits: b9e16369 by Sven Tennie at 2026-05-04T19:10:24+02:00 WIP: relative LibDir - - - - - 1 changed file: - hadrian/src/Rules/Generate.hs Changes: ===================================== hadrian/src/Rules/Generate.hs ===================================== @@ -25,7 +25,6 @@ import Utilities import GHC.Toolchain as Toolchain hiding (HsCpp(HsCpp)) import GHC.Platform.ArchOS import Settings.Program (ghcWithInterpreter) -import Hadrian.Oracles.Path -- | Track this file to rebuild generated files whenever it changes. trackGenerateHs :: Expr () @@ -488,16 +487,11 @@ generateSettings settingsFile includeLibDir package_db_path = do Stage3 -> pkgUnitId Stage2 base let rel_pkg_db = makeRelativeNoSysLink (dropFileName settingsFile) package_db_path - make_absolute rel_path = do - abs_path <- liftIO (makeAbsolute rel_path) - fixAbsolutePathOnWindows abs_path - -- E.g. the Stage2 compiler lives in _build/stage1 -- So, we need to decrement the stage to get the correct directory stage_dir_stage = predStage stage rel_lib_topDir :: FilePath <- expr $ stageLibPath stage_dir_stage - lib_topDir :: FilePath <- expr $ make_absolute rel_lib_topDir settings <- traverse sequence $ [ ("unlit command", ("$topdir/../bin/" <>) <$> expr (programName (ctx { Context.package = unlit }))) @@ -506,7 +500,7 @@ generateSettings settingsFile includeLibDir package_db_path = do , ("Relative Global Package DB", pure rel_pkg_db) , ("base unit-id", pure base_unit_id) ] - ++ ([("LibDir", pure lib_topDir) | includeLibDir]) + ++ ([("LibDir", pure rel_lib_topDir) | includeLibDir]) let showTuple (k, v) = "(" ++ show k ++ ", " ++ show v ++ ")" pure $ case settings of View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/b9e16369e07233e103b754c8b6b4053… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/b9e16369e07233e103b754c8b6b4053… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][wip/dcoutts/windows-rts-dll] 5 commits: Add minimal dlltool support into ./configure
by David Eichmann (＠DavidEichmann) 04 May '26

