Marge Bot pushed to branch wip/marge_bot_batch_merge_job at Glasgow Haskell Compiler / GHC

Commits:

10 changed files:

Changes:

  • compiler/GHC/Driver/Downsweep.hs
    ... ... @@ -947,6 +947,71 @@ enableCodeGenWhen logger tmpfs staticLife dynLife unit_env mod_graph = do
    947 947 
                                      hostFullWays
    948 948 
             in dflags_c
    949 949
    950 
    +{- Note [-fno-code mode]
    951 
    +~~~~~~~~~~~~~~~~~~~~~~~~
    952 
    +GHC offers the flag -fno-code for the purpose of parsing and typechecking a
    953 
    +program without generating object files. This is intended to be used by tooling
    954 
    +and IDEs to provide quick feedback on any parser or type errors as cheaply as
    955 
    +possible.
    956 
    +
    957 
    +When GHC is invoked with -fno-code, no object files or linked output will be
    958 
    +generated. As many errors and warnings as possible will be generated, as if
    959 
    +-fno-code had not been passed. The session DynFlags will have
    960 
    +backend == NoBackend.
    961 
    +
    962 
    +-fwrite-interface
    963 
    +~~~~~~~~~~~~~~~~
    964 
    +Whether interface files are generated in -fno-code mode is controlled by the
    965 
    +-fwrite-interface flag. The -fwrite-interface flag is a no-op if -fno-code is
    966 
    +not also passed. Recompilation avoidance requires interface files, so passing
    967 
    +-fno-code without -fwrite-interface should be avoided. If -fno-code were
    968 
    +re-implemented today, there would be no need for -fwrite-interface as it
    969 
    +would considered always on; this behaviour is as it is for backwards compatibility.
    970 
    +
    971 
    +================================================================
    972 
    +IN SUMMARY: ALWAYS PASS -fno-code AND -fwrite-interface TOGETHER
    973 
    +================================================================
    974 
    +
    975 
    +Template Haskell
    976 
    +~~~~~~~~~~~~~~~~
    977 
    +A module using Template Haskell may invoke an imported function from inside a
    978 
    +splice. This will cause the type-checker to attempt to execute that code, which
    979 
    +would fail if no object files had been generated. See #8025. To rectify this,
    980 
    +during the downsweep we patch the DynFlags in the ModSummary of any home module
    981 
    +that is imported by a module that uses Template Haskell to generate object
    982 
    +code.
    983 
    +
    984 
    +The flavour of the generated code depends on whether `-fprefer-byte-code` is enabled
    985 
    +or not in the module which needs the code generation. If the module requires byte-code then
    986 
    +dependencies will generate byte-code, otherwise they will generate object files.
    987 
    +In the case where some modules require byte-code and some object files, both are
    988 
    +generated by enabling `-fbyte-code-and-object-code`, the test "fat015" tests these
    989 
    +configurations.
    990 
    +
    991 
    +The object files (and interface files if -fwrite-interface is disabled) produced
    992 
    +for Template Haskell are written to temporary files.
    993 
    +
    994 
    +Note that since Template Haskell can run arbitrary IO actions, -fno-code mode
    995 
    +is no more secure than running without it.
    996 
    +
    997 
    +Potential TODOS:
    998 
    +~~~~~
    999 
    +* Remove -fwrite-interface and have interface files always written in -fno-code
    1000 
    +  mode
    1001 
    +* Both .o and .dyn_o files are generated for template haskell, but we only need
    1002 
    +  .dyn_o. Fix it.
    1003 
    +* In make mode, a message like
    1004 
    +  Compiling A (A.hs, /tmp/ghc_123.o)
    1005 
    +  is shown if downsweep enabled object code generation for A. Perhaps we should
    1006 
    +  show "nothing" or "temporary object file" instead. Note that one
    1007 
    +  can currently use -keep-tmp-files and inspect the generated file with the
    1008 
    +  current behaviour.
    1009 
    +* Offer a -no-codedir command line option, and write what were temporary
    1010 
    +  object files there. This would speed up recompilation.
    1011 
    +* Use existing object files (if they are up to date) instead of always
    1012 
    +  generating temporary ones.
    1013 
    +-}
    1014 
    +
    950 1015 
     -- | Populate the Downsweep cache with the root modules.
    951 1016 
     mkRootMap
    952 1017 
       :: [ModuleNodeInfo]

