The significant changes to the various parts of the compiler are listed in the following sections. The new QuasiQuotes language extension adds general quasi-quotation, as described in "Nice to be Quoted: Quasiquoting for Haskell" (Geoffrey Mainland, Haskell Workshop 2007). See for more information. The new ViewPatterns language extension allows "view patterns". The syntax for view patterns is expression -> pattern in a pattern. For more information, see . GHC already supported (e op) postfix operators, but this support was enabled by default. Now you need to use the PostfixOperators language extension if you want it. See for more information on postfix operators. The new TransformListComp language extension enables implements generalised list comprehensions, as described in the paper "Comprehensive comprehensions" (Peyton Jones & Wadler, Haskell Workshop 2007). For more information see . If you want to use impredicative types then you now need to enable the ImpredicativeTypes language extension. See for more information. FFI change: header files are now not used when compiling via C. The -#include flag, the includes field in .cabal files, and header files specified in a foreign import declaration all have no effect when compiling Haskell source code. This change has important ramifications if you are calling FFI functions that are defined by macros (or renamed by macros). If you need to call one of these functions, then write a C wrapper for the function and call the wrapper using the FFI instead. In this way, your code will work with GHC 6.10.1, and will also work with -fasm in older GHCs. This change was made for several reasons. Firstly, -fvia-C now behaves consistently with -fasm, which is important because we intend to stop compiling via C in the future. Also, we don't need to worry about the interactions between header files, or CPP options necessary to expose certain functions from the system header files (this was becoming quite a headache). We don't need to worry about needing header files when inlining FFI calls across module or package boundaries; calls can now be inlined freely. One downside is that you don't get a warning from the C compiler when you call a function via the FFI at the wrong type. Another consequence of this change is that calling varargs functions (such as printf) via the FFI no longer works. It has never been officially supported (the FFI spec outlaws it), but in GHC 6.10.1 it may now really cause a crash on certain platforms. Again, to call one of these functions use appropriate fixed-argument C wrappers. There is a new languages extension PackageImports which allows imports to be qualified with the package they should come from, e.g. import "network" Network.Socket Note that this feature is not intended for general use, it was added for constructing backwards-compatibility packages such as the base-3.0.3.0 package. See for more details. In earlier versions of GHC, the recompilation checker didn't notice changes in other packages meant that recompilation is needed. This is now handled properly, using MD5 checksums of the interface ABIs. GHC now treats the Unicode "Letter, Other" class as lowercase letters. This is an arbitrary choice, but better than not allowing them in identifiers at all. This may be revisited by Haskell'. In addition to the DEPRECATED pragma, you can now attach arbitrary warnings to declarations with the new WARNING pragma. See for more details. If GHC is failing due to -Werror, then it now emits a message telling you so. GHC now warns about unrecognised pragmas, as they are often caused by a typo. The -fwarn-unrecognised-pragmas controls whether this warning is emitted. The warning is enabled by default. There is a new flag -fwarn-dodgy-foreign-imports which controls a new warning about FFI delcarations of the form foreign import "f" f :: FunPtr t on the grounds that it is probably meant to be foreign import "&f" f :: FunPtr t The warning is enabled by default. External core (output only) is working again. There is a new flag -dsuppress-uniques that makes GHC's intermediate core easier to read. This flag cannot be used when actually generating code. There is a new flag -dno-debug-output that suppresses all of the debug information when running a compiler built with the DEBUG option. A bug in earlier versions of GHC meant that sections didn't always need to be parenthesised, e.g. (+ 1, 2) was accepted. This has now been fixed. The -fspec-threshold flag has been replaced by -fspec-constr-threshold and -fliberate-case-threshold flags. The thresholds can be disabled by -fno-spec-constr-threshold and -fno-liberate-case-threshold. The new flag -fsimplifier-phases controls the number of simplifier phases run during optimisation. These are numbered from n to 1 (by default, n=2). Phase 0 is always run regardless of this flag. Simplifier phases can have an arbitrary number of tags assigned to them, and multiple phases can share the same tags. The tags can be used as arguments to the new flag -ddump-simpl-phases to specify which phases are to be dumped. For example, -ddump-simpl-phases=main will dump the output of phases 2, 1 and 0 of the initial simplifier run (they all share the "main" tag) while -ddump-simpl-phases=main:0 will dump only the output of phase 0 of that run. At the moment, the supported tags are main (the main, staged simplifier run (before strictness)), post-worker-wrapper (after the w/w split), post-liberate-case (after LiberateCase), and final (final clean-up run) The names are somewhat arbitrary and will change in the future. The -fno-method-sharing flag is now dynamic (it used to be static). The new flag -fwarn-deprecated-flags, controls whether we warn about deprecated flags and language extensions. The warning is on by default. The following language extensions are now marked as deprecated; expect them to be removed in a future release: RecordPuns (use NamedFieldPuns instead) PatternSignatures (use ScopedTypeVariables instead) The following flags are now marked as deprecated; expect them to be removed in a future release: -Onot (use -O0 instead) -Wnot (use -w instead) -frewrite-rules (use -fenable-rewrite-rules instead) -no-link (use -c instead) -recomp (use -fno-force-recomp instead) -no-recomp (use -fforce-recomp instead) -syslib (use -package instead) -fth (use the TemplateHaskell language extension instead) -ffi, -fffi (use the ForeignFunctionInterface extension instead) -farrows (use the Arrows language extension instead) -fgenerics (use the Generics language extension instead) -fno-implicit-prelude (use the NoImplicitPrelude language extension instead) -fbang-patterns (use the BangPatterns language extension instead) -fno-monomorphism-restriction (use the NoMonomorphismRestriction language extension instead) -fmono-pat-binds (use the MonoPatBinds language extension instead) -fextended-default-rules (use the ExtendedDefaultRules language extension instead) -fimplicit-params (use the ImplicitParams language extension instead) -fscoped-type-variables (use the ScopedTypeVariables language extension instead) -fparr (use the PArr language extension instead) -fallow-overlapping-instances (use the OverlappingInstances language extension instead) -fallow-undecidable-instances (use the UndecidableInstances language extension instead) -fallow-incoherent-instances (use the IncoherentInstances language extension instead) -optdep-s (use -dep-suffix instead) -optdep-f (use -dep-makefile instead) -optdep-w (has no effect) -optdep--include-prelude (use -include-pkg-deps instead) -optdep--include-pkg-deps (use -include-pkg-deps instead) -optdep--exclude-module (use -exclude-module instead) -optdep-x (use -exclude-module instead) The following flags have been removed: -no-link-chk (has been a no-op since at least 6.0) -fruntime-types (has not been used for years) -fhardwire-lib-paths (use -dynload sysdep) The -unreg flag, which was used to build unregisterised code with a registerised compiler, has been removed. Now you need to build an unregisterised compiler if you want to build unregisterised code. There is now a Ghc Monad used to carry around GHC's Session data. This Monad also provides exception handling functions. It is now possible to get the raw characters corresponding to each token the lexer outputs, and thus to reconstruct the original file. GHCi implicitly brings all exposed modules into scope with qualified module names. There is a new flag -fimplicit-import-qualified that controls this behaviour, so other GHC API clients can specify whether or not they want it. There is now haddock documentation for much of the GHC API. You can now force GHCi to interpret a module, rather than loading its compiled code, by prepending a * character to its name, e.g. Prelude> :load *A Compiling A ( A.hs, interpreted ) *A> By default, GHCi will not print bind results, e.g. Prelude> c <- return 'c' Prelude> does not print 'c'. Use -fprint-bind-result if you want the old behaviour. GHCi now uses editline, rather than readline, for input. This shouldn't affect its behaviour. The GHCi prompt history is now saved in ~/.ghc/ghci_history. GHCi now uses libffi to make FFI calls, which means that the FFI now works in GHCi on a much wider range of platforms (all those platforms that libffi supports). The garbage collector can now use multiple threads in parallel. The new -gn RTS flag controls it, e.g. run your program with +RTS -g2 -RTS to use 2 threads. The -g option is implied by the usual -N option, so normally there will be no need to specify it separately, although occasionally it is useful to turn it off with -g1. Do let us know if you experience strange effects, especially an increase in GC time when using the parallel GC (use +RTS -s -RTS to measure GC time). See for more details. It is now possible to generate a heap profile without recompiling your program for profiling. Run the program with +RTS -hT to generate a basic heap profile, and use hp2ps as usual to convert the heap profile into a .ps file for viewing. See for more details. If the user presses control-C while running a Haskell program then the program gets an asynchronous UserInterrupt exception. We now ignore SIGPIPE by default. The -S and -s RTS flags now send their output to stderr, rather than prog.stat, by default. The new -vg RTS flag provides some RTS trace messages even in the non-debug