ghc.git
8 years agoadd a new trace class for spark events
Duncan Coutts [Fri, 8 Jul 2011 15:36:16 +0000 (17:36 +0200)] 
add a new trace class for spark events

8 years agoadd a missing traceSparkCounters invocation
Mikolaj [Tue, 5 Jul 2011 16:49:37 +0000 (18:49 +0200)] 
add a missing traceSparkCounters invocation

8 years agoAdd spark counter tracing
Duncan Coutts [Mon, 6 Jun 2011 09:39:16 +0000 (10:39 +0100)] 
Add spark counter tracing

A new eventlog event containing 7 spark counters/statistics: sparks
created, dud, overflowed, converted, GC'd, fizzled and remaining.
These are maintained and logged separately for each capability.
We log them at startup, on each GC (minor and major) and on shutdown.

8 years agoMove allocation of spark pools into initCapability
Duncan Coutts [Mon, 6 Jun 2011 09:32:33 +0000 (10:32 +0100)] 
Move allocation of spark pools into initCapability

Rather than a separate phase of initSparkPools. It means all the spark
stuff for a capability is initialisaed at the same time, which is then
becomes a good place to stick an initial spark trace event.

8 years agoAdd assertion of the invariant for the spark counters
Duncan Coutts [Thu, 2 Jun 2011 16:28:56 +0000 (17:28 +0100)] 
Add assertion of the invariant for the spark counters

The invariant is: created = converted + remaining + gcd + fizzled
Since sparks move between capabilities, we have to aggregate the
counters over all capabilities. This in turn means we can only check
the invariant at stable points where all but one capabilities are
stopped. We can do this at shutdown time and before and after a global
synchronised GC.

8 years agoFix wrong arg name in a C function prototype
Duncan Coutts [Thu, 2 Jun 2011 15:16:14 +0000 (16:16 +0100)] 
Fix wrong arg name in a C function prototype

8 years agoClassify overflowed sparks separately
Duncan Coutts [Wed, 1 Jun 2011 18:48:15 +0000 (19:48 +0100)] 
Classify overflowed sparks separately

When you use `par` to make a spark, if the spark pool on the current
capability is full then the spark is discarded. This represents a
loss of potential parallelism and it also means there are simply a
lot of sparks around. Both are things that might be of concern to a
programmer when tuning a parallel program that uses par.

The "+RTS -s" stats command now reports overflowed sparks, e.g.
SPARKS: 100001 (15521 converted, 84480 overflowed, 0 dud, 0 GC'd, 0 fizzled)

8 years agoUse a struct for the set of spark counters
Duncan Coutts [Wed, 1 Jun 2011 18:38:15 +0000 (19:38 +0100)] 
Use a struct for the set of spark counters

8 years agoChange tryStealSpark so it does not consume fizzled sparks
Duncan Coutts [Wed, 1 Jun 2011 17:17:27 +0000 (18:17 +0100)] 
Change tryStealSpark so it does not consume fizzled sparks

We want to count fizzled sparks accurately. Now tryStealSpark returns
fizzled sparks, and the callers now update the fizzled spark count.

8 years agoImprove the newSpark dud test by using the pointer tag bits
Duncan Coutts [Wed, 1 Jun 2011 16:16:29 +0000 (17:16 +0100)] 
Improve the newSpark dud test by using the pointer tag bits

newSpark() checks if the spark is a dud, and if so does not add it to
the spark pool. Previously, newSpark would discard the pointer tag bits
and just check closure_SHOULD_SPARK(p). We can take advantage of the
tag bits which can tell us if the pointer points to a value. If it is,
it's a dud spark and we don't need to add it to the spark pool.

8 years agoFix Windows breakage (#5322). When I modified StgRun to use the pure
Simon Marlow [Mon, 18 Jul 2011 12:48:53 +0000 (13:48 +0100)] 
Fix Windows breakage (#5322).  When I modified StgRun to use the pure
assembly version as part of the fix for #5250, we inadvertently lost
the Windows magic for extending the stack.  Win32 requires that the
stack is extended a page at a time, otherwise you get a segfault.  The
C compiler knows how to do this, so we now call a C stub to ensure
there's enough stack space at each invocation of the scheduler.