04 May '26

David Eichmann pushed to branch wip/dcoutts/windows-rts-dll at Glasgow Haskell Compiler / GHC Commits: 5bae0d47 by Duncan Coutts at 2026-05-04T17:50:31+01:00 Add minimal dlltool support into ./configure Find dlltool, and hopefully support finding it within the bundled llvm toolchain on windows. - - - - - 7708588f by Duncan Coutts at 2026-05-04T17:50:31+01:00 Update the default host and target files for dlltool support - - - - - 015417c8 by Duncan Coutts at 2026-05-04T17:50:31+01:00 Add dlltool as a hadrian builder Optional except on windows. - - - - - b15d6d0c by Duncan Coutts at 2026-05-04T17:50:31+01:00 Update and generate libHSghc-internal.def from .def.in file The only symbol that the rts imports from the ghc-internal package now is init_ghc_hs_iface. So the rts only needs an import lib that defines that one symbol. Also, remove the libHSghc-prim.def because it is redundant. The rts no longer imports anything from ghc-prim. Keep libHSffi.def for now. We may yet need it once it is clear how libffi is going to be built/used for ghc. - - - - - d75a3712 by Duncan Coutts at 2026-05-04T17:50:31+01:00 Add rule to build libHSghc-internal.dll.a and link into the rts On windows only, with dynamic linking. This is needed because on windows, all symbols in dlls must be resolved. No dangling symbols allowed. References to external symbols must be explicit. We resolve this with an import library. We create an import library for ghc-internal, a .dll.a file. This is a static archive containing .o files that define the symbols we need, and crucially have ".idata" sections that specifies the symbols the dll imports and from where. Note that we do not install this libHSghc-internal.dll.a, and it does not need to list all the symbols exported by that package. We create a special purpose import lib and only use it when linking the rts dll, so it only has to list the symbols that the rts uses from ghc-internal (which is exactly one symbol: init_ghc_hs_iface). - - - - - 14 changed files: - .gitlab/generate-ci/gen_ci.hs - .gitlab/jobs.yaml - configure.ac - hadrian/cfg/default.host.target.in - hadrian/cfg/default.target.in - hadrian/src/Builder.hs - hadrian/src/Rules/Generate.hs - hadrian/src/Rules/Library.hs - hadrian/src/Rules/Rts.hs - m4/fp_setup_windows_toolchain.m4 - m4/ghc_toolchain.m4 - m4/prep_target_file.m4 - rts/.gitignore - + rts/win32/libHSghc-internal.def.in Changes: ===================================== .gitlab/generate-ci/gen_ci.hs ===================================== @@ -1305,6 +1305,7 @@ cross_jobs = [ . setVariable "STRINGS" (llvm_prefix ++ "strings") . setVariable "STRIP" (llvm_prefix ++ "strip") . setVariable "WindresCmd" (llvm_prefix ++ "windres") + . setVariable "DlltoolCmd" (llvm_prefix ++ "llvm-dlltool") . setVariable "LLVMAS" (llvm_prefix ++ "clang") . setVariable "LD" (llvm_prefix ++ "ld") -- Windows target require to make linker merge feature check disabled. ===================================== .gitlab/jobs.yaml ===================================== @@ -317,6 +317,7 @@ "CROSS_EMULATOR": "/opt/wine-arm64ec-msys2-deb12/bin/wine", "CROSS_TARGET": "aarch64-unknown-mingw32", "CXX": "/opt/llvm-mingw-linux/bin/aarch64-w64-mingw32-clang++", + "DlltoolCmd": "/opt/llvm-mingw-linux/bin/aarch64-w64-mingw32-llvm-dlltool", "HADRIAN_ARGS": "--docs=none", "INSTALL_CONFIGURE_ARGS": "--enable-strict-ghc-toolchain-check", "LD": "/opt/llvm-mingw-linux/bin/aarch64-w64-mingw32-ld", @@ -399,6 +400,7 @@ "CROSS_EMULATOR": "/opt/wine-arm64ec-msys2-deb12/bin/wine", "CROSS_TARGET": "aarch64-unknown-mingw32", "CXX": "/opt/llvm-mingw-linux/bin/aarch64-w64-mingw32-clang++", + "DlltoolCmd": "/opt/llvm-mingw-linux/bin/aarch64-w64-mingw32-llvm-dlltool", "HADRIAN_ARGS": "--docs=none", "INSTALL_CONFIGURE_ARGS": "--enable-strict-ghc-toolchain-check", "LD": "/opt/llvm-mingw-linux/bin/aarch64-w64-mingw32-ld", @@ -1111,6 +1113,7 @@ "CROSS_EMULATOR": "/opt/wine-arm64ec-msys2-deb12/bin/wine", "CROSS_TARGET": "aarch64-unknown-mingw32", "CXX": "/opt/llvm-mingw-linux/bin/aarch64-w64-mingw32-clang++", + "DlltoolCmd": "/opt/llvm-mingw-linux/bin/aarch64-w64-mingw32-llvm-dlltool", "HADRIAN_ARGS": "--docs=none", "INSTALL_CONFIGURE_ARGS": "--enable-strict-ghc-toolchain-check", "LD": "/opt/llvm-mingw-linux/bin/aarch64-w64-mingw32-ld", @@ -1194,6 +1197,7 @@ "CROSS_EMULATOR": "/opt/wine-arm64ec-msys2-deb12/bin/wine", "CROSS_TARGET": "aarch64-unknown-mingw32", "CXX": "/opt/llvm-mingw-linux/bin/aarch64-w64-mingw32-clang++", + "DlltoolCmd": "/opt/llvm-mingw-linux/bin/aarch64-w64-mingw32-llvm-dlltool", "HADRIAN_ARGS": "--docs=none", "INSTALL_CONFIGURE_ARGS": "--enable-strict-ghc-toolchain-check", "LD": "/opt/llvm-mingw-linux/bin/aarch64-w64-mingw32-ld", ===================================== configure.ac ===================================== @@ -314,13 +314,18 @@ else AC_CHECK_TOOL([RANLIB],[ranlib]) AC_CHECK_TOOL([OBJDUMP],[objdump]) AC_CHECK_TOOL([WindresCmd],[windres]) + AC_CHECK_TOOL([DlltoolCmd],[dlltool]) + AC_CHECK_TOOL([DlltoolCmd],[llvm-dlltool]) AC_CHECK_TOOL([Genlib],[genlib]) if test "$HostOS" = "mingw32"; then AC_CHECK_TARGET_TOOL([WindresCmd],[windres]) + AC_CHECK_TARGET_TOOL([DlltoolCmd],[dlltool]) + AC_CHECK_TARGET_TOOL([DlltoolCmd],[llvm-dlltool]) AC_CHECK_TARGET_TOOL([OBJDUMP],[objdump]) WindresCmd="$(cygpath -m $WindresCmd)" + DlltoolCmd="$(cygpath -m $DlltoolCmd)" if test "$Genlib" != ""; then GenlibCmd="$(cygpath -m $Genlib)" @@ -1062,6 +1067,7 @@ echo "\ otool : $OtoolCmd install_name_tool : $InstallNameToolCmd windres : $WindresCmd + dlltool : $DlltoolCmd genlib : $GenlibCmd Happy : $HappyCmd ($HappyVersion) Alex : $AlexCmd ($AlexVersion) ===================================== hadrian/cfg/default.host.target.in ===================================== @@ -45,6 +45,7 @@ Target , tgtOpt = Nothing , tgtLlvmAs = Nothing , tgtWindres = Nothing +, tgtDlltool = Nothing , tgtOtool = Nothing , tgtInstallNameTool = Nothing } ===================================== hadrian/cfg/default.target.in ===================================== @@ -45,6 +45,7 @@ Target , tgtOpt = @OptCmdMaybeProg@ , tgtLlvmAs = @LlvmAsCmdMaybeProg@ , tgtWindres = @WindresCmdMaybeProg@ +, tgtDlltool = @DlltoolCmdMaybeProg@ , tgtOtool = @OtoolCmdMaybeProg@ , tgtInstallNameTool = @InstallNameToolCmdMaybeProg@ } ===================================== hadrian/src/Builder.hs ===================================== @@ -17,7 +17,7 @@ import Development.Shake.Classes import Development.Shake.Command import Development.Shake.FilePath import GHC.Generics -import GHC.Platform.ArchOS (ArchOS(..), Arch(..)) +import GHC.Platform.ArchOS (ArchOS(..), Arch(..), OS(..)) import qualified Hadrian.Builder as H import Hadrian.Builder hiding (Builder) import Hadrian.Builder.Ar @@ -180,6 +180,7 @@ data Builder = Alex | Objdump | Python | Ranlib + | Dlltool | Testsuite TestMode | Sphinx SphinxMode | Tar TarMode @@ -418,6 +419,7 @@ isOptional target = \case Alex -> True -- Most ar implemententions no longer need ranlib, but some still do Ranlib -> not $ Toolchain.arNeedsRanlib (tgtAr target) + Dlltool -> archOS_OS (tgtArchOs target) /= OSMinGW32 JsCpp -> not $ (archOS_arch . tgtArchOs) target == ArchJavaScript -- ArchWasm32 too? _ -> False @@ -442,6 +444,7 @@ systemBuilderPath builder = case builder of Objdump -> fromKey "objdump" Python -> fromKey "python" Ranlib -> fromTargetTC "ranlib" (maybeProg Toolchain.ranlibProgram . tgtRanlib) + Dlltool -> fromTargetTC "dlltool" (maybeProg id . tgtDlltool) Testsuite _ -> fromKey "python" Sphinx _ -> fromKey "sphinx-build" Tar _ -> fromKey "tar" ===================================== hadrian/src/Rules/Generate.hs ===================================== @@ -377,6 +377,7 @@ templateRules = do , interpolateSetting "ProjectPatchLevel1" ProjectPatchLevel1 , interpolateSetting "ProjectPatchLevel2" ProjectPatchLevel2 ] + templateRule "rts/win32/libHSghc-internal.def" projectVersion templateRule "docs/index.html" $ packageUnitIds Stage1 templateRule "docs/users_guide/ghc_config.py" $ mconcat [ projectVersion ===================================== hadrian/src/Rules/Library.hs ===================================== @@ -4,6 +4,8 @@ import Hadrian.BuildPath import Hadrian.Haskell.Cabal import Hadrian.Haskell.Cabal.Type import qualified Text.Parsec as Parsec +import GHC.Platform.ArchOS (ArchOS(archOS_OS), OS(..)) +import GHC.Toolchain.Target (Target(tgtArchOs)) import