  • compiler/GHC/Driver/Make.hs
    ... ... @@ -1246,70 +1246,6 @@ addSptEntries hsc_env mlinkable =
    1246 1246 
          , spt <- bc_spt_entries bco
    1247 1247 
          ]
    1248 1248
    1249 
    -{- Note [-fno-code mode]
    1250 
    -~~~~~~~~~~~~~~~~~~~~~~~~
    1251 
    -GHC offers the flag -fno-code for the purpose of parsing and typechecking a
    1252 
    -program without generating object files. This is intended to be used by tooling
    1253 
    -and IDEs to provide quick feedback on any parser or type errors as cheaply as
    1254 
    -possible.
    1255 
    -
    1256 
    -When GHC is invoked with -fno-code no object files or linked output will be
    1257 
    -generated. As many errors and warnings as possible will be generated, as if
    1258 
    --fno-code had not been passed. The session DynFlags will have
    1259 
    -backend == NoBackend.
    1260 
    -
    1261 
    --fwrite-interface
    1262 
    -~~~~~~~~~~~~~~~~
    1263 
    -Whether interface files are generated in -fno-code mode is controlled by the
    1264 
    --fwrite-interface flag. The -fwrite-interface flag is a no-op if -fno-code is
    1265 
    -not also passed. Recompilation avoidance requires interface files, so passing
    1266 
    --fno-code without -fwrite-interface should be avoided. If -fno-code were
    1267 
    -re-implemented today, -fwrite-interface would be discarded and it would be
    1268 
    -considered always on; this behaviour is as it is for backwards compatibility.
    1269 
    -
    1270 
    -================================================================
    1271 
    -IN SUMMARY: ALWAYS PASS -fno-code AND -fwrite-interface TOGETHER
    1272 
    -================================================================
    1273 
    -
    1274 
    -Template Haskell
    1275 
    -~~~~~~~~~~~~~~~~
    1276 
    -A module using template haskell may invoke an imported function from inside a
    1277 
    -splice. This will cause the type-checker to attempt to execute that code, which
    1278 
    -would fail if no object files had been generated. See #8025. To rectify this,
    1279 
    -during the downsweep we patch the DynFlags in the ModSummary of any home module
    1280 
    -that is imported by a module that uses template haskell, to generate object
    1281 
    -code.
    1282 
    -
    1283 
    -The flavour of the generated code depends on whether `-fprefer-byte-code` is enabled
    1284 
    -or not in the module which needs the code generation. If the module requires byte-code then
    1285 
    -dependencies will generate byte-code, otherwise they will generate object files.
    1286 
    -In the case where some modules require byte-code and some object files, both are
    1287 
    -generated by enabling `-fbyte-code-and-object-code`, the test "fat015" tests these
    1288 
    -configurations.
    1289 
    -
    1290 
    -The object files (and interface files if -fwrite-interface is disabled) produced
    1291 
    -for template haskell are written to temporary files.
    1292 
    -
    1293 
    -Note that since template haskell can run arbitrary IO actions, -fno-code mode
    1294 
    -is no more secure than running without it.
    1295 
    -
    1296 
    -Potential TODOS:
    1297 
    -~~~~~
    1298 
    -* Remove -fwrite-interface and have interface files always written in -fno-code
    1299 
    -  mode
    1300 
    -* Both .o and .dyn_o files are generated for template haskell, but we only need
    1301 
    -  .dyn_o. Fix it.
    1302 
    -* In make mode, a message like
    1303 
    -  Compiling A (A.hs, /tmp/ghc_123.o)
    1304 
    -  is shown if downsweep enabled object code generation for A. Perhaps we should
    1305 
    -  show "nothing" or "temporary object file" instead. Note that one
    1306 
    -  can currently use -keep-tmp-files and inspect the generated file with the
    1307 
    -  current behaviour.
    1308 
    -* Offer a -no-codedir command line option, and write what were temporary
    1309 
    -  object files there. This would speed up recompilation.
    1310 
    -* Use existing object files (if they are up to date) instead of always
    1311 
    -  generating temporary ones.
    1312 
    --}
    1313 1249
    1314 1250 
     -- Note [When source is considered modified]
    1315 1251 
     -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

  • compiler/GHC/Runtime/Eval.hs
    ... ... @@ -78,7 +78,7 @@ import GHC.Core.Type hiding( typeKind )
    78 78 
     import qualified GHC.Core.Type as Type
    79 79
    80 80 
     import GHC.Iface.Env       ( newInteractiveBinder )
    81 
    -import GHC.Iface.Load      ( loadSrcInterface )
    81 
    +import GHC.Iface.Load      ( loadInterfaceForModule )
    82 82 
     import GHC.Tc.Utils.TcType
    83 83 
     import GHC.Tc.Types.Constraint
    84 84 
     import GHC.Tc.Types.Origin
    ... ... @@ -843,7 +843,7 @@ mkTopLevEnv hsc_env modl
    843 843 
                           $ fmap (foldr plusGlobalRdrEnv emptyGlobalRdrEnv)
    844 844 
                           $ forM imports $ \iface_import -> do
    845 845 
                             let ImpUserSpec spec details = tcIfaceImport iface_import
    846 
    -                        iface <- loadSrcInterface (text "imported by GHCi") (moduleName $ is_mod spec) (is_isboot spec) (is_pkg_qual spec)
    846 
    +                        iface <- loadInterfaceForModule (text "imported by GHCi") (is_mod spec)
    847 847 
                             pure $ case details of
    848 848 
                               ImpUserAll -> importsFromIface hsc_env iface spec Nothing
    849 849 
                               ImpUserEverythingBut ns -> importsFromIface hsc_env iface spec (Just ns)