8 years agofix bug in sdist-prep rule (hpc/dist is now hpc/dist-install)
Simon Marlow [Mon, 18 Jul 2011 10:19:22 +0000 (11:19 +0100)] 
fix bug in sdist-prep rule (hpc/dist is now hpc/dist-install)

8 years agoRemove bogus ASSERT
Simon Peyton Jones [Mon, 18 Jul 2011 12:54:00 +0000 (13:54 +0100)] 
Remove bogus ASSERT

8 years agoMerge branch 'master' of http://darcs.haskell.org/ghc
Simon Peyton Jones [Mon, 18 Jul 2011 08:12:15 +0000 (09:12 +0100)] 
Merge branch 'master' of darcs.haskell.org/ghc

8 years agoFix permissions of package.cache when installing; fixes trac #5311
Ian Lynagh [Sun, 17 Jul 2011 18:58:37 +0000 (19:58 +0100)] 
Fix permissions of package.cache when installing; fixes trac #5311

When we install the packages, ghc-pkg obeys umask when creating
package.cache, but for everything else we specify the permissions. We
therefore fix the permissions of package.cache afterwards.

8 years agoFix build with older bootstrapping compilers
Ian Lynagh [Sat, 16 Jul 2011 08:12:11 +0000 (09:12 +0100)] 
Fix build with older bootstrapping compilers

8 years agoAdd --resume support to sync-all; trac #5119
Ian Lynagh [Fri, 15 Jul 2011 21:14:33 +0000 (22:14 +0100)] 
Add --resume support to sync-all; trac #5119

Now after, for example, "./sync-all pull", if pulling in one repo fails
then you can restart from that point with "./sync-all --resume pull".

8 years agoSmall refactoring
Ian Lynagh [Fri, 15 Jul 2011 18:37:50 +0000 (19:37 +0100)] 
Small refactoring

8 years agoRemove more defaultTargetPlatform uses
Ian Lynagh [Fri, 15 Jul 2011 18:06:04 +0000 (19:06 +0100)] 
Remove more defaultTargetPlatform uses

8 years agoRemove more defaultTargetPlatform uses
Ian Lynagh [Fri, 15 Jul 2011 17:13:49 +0000 (18:13 +0100)] 
Remove more defaultTargetPlatform uses

8 years agoRemove some more defaultTargetPlatform uses
Ian Lynagh [Fri, 15 Jul 2011 16:42:24 +0000 (17:42 +0100)] 
Remove some more defaultTargetPlatform uses

8 years agoA bit more work to to keep the right in-scope set around
Simon Peyton Jones [Fri, 15 Jul 2011 16:14:10 +0000 (17:14 +0100)] 
A bit more work to to keep the right in-scope set around

This change gets rid of a debug WARNING; it wasn't
actually making anything go wrong, but it's best to
have the correct in-scope set.

8 years agoA bit of trace refactoring
Simon Peyton Jones [Fri, 15 Jul 2011 16:13:15 +0000 (17:13 +0100)] 
A bit of trace refactoring

8 years agoAn example of failure:
Sergei Trofimovich [Fri, 15 Jul 2011 15:46:37 +0000 (16:46 +0100)] 
An example of failure:
"inplace/bin/ghc-stage1"   -H32m -O    -package-name base-4.4.0.0 -hide-all-packages -i -ilibraries/base/. -ilibraries/base/dist-install/build -ilibraries/base/dist-install/build/autogen -Ilibraries/base/dist-install/build -Ilibraries/base/dist-install/build/autogen -Ilibraries/base/include   -optP-DOPTIMISE_INTEGER_GCD_LCM -optP-include -optPlibraries/base/dist-install/build/autogen/cabal_macros.h -package ghc-prim-0.2.0.0 -package integer-gmp-0.3.0.0 -package rts-1.0 -split-objs -package-name base -XHaskell98 -XCPP -O2 -no-user-package-conf -rtsopts     -odir libraries/base/dist-install/build -hidir libraries/base/dist-install/build -stubdir libraries/base/dist-install/build -hisuf hi -osuf  o -hcsuf hc -c libraries/base/./GHC/Int.hs -o libraries/base/dist-install/build/GHC/Int.o
/usr/lib/gcc/sparc-unknown-linux-gnu/4.5.2/../../../../sparc-unknown-linux-gnu/bin/ld: --relax and -r may not be used together
collect2: ld returned 1 exit status
make[1]: *** [libraries/base/dist-install/build/GHC/Int.o] Error 1

Or with '-v' param:
/usr/bin/gcc -fno-stack-protector -nostdlib -nodefaultlibs -Wl,-r -Wl,--build-id=none -Wl,-x -o libraries/base/dist-install/build/GHC/Int.o /tmp/ghc45126_0/ghc45126_0.ldscript
/usr/lib/gcc/sparc-unknown-linux-gnu/4.5.2/../../../../sparc-unknown-linux-gnu/bin/ld: --relax and -r may not be used together

On SPARC gcc-4.5.2 has -mrelax in it's default specs, so we explicitly pass --no-relax
for ld to make sure gcc's default does not affect us.