RTS variants. runghc now uses the compiler that it came with to run the code, rather than the first compiler that it finds on the PATH. If the program to run does not have a .lhs extension then runghc now treats it as a .hs file. In particular, this means that programs without an extension now work. runghc foo will now work if foo.hs or foo.lhs exists. runghc can now take the code to run from stdin. ghc-pkg will refuse to unregister a package on which other packages depend, unless the ––force option is also supplied. ghc-pkg now has a -no-user-package-conf flag which instructs it to ignore the user's personal package.conf. ghc-pkg no longer allows you to register two packages that differ in case only. ghc-pkg no longer allows you to register packages which have unversioned dependencies. There is a new command dump which is similar to describe '*', but in a format that is designed to be parsable by other tools. Haddock 2 now comes with GHC. DPH is now an extralib. There is a new flag -Odph that sets the flags recommended when using DPH. Currently it is equivalent to -O2 -fno-method-sharing -fdicts-cheap -fmax-simplifier-iterations20 -fno-spec-constr-threshold There are now flags -fdph-seq and -fdph-par for selecting which DPH backend to use. The -fflatten flag has been removed. It never worked and has now been superceded by vectorisation. Version number 0.2.0.0 (was 0.1.0.0) Version number 4.0.0.0 (was 3.0.2.0) We also ship a base version 3.0.3.0, so legacy code should continue to work. The Show instance for Ratio now puts spaces around the %, as required by Haskell 98. There is a new module Control.Category. >>> is no longer a method of the Arrow class; instead Category is a superclass of Arrow. pure is no longer a method of the Arrow class; use arr instead. Control.Exception now uses extensible exceptions. The old style of exceptions are still available in Control.OldException, but we expect to remove them in a future release. There is a new function System.Exit.exitSuccess :: IO a analogous to the existing System.Exit.exitFailure :: IO a. There are new functions Data.Either.lefts :: [Either a b] -> [a], Data.Either.rights :: [Either a b] -> [b] and Data.Either.partitionEithers :: [Either a b] -> ([a], [b]) . The new function Data.List.subsequences :: [a] -> [[a]] gives all sublists of a list, e.g. subsequences "abc" == ["","a","b","ab","c","ac","bc","abc"] . The new function Data.List.permutations :: [a] -> [[a]] gives all permutations of a list, e.g. permutations "abc" == ["abc","bac","cba","bca","cab","acb"] . The new functions Data.Traversable.mapAccumL and Data.Traversable.mapAccumR generalise their Data.List counterparts to work on any Traversable type. The new function Control.Exception.blocked :: IO Bool tells you whether or not exceptions are blocked (as controlled by Control.Exception.(un)block). There is a new function traceShow :: Show a => a -> b -> b in Debug.Trace. The type of Control.Monad.forever has been generalised from Monad m => m a -> m () to Monad m => m a -> m b. The new value GHC.Exts.maxTupleSize tells you the largest tuple size that can be used. This is mostly of use in Template Haskell programs. GHC.Exts now exports Down(..), groupWith, sortWith and the which are used in the desugaring of generalised comprehensions. GHC.Exts no longer exports the Integer internals. If you want them then you need to get them directly from the new integer package. The new function GHC.Conc.threadStatus allows you to ask whether a thread is running, blocked on an MVar, etc. The Data.Generics hierarchy has been moved to a new package syb. The GHC.Prim and GHC.PrimopWrappers modules have been moved into a new ghc-prim package. Version number 0.9.0.1.2 (was 0.9.0.1.1) Version number 1.6.0.1 (was 1.2.4.0) Many API changes. See the Cabal docs for more information. Version number 0.2.0.0 (was 0.1.0.2) Various result type now use Maybe rather than allowing any Monad. Version number 1.0.0.2 (was 1.0.0.1) No longer defines the UNICODE CPP symbol for packages that use it. This is a new bootlib, version 0.2.1.0. Version number 1.1.0.1 (was 1.1.0.0) This is a new bootlib, version 0.1.0.0. Version number 1.0.1.0 (unchanged) Version number 0.5.0.2 (was 0.5.0.1) This is a new bootlib, version 0.1.0.0. Version number 1.0.0.1 (was 1.0.0.0) Version number 1.0.0.1 (was 1.0.0.0) Version number 0.1.0.1 (was 0.1.0.0) Version number 1.0.1.0 (was 1.0.0.0) There is a new combinator zeroWidthText :: String -> Doc for printing things like ANSI escape sequences. Version number 1.0.1.0 (was 1.0.0.1) The System.Process API has been overhauled. The new API is a superset of the old API, however. Version number 1.0.0.1 (was 1.0.0.0) This is no longer a bootlib; editline replaces it. This is a new bootlib, version 0.1.0.0. Version number 2.3.0.0 (was 2.2.0.0) The datatypes now have support for Word primitives. currentModule :: Q String has been replaced with location :: Q Loc, where Loc is a new datatype. Version number 2.3.1.0 (was 2.3.0.1) The System.Posix.Terminal.BaudRate type now includes B57600 and B115200 constructors. Version number 2.2.0.0 (was 2.1.1.1) No longer defines the UNICODE CPP symbol for packages that use it.