ghc-hetmet.git
17 years agoautoreconf -> sh boot
Ian Lynagh [Thu, 12 Apr 2007 22:36:36 +0000 (22:36 +0000)]
autoreconf -> sh boot

17 years agoDarwin/i386 NCG: fix assembly syntax in symbol stubs
wolfgang.thaller@gmx.net [Thu, 12 Apr 2007 21:53:13 +0000 (21:53 +0000)]
Darwin/i386 NCG: fix assembly syntax in symbol stubs

In i386 position-independent symbol stubs, the NCG used to print
jmp %edx
instead of
jmp *%edx

Apple's assembler used to silently accept this, but has recently started to print warnings about it.

MERGE TO STABLE.

17 years agoDarwin Linker: Do not add local symbols to lochash
wolfgang.thaller@gmx.net [Thu, 12 Apr 2007 21:17:12 +0000 (21:17 +0000)]
Darwin Linker: Do not add local symbols to lochash
... so that GHCi doesn't complain about duplicate symbols when two C modules define the same static variable.

MERGE TO STABLE.

17 years agoHandle Weak Definitions in the Darwin Linker
wolfgang.thaller@gmx.net [Thu, 12 Apr 2007 21:01:55 +0000 (21:01 +0000)]
Handle Weak Definitions in the Darwin Linker

The __i686.get_pc_thunk symbols generated by gcc's PIC code generator are weak definitions that appear in every object file, so we need to deal with them.

MERGE TO STABLE.

17 years agoCheck boot seems to have been run before we let make proceed
Ian Lynagh [Thu, 12 Apr 2007 22:30:14 +0000 (22:30 +0000)]
Check boot seems to have been run before we let make proceed

17 years agowibble to prep-bin-dist-mingw target
Simon Marlow [Thu, 12 Apr 2007 13:21:15 +0000 (13:21 +0000)]
wibble to prep-bin-dist-mingw target

17 years agotidy up the binary-dist targets a bit more
Simon Marlow [Thu, 12 Apr 2007 13:17:02 +0000 (13:17 +0000)]
tidy up the binary-dist targets a bit more

17 years agofurther tweak the comments
Simon Marlow [Wed, 11 Apr 2007 14:57:36 +0000 (14:57 +0000)]
further tweak the comments

17 years agofix comment
Simon Marlow [Wed, 11 Apr 2007 14:51:55 +0000 (14:51 +0000)]
fix comment

17 years agodo the right thing by default with a current mingw
Simon Marlow [Wed, 11 Apr 2007 14:48:25 +0000 (14:48 +0000)]
do the right thing by default with a current mingw
and hopefully be a bit cleverer about finding things automatically.

17 years agoMERGE: add prep-bin-dist-mingw to binary-dist target, and add a separate target for
Simon Marlow [Thu, 12 Apr 2007 13:13:27 +0000 (13:13 +0000)]
MERGE: add prep-bin-dist-mingw to binary-dist target, and add a separate target for
 publish-binary-dist

17 years agoMERGE: fixes to prep-bin-dist-mingw, and add separate rule to tar the bindist
Simon Marlow [Thu, 12 Apr 2007 13:10:12 +0000 (13:10 +0000)]
MERGE: fixes to prep-bin-dist-mingw, and add separate rule to tar the bindist

17 years agocall prep-bin-dist-mingw as part of binary-dist, before we make the tarball
Simon Marlow [Wed, 11 Apr 2007 14:52:35 +0000 (14:52 +0000)]
call prep-bin-dist-mingw as part of binary-dist, before we make the tarball

17 years agoRemove redundant import
Ian Lynagh [Wed, 11 Apr 2007 21:38:33 +0000 (21:38 +0000)]
Remove redundant import

17 years agoRationalise GhcMode, HscTarget and GhcLink
Simon Marlow [Wed, 11 Apr 2007 10:18:02 +0000 (10:18 +0000)]
Rationalise GhcMode, HscTarget and GhcLink

This patch cleans up the GHC API, and adds some functionality: we can
now compile to object code inside GHCi.