  • testsuite/tests/gadt/T23298.hs
    1 
    +{-# LANGUAGE GADTs #-}
    2 
    +module T23298 where
    3 
    +
    4 
    +import Data.Kind (Type)
    5 
    +
    6 
    +type HList :: Type -> Type
    7 
    +data HList a where
    8 
    +  HCons :: HList x -> HList (Maybe x)
    9 
    +
    10 
    +eq :: HList a -> Bool
    11 
    +eq x = case x of
    12 
    +         HCons ms -> True
    13 
    +
    14 
    +go (HCons x) = go x
    15 
    +
    16 
    +{- go :: HList alpha -> beta
    17 
    +
    18 
    +Under HCons
    19 
    +  [G] alpha ~ Maybe x
    20 
    +  [W] HList x ~ HList alpha
    21 
    +==>
    22 
    +  [W] x ~ alpha
    23 
    +==>
    24 
    +  [W] x ~ Maybe x
    25 
    +-}

  • testsuite/tests/gadt/T23298.stderr
    1 
    + T23298.hs:14:16: error: [GHC-25897]
    2 
    +    • Couldn't match type ‘x’ with ‘Maybe x’
    3 
    +      Expected: HList x -> t
    4 
    +        Actual: HList a -> t
    5 
    +      ‘x’ is a rigid type variable bound by
    6 
    +        a pattern with constructor:
    7 
    +          HCons :: forall x. HList x -> HList (Maybe x),
    8 
    +        in an equation for ‘go’
    9 
    +        at T23298.hs:14:5-11
    10 
    +    • In the expression: go x
    11 
    +      In an equation for ‘go’: go (HCons x) = go x
    12 
    +    • Relevant bindings include x :: HList x (bound at T23298.hs:14:11)

  • testsuite/tests/gadt/all.T
    ... ... @@ -131,3 +131,4 @@ test('T19847a', normalise_version('base'), compile, ['-ddump-types'])
    131 131 
     test('T19847b', normal, compile, [''])
    132 132 
     test('T23022', normal, compile, ['-dcore-lint'])
    133 133 
     test('T23023', normal, compile_fail, ['-O -dcore-lint']) # todo: move this test?
    134 
    +test('T23298', normal, compile_fail, [''])

  • testsuite/tests/ghci/scripts/GhciPackageRename.hs
    1 
    +module GhciPackageRename where
    2 
    +
    3 
    +foo :: Map k v
    4 
    +foo = empty
    \ No newline at end of file

  • testsuite/tests/ghci/scripts/GhciPackageRename.script
    1 
    +:l GhciPackageRename.hs
    2 
    +-- Test that Data.Map is available as Prelude
    3 
    +:t fromList
    4 
    +
    5 
    +-- Test using a Map function
    6 
    +fromList [(1,"a"), (2,"b")]
    \ No newline at end of file

  • testsuite/tests/ghci/scripts/GhciPackageRename.stdout
    1 
    +fromList
    2 
    +  :: ghc-internal:GHC.Internal.Classes.Ord k => [(k, a)] -> Map k a
    3 
    +fromList [(1,"a"),(2,"b")]

  • testsuite/tests/ghci/scripts/all.T
    ... ... @@ -386,3 +386,9 @@ test('T13869', extra_files(['T13869a.hs', 'T13869b.hs']), ghci_script, ['T13869.
    386 386 
     test('ListTuplePunsPpr', normal, ghci_script, ['ListTuplePunsPpr.script'])
    387 387 
     test('ListTuplePunsPprNoAbbrevTuple', [expect_broken(23135), limit_stdout_lines(13)], ghci_script, ['ListTuplePunsPprNoAbbrevTuple.script'])
    388 388 
     test('T24459', normal, ghci_script, ['T24459.script'])
    389 
    +
    390 
    +# Test package renaming in GHCi session
    391 
    +test('GhciPackageRename',
    392 
    +     [extra_hc_opts("-hide-all-packages -package 'containers (Data.Map as Prelude)'")],
    393 
    +     ghci_script,
    394 
    +     ['GhciPackageRename.script'])

    • —
    View it on GitLab.
    You're receiving this email because of your account on gitlab.haskell.org. Manage all notifications · Help