Signed-off-by: Sergei Trofimovich <slyfox@gentoo.org>
8 years agoFix error reporting for overlapping instances in type checker
Simon Peyton Jones [Fri, 15 Jul 2011 13:46:13 +0000 (14:46 +0100)] 
Fix error reporting for overlapping instances in type checker

See #5320 and discussion there.  This patch only removes
a bogus assertion failure and refactors the code slightly.
There's still an underlying delicate point, described in #5320.

8 years agoImprove pretty printing of infix HsOpApp (fixes #5318)
Simon Peyton Jones [Fri, 15 Jul 2011 13:44:44 +0000 (14:44 +0100)] 
Improve pretty printing of infix HsOpApp (fixes #5318)

8 years agoRemove some defaultTargetPlatform's
Ian Lynagh [Fri, 15 Jul 2011 12:16:50 +0000 (13:16 +0100)] 
Remove some defaultTargetPlatform's

We now pass the correct platform in instead

8 years agoImprove pretty printing of Core (fixes #5325)
Simon Peyton Jones [Fri, 15 Jul 2011 11:08:43 +0000 (12:08 +0100)] 
Improve pretty printing of Core (fixes #5325)

8 years agoSpelling error in comment
Simon Peyton Jones [Fri, 15 Jul 2011 11:08:28 +0000 (12:08 +0100)] 
Spelling error in comment

8 years agoUse varToCoreExpr rather than Var (fixes #5315)
Simon Peyton Jones [Fri, 15 Jul 2011 10:49:38 +0000 (11:49 +0100)] 
Use varToCoreExpr rather than Var (fixes #5315)

A coercion variable in a term must be wrapped in a coercion!
(Led to lint errors.)

8 years agoMore work towards cross-compilation
Ian Lynagh [Fri, 15 Jul 2011 01:26:24 +0000 (02:26 +0100)] 
More work towards cross-compilation

There's now a variant of the Outputable class that knows what
platform we're targetting:

class PlatformOutputable a where
    pprPlatform :: Platform -> a -> SDoc
    pprPlatformPrec :: Platform -> Rational -> a -> SDoc

and various instances have had to be converted to use that class,
and we pass Platform around accordingly.

8 years agoImprovements to Safe Haskell userguide.
David Terei [Fri, 15 Jul 2011 01:13:26 +0000 (18:13 -0700)] 
Improvements to Safe Haskell userguide.

8 years agoHandle II8 Size value in PowerPC code generator.
Erik de Castro Lopo [Thu, 26 May 2011 08:38:11 +0000 (18:38 +1000)] 
Handle II8 Size value in PowerPC code generator.

8 years agoJust renaming a couple of functions
Ian Lynagh [Thu, 14 Jul 2011 19:15:11 +0000 (20:15 +0100)] 
Just renaming a couple of functions

8 years agoSeparate the warning flags into their own datatype
Ian Lynagh [Thu, 14 Jul 2011 18:59:13 +0000 (19:59 +0100)] 
Separate the warning flags into their own datatype

The -w flag wasn't turning off a few warnings (Opt_WarnMissingImportList,
Opt_WarnMissingLocalSigs, Opt_WarnIdentities). Rather than just adding
them, I've separated the Opt_Warn* contructors off into their own type,
so -w now just sets the list of warning flags to [].

8 years agoMore Lexer.x tidy-ups
Ian Lynagh [Thu, 14 Jul 2011 17:17:36 +0000 (18:17 +0100)] 
More Lexer.x tidy-ups

8 years agoRemove some unused, commented-out code from Lexer.x
Ian Lynagh [Thu, 14 Jul 2011 17:11:28 +0000 (18:11 +0100)] 
Remove some unused, commented-out code from Lexer.x

8 years agoWhitespace in Lexer.x
Ian Lynagh [Thu, 14 Jul 2011 17:03:26 +0000 (18:03 +0100)] 
Whitespace in Lexer.x

8 years agovalidate now checks that the testsuite exists first; trac #5089
Ian Lynagh [Thu, 14 Jul 2011 13:15:46 +0000 (14:15 +0100)] 
validate now checks that the testsuite exists first; trac #5089

8 years agoFix typo
Ian Lynagh [Wed, 13 Jul 2011 23:31:31 +0000 (00:31 +0100)] 
Fix typo

8 years agoFix validate with old versions of gcc
Ian Lynagh [Wed, 13 Jul 2011 14:33:07 +0000 (15:33 +0100)] 
Fix validate with old versions of gcc

Old versions don't understand -Wno-error=inline

8 years agoRemove 'threadsafe' FFI imports
Ian Lynagh [Tue, 12 Jul 2011 23:57:01 +0000 (00:57 +0100)] 
Remove 'threadsafe' FFI imports

They've been deprecated since GHC 6.12.