Previously we had:

  data GhcMode
    = BatchCompile
    | Interactive
    | OneShot
    | JustTypecheck
    | MkDepend

  data HscTarget
    = HscC
    | HscAsm
    | HscJava
    | HscInterpreted
    | HscNothing

There was redundancy here; if GhcMode is Interactive, then only
HscInterpreted makes sense, and JustTypecheck required HscNothing.
Now we have:

  data GhcMode
    = CompManager       -- ^ --make, GHCi, etc.
    | OneShot           -- ^ ghc -c Foo.hs
    | MkDepend          -- ^ ghc -M, see Finder for why we need this

and HscTarget remains as before.

Previously GhcLink looked like this:

  data GhcLink = NoLink | StaticLink

Now we have:

  data GhcLink = NoLink | LinkBinary | LinkInMemory

The idea being that you can have an HscTarget of HscAsm (for example)
and still link in memory.

There are two new flags:

  -fobject-code selects object code as the target (selects
                either -fasm or -fvia-C, whichever is the default)
                This can be usd with ':set' in GHCi, or on the command line.

  -fbyte-code   sets byte-code as the target.  Only works in GHCi.
                One day maybe this could save the byte code in a file
                when used outside GHCi.

  (names chosen for consistency with -fno-code).

Changes to the GHC API: newSession no longer takes the GhcMode
argument.  The GhcMode defaults to CompManager, which is usually what
you want.  To do JustTypecheck now, just set hscTarget to HscNothing.

17 years agoUse "$@" rather than $* when calling things in darcs-all
Ian Lynagh [Wed, 11 Apr 2007 01:22:58 +0000 (01:22 +0000)]
Use "$@" rather than $* when calling things in darcs-all

17 years agoBuild library Setup.hs's with -Wall
Ian Lynagh [Wed, 11 Apr 2007 00:25:16 +0000 (00:25 +0000)]
Build library Setup.hs's with -Wall

17 years agoparsec, regex-base, regex-compat, regex-posix are no longer core-packages
Ian Lynagh [Tue, 10 Apr 2007 22:03:06 +0000 (22:03 +0000)]
parsec, regex-base, regex-compat, regex-posix are no longer core-packages

17 years agoReplace genprimopcode's parsec parser with an alex+happy parser
Ian Lynagh [Tue, 10 Apr 2007 22:00:15 +0000 (22:00 +0000)]
Replace genprimopcode's parsec parser with an alex+happy parser
This use was the only thing keeping parsec in core-packages, and
we already have a dependency on alex+happy anyway.

17 years agoDon't use regexps in the GHC sources
Ian Lynagh [Tue, 10 Apr 2007 21:56:47 +0000 (21:56 +0000)]
Don't use regexps in the GHC sources
3 small regexps were responsible for pulling 3 packages into core-packages.

The new code should also do a better job of hiding
"call-clobbered register used" warnings.

17 years agoUse a boot script instead of having autoreconf recurse
Ian Lynagh [Tue, 10 Apr 2007 21:52:52 +0000 (21:52 +0000)]
Use a boot script instead of having autoreconf recurse
It used to be the case that
    autoreconf
in the root would also autoreconf in any libraries that need it, and
    ./configure
in the root would also configure all the libraries.

However, cabal now configures the libraries, so they were getting
configured twice.

Thus now a small shellscript, boot, autoreconfs the root and all libraries
that need it, and ./configure in the root doesn't configure the libraries.

17 years agoTweak AC_MSG_RESULTs
Ian Lynagh [Mon, 9 Apr 2007 13:57:34 +0000 (13:57 +0000)]
Tweak AC_MSG_RESULTs

17 years agoAllow the version date to be overridden with VERSION_DATE
Ian Lynagh [Mon, 9 Apr 2007 13:46:40 +0000 (13:46 +0000)]
Allow the version date to be overridden with VERSION_DATE

17 years agoUse cp -R rather than cp -a for portability reasons
Ian Lynagh [Fri, 6 Apr 2007 19:04:59 +0000 (19:04 +0000)]
Use cp -R rather than cp -a for portability reasons

17 years agoFixes for building the libraries with cabal on Windows
Ian Lynagh [Fri, 6 Apr 2007 15:18:22 +0000 (15:18 +0000)]
Fixes for building the libraries with cabal on Windows
* configure can be told where ld is
* make an hsc2hs-inplace.bat
* tell Setup configure about foo.bar rather than foo on Windows
* tell Setup configure to pass a --with-cc=$(CC) argument to ./configure

17 years agoFix external core syntax (though not full compilation)
Aaron Tomb [Mon, 13 Nov 2006 23:01:32 +0000 (23:01 +0000)]
Fix external core syntax (though not full compilation)

This patch updates the External Core creator, pretty-printer, and parser to
agree on a concrete syntax for External Core, including the constructs
required by the change to System FC. Code to create valid ASTs from External
Core files will come later, as will bits for renaming, typechecking, and
desugaring.