Base import Context @@ -185,9 +187,13 @@ jsObjects context = do srcs <- interpretInContext context (getContextData jsSrcs) mapM (objectPath context) srcs --- | Return extra object files needed to build the given library context. The --- resulting list is currently non-empty only when the package from the --- 'Context' is @ghc-internal@ built with in-tree GMP backend. +-- | Return extra object files needed to build the given library context. +-- +-- This is non-empty for: +-- +-- * @ghc-internal@ when built with in-tree GMP backend +-- * @rts@ on Windows when linking dynamically +-- extraObjects :: Context -> Action [FilePath] extraObjects context | package context == ghcInternal = do @@ -195,6 +201,13 @@ extraObjects context "gmp" -> gmpObjects (stage context) _ -> return [] + | package context == rts = do + target <- interpretInContext context getStagedTarget + builddir <- buildPath context + return [ builddir -/- "libHSghc-internal.dll.a" + | archOS_OS (tgtArchOs target) == OSMinGW32 + , Dynamic `wayUnit` way context ] + | otherwise = return [] -- | Return all the object files to be put into the library we're building for ===================================== hadrian/src/Rules/Rts.hs ===================================== @@ -24,6 +24,20 @@ rtsRules = priority 3 $ do (addRtsDummyVersion $ takeFileName rtsLibFilePath') rtsLibFilePath' + -- Solve the recursive dependency between the rts and ghc-internal + -- on Windows by creating an import lib for the ghc-internal dll, + -- to be linked into the rts dll. + forM_ [Stage1, Stage2, Stage3 ] $ \ stage -> do + let buildPath = root -/- buildDir (rtsContext stage) + buildPath -/- "libHSghc-internal.dll.a" %> buildGhcInternalImportLib + +buildGhcInternalImportLib :: FilePath -> Action () +buildGhcInternalImportLib target = do + let input = "rts/win32/libHSghc-internal.def" + output = target -- the .dll.a import lib + need [input] + runBuilder Dlltool ["-d", input, "-l", output] [input] [output] + -- Need symlinks generated by rtsRules. needRtsSymLinks :: Stage -> Set.Set Way -> Action () needRtsSymLinks stage rtsWays ===================================== m4/fp_setup_windows_toolchain.m4 ===================================== @@ -131,8 +131,8 @@ AC_DEFUN([FP_SETUP_WINDOWS_TOOLCHAIN],[ AR="${mingwbin}llvm-ar.exe" RANLIB="${mingwbin}llvm-ranlib.exe" OBJDUMP="${mingwbin}llvm-objdump.exe" - DLLTOOL="${mingwbin}llvm-dlltool.exe" WindresCmd="${mingwbin}llvm-windres.exe" + DlltoolCmd="${mingwbin}llvm-dlltool.exe" LLC="${mingwbin}llc.exe" OPT="${mingwbin}opt.exe" LLVMAS="${mingwbin}clang.exe" ===================================== m4/ghc_toolchain.m4 ===================================== @@ -95,6 +95,7 @@ AC_DEFUN([FIND_GHC_TOOLCHAIN], echo "--merge-objs=$MergeObjsCmd" >> acargs echo "--readelf=$READELF" >> acargs echo "--windres=$WindresCmd" >> acargs + echo "--dlltool=$DlltoolCmd" >> acargs echo "--llc=$LlcCmd" >> acargs echo "--opt=$OptCmd" >> acargs echo "--llvm-as=$LlvmAsCmd" >> acargs ===================================== m4/prep_target_file.m4 ===================================== @@ -191,6 +191,7 @@ AC_DEFUN([PREP_TARGET_FILE],[ PREP_MAYBE_SIMPLE_PROGRAM([OptCmd]) PREP_MAYBE_PROGRAM([LlvmAsCmd], [LlvmAsFlags]) PREP_MAYBE_SIMPLE_PROGRAM([WindresCmd]) + PREP_MAYBE_SIMPLE_PROGRAM([DlltoolCmd]) PREP_MAYBE_SIMPLE_PROGRAM([OtoolCmd]) PREP_MAYBE_SIMPLE_PROGRAM([InstallNameToolCmd]) PREP_MAYBE_STRING([TargetVendor_CPP]) ===================================== rts/.gitignore ===================================== @@ -20,3 +20,4 @@ /ghcautoconf.h.autoconf.in /ghcautoconf.h.autoconf /include/ghcautoconf.h +/win32/libHSghc-internal.def ===================================== rts/win32/libHSghc-internal.def.in ===================================== @@ -0,0 +1,4 @@ +LIBRARY libHSghc-internal-@ProjectVersionForLib@.0-ghc@ProjectVersion@.dll + +EXPORTS + init_ghc_hs_iface View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/06606b8eea8aee5697f2607d78562e… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/06606b8eea8aee5697f2607d78562e… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][wip/spj-try-opt-coercion] Improve the gobbler
by Simon Peyton Jones (＠simonpj) 04 May '26