17 years agoRemove code that is dead, as we require __GLASGOW_HASKELL__ >= 504
Ian Lynagh [Fri, 6 Apr 2007 00:11:37 +0000 (00:11 +0000)]
Remove code that is dead, as we require __GLASGOW_HASKELL__ >= 504

17 years agoDon't break when asked to install-docs
Ian Lynagh [Fri, 16 Mar 2007 17:00:03 +0000 (17:00 +0000)]
Don't break when asked to install-docs
(we don't actually do anything useful either, because Cabal doesn't have
an install-docs option).

17 years agoAdd a comment
Ian Lynagh [Fri, 16 Mar 2007 16:59:57 +0000 (16:59 +0000)]
Add a comment

17 years agoMake building binary dists work with cabalised libraries
Ian Lynagh [Sat, 10 Mar 2007 01:25:58 +0000 (01:25 +0000)]
Make building binary dists work with cabalised libraries

17 years agoBuild the libraries with cabal
Ian Lynagh [Thu, 8 Mar 2007 16:32:47 +0000 (16:32 +0000)]
Build the libraries with cabal

17 years agofix adjustor generation on ia64 (test case ffi009)
red5_2@hotmail.com [Sun, 1 Apr 2007 16:31:32 +0000 (16:31 +0000)]
fix adjustor generation on ia64 (test case ffi009)

Some fixes to adjustor functions.  The 8-byte address returned by the
allocator is adjusted to be aligned to 16-byte boundaries.  Fixed a typo
in inserting an immediate address into an instruction.

This fixes the calls to 5-argument and 6-argument functions in ffi009.
Some functions still break.  I suspect it's related to passing arguments on
the stack.

17 years agosave two more FP registers on ia64
red5_2@hotmail.com [Tue, 27 Mar 2007 04:18:11 +0000 (04:18 +0000)]
save two more FP registers on ia64

Save/restore two more registers in StgCRun().  The extra
registers are used by ffi009.hs, when compiling with gcc 4.1.2.

17 years agoadd a safety net to avoid generating bogus code
Simon Marlow [Wed, 4 Apr 2007 15:50:01 +0000 (15:50 +0000)]
add a safety net to avoid generating bogus code

17 years agoadd comments about ticky RTS ways
Simon Marlow [Wed, 4 Apr 2007 14:10:13 +0000 (14:10 +0000)]
add comments about ticky RTS ways

17 years agoslight doc wibble for the -r flag
Simon Marlow [Wed, 4 Apr 2007 14:08:05 +0000 (14:08 +0000)]
slight doc wibble for the -r flag

17 years agoMERGE: Fix bug exposed by conc052.
Simon Marlow [Wed, 4 Apr 2007 10:30:27 +0000 (10:30 +0000)]
MERGE: Fix bug exposed by conc052.
A thread that was blocked on a blackhole but can now be woken up could
possibly be treated as unreachable by the GC, and sent the
NonTermination exception.

This can give rise to spurious <<loop>>s in concurrent programs, so
it's a good one to fix.

17 years agoHANDLE --> (void *), we can't rely on having included windows.h here
Simon Marlow [Wed, 4 Apr 2007 08:40:18 +0000 (08:40 +0000)]
HANDLE --> (void *), we can't rely on having included windows.h here
HANDLE is defined to be (void *) anyway, so this shouldn't hurt

17 years agoFix C/Haskell type mismatches
Ian Lynagh [Wed, 4 Apr 2007 00:34:25 +0000 (00:34 +0000)]
Fix C/Haskell type mismatches

17 years agoFix typo
Ian Lynagh [Wed, 4 Apr 2007 00:03:47 +0000 (00:03 +0000)]
Fix typo

17 years agoFix C/Haskell type mismatches
Ian Lynagh [Tue, 3 Apr 2007 19:53:24 +0000 (19:53 +0000)]
Fix C/Haskell type mismatches

17 years agoMake type-tidying work for coercion variables
simonpj@microsoft.com [Mon, 2 Apr 2007 07:38:35 +0000 (07:38 +0000)]
Make type-tidying work for coercion variables

When tidying a TyVar binder, we must tidy its kind if it's a coercion
variable!  I had forgotten to do this, which is a serious bug.  As a
result some more complicated programs were getting a Lint error when
reading in interface files.  Score one for Core Lint!

17 years agoMake error message output order consistent
Ian Lynagh [Sun, 1 Apr 2007 18:57:41 +0000 (18:57 +0000)]
Make error message output order consistent

17 years agoFix darcs-all to be usable with shells other than bash
Alec Berryman [Tue, 6 Mar 2007 17:30:11 +0000 (17:30 +0000)]
Fix darcs-all to be usable with shells other than bash

17 years agoFixed desugaring of parallel array comprehensions
Manuel M T Chakravarty [Mon, 2 Apr 2007 08:24:24 +0000 (08:24 +0000)]
Fixed desugaring of parallel array comprehensions

  ** MERGE into 6.6.1 **

17 years agoThe ru_local field of a CoreRule is False for implicit Ids
simonpj@microsoft.com [Fri, 30 Mar 2007 14:43:44 +0000 (14:43 +0000)]
The ru_local field of a CoreRule is False for implicit Ids

MERGE to 6.6.1

For class-ops, record selectors, data constructors, we want the ru_local
field of the Rule to be False.  We do not attach the rule to the binding
for the Id, because there simply isn't a binding until the code gen stage.

(NB: the ru_local field is different to the orphan-hood of the rule.)

This fixes a bug that meant that RULES on class ops were never exported.

17 years agoMatch the type of an Id during rule matching
simonpj@microsoft.com [Fri, 30 Mar 2007 14:00:33 +0000 (14:00 +0000)]
Match the type of an Id during rule matching

Please MERGE to 6.6.1

Consider this RULE
    forall (c::Char->Int) (x::Char).
f (c x) = "RULE FIRED"

Well, this should only match on arguments of the specified type
But we simply weren't checking this condition before.  Now we are.

Test is simplrun008

17 years agoAdd filepath library
Ian Lynagh [Wed, 28 Mar 2007 23:51:54 +0000 (23:51 +0000)]
Add filepath library

17 years agoFix the type of wgencat
Ian Lynagh [Thu, 29 Mar 2007 16:42:38 +0000 (16:42 +0000)]
Fix the type of wgencat

17 years agoDocument that ghci now adds () to the start of the type defaulting list
Ian Lynagh [Thu, 29 Mar 2007 14:16:31 +0000 (14:16 +0000)]
Document that ghci now adds () to the start of the type defaulting list

17 years agoRefactoring in HscMain.lhs
Lemmih [Wed, 28 Mar 2007 22:47:42 +0000 (22:47 +0000)]
Refactoring in HscMain.lhs

I realized that compiling many files is just like compiling a single file
several times. This does sound a bit stupid and I wonder why I didn't
catch it before.

17 years agoTypo in comment.
Lemmih [Wed, 28 Mar 2007 22:01:09 +0000 (22:01 +0000)]
Typo in comment.

17 years agoremove old library doc link (#1098), and point to HackageDB
Simon Marlow [Wed, 28 Mar 2007 10:29:38 +0000 (10:29 +0000)]
remove old library doc link (#1098), and point to HackageDB

17 years agoMake GHC main program depend on the libHSghc_dyn when GhcBuildDylibs==YES
wolfgang.thaller@gmx.net [Tue, 27 Mar 2007 21:01:22 +0000 (21:01 +0000)]
Make GHC main program depend on the libHSghc_dyn when GhcBuildDylibs==YES

When building a dynamic GHC, we obviously want to build the dynamic library
libHSghc_dyn.[so|dylib] before building the compiler executable.

MERGE TO STABLE

17 years agoUse a sensible set of compiler options for building dynamic libraries on Mac OS X
wolfgang.thaller@gmx.net [Tue, 27 Mar 2007 20:17:00 +0000 (20:17 +0000)]
Use a sensible set of compiler options for building dynamic libraries on Mac OS X

The official options to use for building a dylib on Mac OS X are now
-undefined dynamic_lookup -single_module -Wl,-macosx_version_min -Wl,10.3
... which should work on all arches (10.3 or later).