04 May '26

Simon Peyton Jones pushed to branch wip/spj-try-opt-coercion at Glasgow Haskell Compiler / GHC Commits: ce739998 by Simon Peyton Jones at 2026-05-04T17:39:49+01:00 Improve the gobbler - - - - - 1 changed file: - compiler/GHC/Core/Coercion/Opt.hs Changes: ===================================== compiler/GHC/Core/Coercion/Opt.hs ===================================== @@ -326,7 +326,7 @@ opt_co_refl subst co go_hole h@(CH { ch_co_var = cv }) = h { ch_co_var = updateVarType go_ty cv } - go (Refl ty) = Refl (substTy subst ty) + go (Refl ty) = Refl $!! substTy subst ty go (GRefl r ty mco) = GRefl r $!! go_ty ty $!! go_m mco go (CoVarCo cv) = substCoVar subst cv go (HoleCo h) = HoleCo $!! go_hole h @@ -351,40 +351,64 @@ opt_co_refl subst co -- The TransCo case fires up the main loop for -- eliminating reflexive chains of TransCo - go (TransCo co1 co2) - | lk' `eqTypeIgnoringMultiplicity` rk' = go co2 - | otherwise = gobble0 gs1 [co2] - where - co1' = go co1 - Pair lk' rk' = coercionKind co1' - role = coercionRole co1 - - gs0, gs1 :: GobbleState - gs0 = GS (mkReflCo role lk') tm0 - gs1 = GS co1' (insertTM rk' gs1 tm0) - - tm0 = insertTM lk' gs0 emptyTM - - gobble0 :: GobbleState -> [InCoercion] -> OutCoercion - gobble0 (GS co _) [] = co - gobble0 gs (co:cos) = gobble gs co cos - - gobble :: GobbleState -> InCoercion -> [InCoercion] -> OutCoercion - -- gobble (GS co1 tm) co2 cos returns a coercion equivalent to (co1;co2;cos) - gobble gs (TransCo co2 co3) cos - = gobble gs co2 (co3 : cos) - gobble (GS co1' tm) co2 cos - = case lookupTM rk' tm of - Just gs -> gobble0 gs cos - Nothing -> gobble0 gs' cos - where - co2' = go co2 - rk' = coercionRKind co2' - gs' = GS (co1' `mkTransCo` co2') (insertTM rk' gs' tm) - - -data GobbleState = GS OutCoercion (TypeMap GobbleState) - -- The map is keyed by OutType + -- See Note [Optimising TransCo sequences] + go co@(TransCo {}) = gobble gs0 out_cos + where + out_cos :: [OutCoercion] -- None of these are TransCos + out_cos@(out_co1:_) = get_in co [] + gs0 = GS (mkReflCo role lk) tm0 + tm0 = insertTM lk gs0 emptyTM + lk = coercionLKind out_co1 + role = coercionRole out_co1 + + get_in :: InCoercion -> [OutCoercion] -> [OutCoercion] + get_in (TransCo co1 co2) cos = get_in co1 (get_in co2 cos) + get_in (SymCo (TransCo co1 co2)) cos = get_in (mkSymCo co2) (get_in (mkSymCo co1) cos) + get_in co cos = get_out (go co) cos + + get_out :: OutCoercion -> [OutCoercion] -> [OutCoercion] + -- Maybe `go` returned an OutCoercion that is a `TransCo` + get_out (TransCo co1 co2) cos = get_out co1 (get_out co2 cos) + get_out (SymCo (TransCo co1 co2)) cos = get_out (mkSymCo co2) (get_out (mkSymCo co1) cos) + get_out co cos = co : cos + + gobble :: GobbleState -> [OutCoercion] -> OutCoercion + gobble (GS co0 tm) cos + = case cos of + [] -> co0 + (co1:cos) -> case lookupTM rk1 tm of + Just gs -> gobble gs cos -- A hit in the GobbleState map; + -- revert to the earlier state gs + Nothing -> gobble gs' cos -- Miss: make a new GobbleState gs' + where + rk1 = coercionRKind co1 + gs' = GS (co0 `mkTransCo` co1) (insertTM rk1 gs' tm) + +data GobbleState = GS OutCoercion !(TypeMap GobbleState) + +{- Note [Optimising TransCo sequences] +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +In a composition (co1; co2; co3; ... ; con) we work left-to-right, maintaining +a `GobbleState`. Imagine we start with gs0::GobbleState, we process `co1` to get +`gs1` and so on. + +When composing co_i, if coercionRKind( co_i ) is the same as any earlier type in +the chain, we want to discard it and simply revert to the GobbleState of that +earlier point. + +So a GobbleState has + +* The coercion so far: a coercion equivalent to (co1;co2;cok) +* A map from types to all earlier GobbleStates, thus + tyR_i :-> gs_i + where (INVARIANT) tyR_i is the coercionRKind of gs_i + +Wrinkles: + +* Wart: the TypeMap is (currently) multiplicity-sensitive so we might + miss a possible TransCo optimisation. + +-} optUnivCo :: UnivCoProvenance -> [Coercion] -> Role -> Type -> Type -> Coercion View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/ce7399982c7a57a4d4b9eb41c32a5b9… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/ce7399982c7a57a4d4b9eb41c32a5b9… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc][wip/supersven/libDir-setting] WIP: relative LibDir
by Sven Tennie (＠supersven) 04 May '26