MERGE TO STABLE

17 years agomake GHCi use base:Prelude, not just Prelude
Simon Marlow [Tue, 27 Mar 2007 15:45:54 +0000 (15:45 +0000)]
make GHCi use base:Prelude, not just Prelude
The module that GHCi uses for its default scope should be exactly
base:Prelude, not whatever Prelude is found on the search path.

17 years agomore improvements for #1119
Simon Marlow [Tue, 27 Mar 2007 15:37:23 +0000 (15:37 +0000)]
more improvements for #1119
When GHCi compiles its code framgents for setting buffering, it wants
to refer to base:System.IO rather than whatever System.IO is on the
search path, unfortunately there's no way to do this in source code,
so to hack around it we set the search path to empty before compiling
these expressions (not forgetting to flush the finder cache
afterward).

17 years agopartial fix for #1119
Simon Marlow [Tue, 27 Mar 2007 15:22:15 +0000 (15:22 +0000)]
partial fix for #1119
Unless we're in one-shot mode, emit an error if we attempt to
demand-load interfaces for home modules.  This can only happen in one
way (that I'm aware of): typing a qualified name at the GHCi prompt
that refers to a module that isn't loaded.  Previously you got a
cryptic message about not finding an interface file, now you get:

Prelude> Foo.a

<interactive>:1:0:
    attempting to use module `Foo' (Foo.hs) which is not loaded

Of course you can still refer to package modules like this without
loading them explicitly, only home modules are affected, and the
behaviour is exactly the same as if you try to ':browse Foo' and
Foo isn't loaded.

17 years agoremove docs for unimplemented options
Simon Marlow [Mon, 26 Mar 2007 14:27:28 +0000 (14:27 +0000)]
remove docs for unimplemented options
-optdep--include-module and --optdep--exclude-directory were features
of the old mkdependHS script but weren't implemented when mkdependHS
was merged into GHC.

17 years agoAdd comments about the -include-prelude flag for ghc -M
simonpj@microsoft.com [Tue, 27 Mar 2007 11:36:12 +0000 (11:36 +0000)]
Add comments about the -include-prelude flag for ghc -M

17 years agoRemove dead code from rnExpr (Var v); seems to be a leftover from some breakpoint...
simonpj@microsoft.com [Fri, 23 Mar 2007 12:23:54 +0000 (12:23 +0000)]
Remove dead code from rnExpr (Var v); seems to be a leftover from some breakpoint stuff

17 years agomove "boot :: all" after the include of target.mk to fix #1095
Simon Marlow [Mon, 26 Mar 2007 15:59:49 +0000 (15:59 +0000)]
move "boot :: all" after the include of target.mk to fix #1095

17 years agoadd missing progress message
Simon Marlow [Mon, 26 Mar 2007 15:52:33 +0000 (15:52 +0000)]
add missing progress message

17 years agoEmit a decent error message when there is a decl-splice inside a decl-bracket
simonpj@microsoft.com [Fri, 23 Mar 2007 11:19:12 +0000 (11:19 +0000)]
Emit a decent error message when there is a decl-splice inside a decl-bracket

This fixes Trac #1065.  The fix is just to emit a decent error message
rather than crash.  The situation is this:

  f x = ... [d| $(..stuff..);
                f x = v :: T |] ...

TH wants to rename and typecheck the bracket; but it can't run the
nested splice yet.  That seems hard, because we know nothing about v, T,
which are, presumably bound by the splice.

The original TH paper says this isn't allowed, and now it's checked for
properly (in the parser, in fact) rather than causing a crash.  In the
fullness of time we might want to do something more flexible, but not now.

17 years agoWibble fixing overlapped pattern
simonpj@microsoft.com [Fri, 23 Mar 2007 11:18:50 +0000 (11:18 +0000)]
Wibble fixing overlapped pattern

17 years agoTidy up refactoring only
simonpj@microsoft.com [Fri, 23 Mar 2007 11:18:21 +0000 (11:18 +0000)]
Tidy up refactoring only

17 years agoDocumentation for --install-signal-handlers=<yes|no>
Ian Lynagh [Thu, 22 Mar 2007 18:46:07 +0000 (18:46 +0000)]
Documentation for --install-signal-handlers=<yes|no>

17 years agoAdd an --install-signal-handlers=<yes|no> RTS flag; fixes trac #804
Ian Lynagh [Thu, 22 Mar 2007 18:29:00 +0000 (18:29 +0000)]
Add an --install-signal-handlers=<yes|no> RTS flag; fixes trac #804

17 years agoMake a Functor (IOEnv m) instance so it satisfies the new Quasi requirements
Ian Lynagh [Thu, 22 Mar 2007 17:01:14 +0000 (17:01 +0000)]
Make a Functor (IOEnv m) instance so it satisfies the new Quasi requirements

17 years agoWarning police: Removed catch-all case which was overlapped
sven.panne@aedion.de [Thu, 22 Mar 2007 16:10:40 +0000 (16:10 +0000)]
Warning police: Removed catch-all case which was overlapped

17 years agoAdd -ddump-mod-cycles to -M behaviour
simonpj@microsoft.com [Thu, 22 Mar 2007 12:28:59 +0000 (12:28 +0000)]
Add -ddump-mod-cycles to -M behaviour

This patch adds a flag -ddump-mod-cycles to the "ghc -M" dependency analyser.

The effect of
ghc -M -ddump-mod-cycles
is to dump a list of cycles foud in the module graph.  The display is
trimmed so that only dependencies within the cycle are shown; and the
list of modules in a cycle is itself sorted into dependency order, so that
it is easy to track the chain of dependencies.

Open question: should the flag be "-ddump-mod-cycles" or "-optdep-dump-mod-cycles"?  For this reason I have not yet added to the documentation.

17 years agoImprove documentation of instances
simonpj@microsoft.com [Thu, 22 Mar 2007 11:07:18 +0000 (11:07 +0000)]
Improve documentation of instances

17 years agoFix checkHiBootIface for instances declared in hs-boot files
simonpj@microsoft.com [Wed, 21 Mar 2007 14:04:24 +0000 (14:04 +0000)]
Fix checkHiBootIface for instances declared in hs-boot files

Hs-boot files can contain instance declarations, but Duncan used this feature
for the first time today, and it didn't quite work.  I'm not sure whether
it worked before; anyway it does now.

The point is that the hs-boot file advertises an instance for, say, Num Int,
with the arbitrary name $fx1.  The main module declares Num Int, and gives
it the name, say, $f3.  So we need to generate a declaration
$fx1 = $f3
to make it all line up.  And (this is the bit that was wrong) we need
to make that new binding visible to imported unfoldings that mention
$fx1.

17 years agomangler and runtime updates for gcc 4 on ia64
red5_2@hotmail.com [Sun, 18 Mar 2007 03:31:32 +0000 (03:31 +0000)]
mangler and runtime updates for gcc 4 on ia64

Gcc 4 is doing more clever optimizations than earlier gccs.
These changes let ghc compile and run on ia64 with gcc 4.0.3.

Register stack frames are enlarged so that all functions use the
same size stack frame.
The code to mangle tail calls has been cleaned up and made more
general.
Additional floating-point and special-purpose registers used by
GCC are saved upon entering the STG runtime.
More general handling of NOP instructions.
Handling of functions with multiple epilogues or no epilogue.

17 years agoia64 tailcall fix
red5_2@hotmail.com [Sat, 17 Mar 2007 02:55:29 +0000 (02:55 +0000)]
ia64 tailcall fix

Tail-calls now branch to the end of the function after the call.  This
stops gcc from trying to move instructions into the space immediately
after the call, which confuses the mangler.

17 years agoRemove bogus export from TcSplice.lhs-boot
simonpj@microsoft.com [Wed, 21 Mar 2007 11:20:57 +0000 (11:20 +0000)]
Remove bogus export from TcSplice.lhs-boot

17 years agoWibble to GHCi part of TcRnDriver
simonpj@microsoft.com [Wed, 21 Mar 2007 11:16:39 +0000 (11:16 +0000)]
Wibble to GHCi part of TcRnDriver

17 years agoYet another wibble to checkHiBootIface; it's trickier than it looks!
simonpj@microsoft.com [Wed, 21 Mar 2007 09:37:33 +0000 (09:37 +0000)]
Yet another wibble to checkHiBootIface; it's trickier than it looks!