04 May '26

Sven Tennie pushed to branch wip/supersven/libDir-setting at Glasgow Haskell Compiler / GHC Commits: 48964730 by Sven Tennie at 2026-05-04T18:32:30+02:00 WIP: relative LibDir - - - - - 1 changed file: - hadrian/src/Rules/Generate.hs Changes: ===================================== hadrian/src/Rules/Generate.hs ===================================== @@ -488,16 +488,11 @@ generateSettings settingsFile includeLibDir package_db_path = do Stage3 -> pkgUnitId Stage2 base let rel_pkg_db = makeRelativeNoSysLink (dropFileName settingsFile) package_db_path - make_absolute rel_path = do - abs_path <- liftIO (makeAbsolute rel_path) - fixAbsolutePathOnWindows abs_path - -- E.g. the Stage2 compiler lives in _build/stage1 -- So, we need to decrement the stage to get the correct directory stage_dir_stage = predStage stage rel_lib_topDir :: FilePath <- expr $ stageLibPath stage_dir_stage - lib_topDir :: FilePath <- expr $ make_absolute rel_lib_topDir settings <- traverse sequence $ [ ("unlit command", ("$topdir/../bin/" <>) <$> expr (programName (ctx { Context.package = unlit }))) @@ -506,7 +501,7 @@ generateSettings settingsFile includeLibDir package_db_path = do , ("Relative Global Package DB", pure rel_pkg_db) , ("base unit-id", pure base_unit_id) ] - ++ ([("LibDir", pure lib_topDir) | includeLibDir]) + ++ ([("LibDir", pure rel_lib_topDir) | includeLibDir]) let showTuple (k, v) = "(" ++ show k ++ ", " ++ show v ++ ")" pure $ case settings of View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/489647303f40b3910b94071e957ece2… -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/489647303f40b3910b94071e957ece2… You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc] Deleted branch wip/gbc-faster-internal-names
by Cheng Shao (＠TerrorJack) 04 May '26

04 May '26

Cheng Shao deleted branch wip/gbc-faster-internal-names at Glasgow Haskell Compiler / GHC -- You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc] Deleted branch wip/use-less-appendfs
by Cheng Shao (＠TerrorJack) 04 May '26

04 May '26

Cheng Shao deleted branch wip/use-less-appendfs at Glasgow Haskell Compiler / GHC -- You're receiving this email because of your account on gitlab.haskell.org.

1 0

[Git][ghc/ghc] Deleted branch wip/ghci-messages-no-string
by Cheng Shao (＠TerrorJack) 04 May '26

04 May '26

Cheng Shao deleted branch wip/ghci-messages-no-string at Glasgow Haskell Compiler / GHC -- You're receiving this email because of your account on gitlab.haskell.org.

1 0