17 years agoImprove pretty-printing for IfaceConDecl
simonpj@microsoft.com [Wed, 21 Mar 2007 09:36:11 +0000 (09:36 +0000)]
Improve pretty-printing for IfaceConDecl

17 years agoRemove redundant exports from a hs-boot file
simonpj@microsoft.com [Wed, 21 Mar 2007 09:35:25 +0000 (09:35 +0000)]
Remove redundant exports from a hs-boot file

17 years agoFurther wibbles to checkBootIface (fixed HEAD build)
simonpj@microsoft.com [Tue, 20 Mar 2007 15:41:26 +0000 (15:41 +0000)]
Further wibbles to checkBootIface (fixed HEAD build)

17 years agoUse explict layout (fixes breakage when compiling TcSimplify)
simonpj@microsoft.com [Mon, 19 Mar 2007 15:19:26 +0000 (15:19 +0000)]
Use explict layout (fixes breakage when compiling TcSimplify)

17 years agoFix typo "lazy y x" -> "lazy y"
Ian Lynagh [Sat, 17 Mar 2007 00:40:03 +0000 (00:40 +0000)]
Fix typo "lazy y x" -> "lazy y"

17 years agoDesugar applications of 'seq' specially; fix Trac #1031
simonpj@microsoft.com [Fri, 16 Mar 2007 15:17:12 +0000 (15:17 +0000)]
Desugar applications of 'seq' specially; fix Trac #1031

Merge to 6.6 branch.  Test case is dsrun014.

Note [Desugaring seq]  cf Trac #1031
~~~~~~~~~~~~~~~~~~~~~
   f x y = x `seq` (y `seq` (# x,y #))

The [CoreSyn let/app invariant] means that, other things being equal, because
the argument to the outer 'seq' has an unlifted type, we'll use call-by-value thus:

   f x y = case (y `seq` (# x,y #)) of v -> x `seq` v

But that is bad for two reasons:
  (a) we now evaluate y before x, and
  (b) we can't bind v to an unboxed pair

Seq is very, very special!  So we recognise it right here, and desugar to
case x of _ -> case y of _ -> (# x,y #)

The special case would be valid for all calls to 'seq', but it's only *necessary*
for ones whose second argument has an unlifted type. So we only catch the latter
case here, to avoid unnecessary tests.

17 years agoImprove error messages for 'deriving' clauses
simonpj@microsoft.com [Fri, 16 Mar 2007 14:28:24 +0000 (14:28 +0000)]
Improve error messages for 'deriving' clauses

This patch improves the misleading error message reported in Trac #1133.

Please merge the patch to the 6.6.1 branch.

17 years agoRefactor TcRnDriver, and check exports on hi-boot files
simonpj@microsoft.com [Fri, 16 Mar 2007 13:38:50 +0000 (13:38 +0000)]
Refactor TcRnDriver, and check exports on hi-boot files

This patch refactors TcRnDriver to make the top-level structure
easier to understand.

The change was driven by Trac #924, and this patch fixes that bug.
When comparing a module against its hs-boot file, we must ensure that
the module exports everything that the hs-boot file exports.

17 years agoComment out debug traces
simonpj@microsoft.com [Fri, 16 Mar 2007 13:38:22 +0000 (13:38 +0000)]
Comment out debug traces

17 years agoUse update-alternatives for handling generic tool names
sven.panne@aedion.de [Thu, 15 Mar 2007 15:28:23 +0000 (15:28 +0000)]
Use update-alternatives for handling generic tool names

ATTENTION: Packagers should read the following stuff carefully!

GHC, Hugs and nhc come with various tools like runhaskell or hsc2hs. On the
one hand this is quite handy, avoiding lots of tiny native packages, but OTOH
this leads to a few problems:

   * The tools are not always identical in functionality.

   * The tools fight for a global generic name like "/usr/bin/runhaskell".

These problems are not new and not unique to Haskell implementations, so for
*nix-based system there is a tool called update-alternatives which handles
those cases. The idea is as follows:

   * Each program/man page/etc. installs itself with a very specific name
     like /usr/bin/hsc2hs-ghc or /usr/share/man/man1/lua5.1.1.gz, so nothing
     clashes.

   * The (un-)installation scripts call update-alternatives to notify the
     system about new alternatives for a generic tool/manpage/etc.

   * Alternatives can be grouped together ("link groups"), so e.g. switching
     from Sun's Java to Kaffe switches compiler, JRE, manpages etc. together.
     Alas, this doesn't work well with the Haskell implementations yet,
     because they come with different sets of tools (in addition to runFOO):

       GHC:  hsc2hs
       Hugs: hsc2hs, cpphs
       nhc:  cpphs

     Either these tools should be disentangled fromt the Haskell
     implementations or all implementations should offer the same set.
     Opinions and recommendations on this topic are highly welcome.

   * This mechanism can be used to easily switch between several versions of
     the same implementation, too, but we are not yet fully prepared for that.

As a first step, GHC now installs hsc2hs as 'hsc2hs-ghc' and does *not*
install runhaskell directly anymore, only runghc. hsc2hs and runhaskell are
created via update-alternatives now. What is currently missing is a mechanism
for platforms like Windows and probably Mac OS X.

17 years agoMake the type-defaulting in GHCi use () as the first default type
simonpj@microsoft.com [Thu, 15 Mar 2007 14:28:12 +0000 (14:28 +0000)]
Make the type-defaulting in GHCi use () as the first default type

See Trac #1200

This is a somewhat experimental fix.  I'm not sure we want it in 6.6.1

The idea is explained in Note [Default unitTy] in TcSimplify.  In
interative mode (or with -fextended-default-rules) we add () as the
first type we try when defaulting.  This has very little real impact,
except in the following case.  Consider:
Text.Printf.printf "hello"
This has type (forall a. IO a); it prints "hello", and returns
'undefined'.  We don't want the GHCi repl loop to try to print that
'undefined'.  The neatest thing is to default the 'a' to (), rather
than to Integer (which is what would otherwise happen; and then GHCi
doesn't attempt to print the ().  So in interactive mode, we add () to
the list of defaulting types.

17 years agoAdded support for parallel builds
sven.panne@aedion.de [Thu, 15 Mar 2007 12:24:57 +0000 (12:24 +0000)]
Added support for parallel builds

With this patch, one can define the degree of build parallelism via a 'jobs'
rpm variable. A comfortable way to use this is having a ~/.rpmmacros file with
a line like:

   %jobs 2

Alternatively, one could use a '--define "jobs 2"' command line flag for
rpmbuild. On a Core 2 Duo using 2 jobs brings down the time for a full build
including extralibs from 36m to 27m. If 'jobs' is not defined, a normal
sequential build is done, following the usual conventions on e.g. openSUSE.

17 years agoWibble to type-error-message improvement
simonpj@microsoft.com [Wed, 14 Mar 2007 16:48:50 +0000 (16:48 +0000)]
Wibble to type-error-message improvement

17 years agoImprove error messages slightly, saying "a1...an" instead of "a b c"
simonpj@microsoft.com [Wed, 14 Mar 2007 16:29:06 +0000 (16:29 +0000)]
Improve error messages slightly, saying "a1...an" instead of "a b c"

17 years agoTweak hasktags to cope with abc::def correctly
Ian Lynagh [Wed, 14 Mar 2007 16:01:02 +0000 (16:01 +0000)]
Tweak hasktags to cope with abc::def correctly
Patch from Marc Weber in
http://www.haskell.org/pipermail/glasgow-haskell-users/2007-February/011998.html

17 years agoUnbreak Haddock markup
sven.panne@aedion.de [Wed, 14 Mar 2007 15:15:17 +0000 (15:15 +0000)]
Unbreak Haddock markup

17 years agoWarning police: Avoid name shadowing
sven.panne@aedion.de [Wed, 14 Mar 2007 14:04:34 +0000 (14:04 +0000)]
Warning police: Avoid name shadowing

17 years agoHandle primitive types and pseudo ops when generating LaTeX
sven.panne@aedion.de [Wed, 14 Mar 2007 13:43:28 +0000 (13:43 +0000)]
Handle primitive types and pseudo ops when generating LaTeX

17 years agoWarning police: Replace patter matching failures by something more informative
sven.panne@aedion.de [Wed, 14 Mar 2007 13:40:44 +0000 (13:40 +0000)]
Warning police: Replace patter matching failures by something more informative

17 years agoAdded LaTeX commands for primitive types and pseudo ops
sven.panne@aedion.de [Wed, 14 Mar 2007 13:27:03 +0000 (13:27 +0000)]
Added LaTeX commands for primitive types and pseudo ops

17 years agoFixed LaTeX markup
sven.panne@aedion.de [Wed, 14 Mar 2007 13:25:21 +0000 (13:25 +0000)]
Fixed LaTeX markup

17 years agoWarning police: Don't rely on defaulting
sven.panne@aedion.de [Wed, 14 Mar 2007 12:36:08 +0000 (12:36 +0000)]
Warning police: Don't rely on defaulting