8 years agoWhitespace only in compiler/nativeGen/RegAlloc/Liveness.hs
Ian Lynagh [Tue, 12 Jul 2011 20:30:09 +0000 (21:30 +0100)] 
Whitespace only in compiler/nativeGen/RegAlloc/Liveness.hs

8 years agoWhitespace only in compiler/nativeGen/Instruction.hs
Ian Lynagh [Tue, 12 Jul 2011 18:33:26 +0000 (19:33 +0100)] 
Whitespace only in compiler/nativeGen/Instruction.hs

8 years agoMore CPP removal
Ian Lynagh [Tue, 12 Jul 2011 18:24:26 +0000 (19:24 +0100)] 
More CPP removal

8 years agoMore CPP removal
Ian Lynagh [Tue, 12 Jul 2011 17:42:01 +0000 (18:42 +0100)] 
More CPP removal

8 years agoCPP removal
Ian Lynagh [Tue, 12 Jul 2011 17:26:22 +0000 (18:26 +0100)] 
CPP removal

8 years agoRemove SafeLanguage from safe haskell
David Terei [Tue, 12 Jul 2011 20:00:11 +0000 (13:00 -0700)] 
Remove SafeLanguage from safe haskell

8 years agomake sure libraries/dph/ghc.mk is included after the other DPH ghc.mk
Simon Marlow [Tue, 12 Jul 2011 14:38:37 +0000 (15:38 +0100)] 
make sure libraries/dph/ghc.mk is included after the other DPH ghc.mk
files, so that it can refer to variables defined there.  (necessary,
but perhaps not sufficient, to fix the DPH build bugs)

8 years agoFix DLL/SO loading (see #5313).
Simon Marlow [Tue, 12 Jul 2011 11:16:17 +0000 (12:16 +0100)] 
Fix DLL/SO loading (see #5313).

The code in here is a bit of a mess.  I've fixed up some
inconsistencies I can see, but it could do with an overhaul.

8 years agoFix bug in parsing of module headers (see #5243)
Simon Marlow [Tue, 12 Jul 2011 10:48:00 +0000 (11:48 +0100)] 
Fix bug in parsing of module headers (see #5243)

8 years agoomit the local MD5 implementation if we can get it from
Simon Marlow [Tue, 12 Jul 2011 09:41:22 +0000 (10:41 +0100)] 
omit the local MD5 implementation if we can get it from
GHC.Fingerprint in base instead.

8 years agoDebugger commands do not work with -fno-ghci-sandbox, so emit useful
Simon Marlow [Tue, 12 Jul 2011 09:40:50 +0000 (10:40 +0100)] 
Debugger commands do not work with -fno-ghci-sandbox, so emit useful
error messages.

8 years agoderive Typeable (eliminate deprecation warnings for mkTyCon)
Simon Marlow [Mon, 11 Jul 2011 13:30:39 +0000 (14:30 +0100)] 
derive Typeable (eliminate deprecation warnings for mkTyCon)

8 years agofix Binary instance for TyCon following Typeable implementation
Simon Marlow [Mon, 11 Jul 2011 13:06:01 +0000 (14:06 +0100)] 
fix Binary instance for TyCon following Typeable implementation
changes

8 years agothe time package uses mkTyCon, turn off deprecation warnings for now
Simon Marlow [Mon, 11 Jul 2011 12:24:40 +0000 (13:24 +0100)] 
the time package uses mkTyCon, turn off deprecation warnings for now

8 years agoHad to move some Typeable things around to avoid orphan instances.
Simon Marlow [Mon, 11 Jul 2011 12:00:20 +0000 (13:00 +0100)] 
Had to move some Typeable things around to avoid orphan instances.

8 years agoremove getOrSetTypeableStore. This is no longer used after the switch
Simon Marlow [Fri, 8 Jul 2011 09:47:57 +0000 (10:47 +0100)] 
remove getOrSetTypeableStore.  This is no longer used after the switch
to using MD5 hashes to identify TypeReps in the Typeable library.

8 years agoChange the code generated for deriving Typeable, to match the changes
Simon Marlow [Fri, 8 Jul 2011 09:47:19 +0000 (10:47 +0100)] 
Change the code generated for deriving Typeable, to match the changes
to the Typeable library.  We now generate an MD5 hash of the
fully-qualified TyCon name at compile time.

8 years agomove computeFingerprint from MkIface to Binary
Simon Marlow [Fri, 8 Jul 2011 09:42:13 +0000 (10:42 +0100)] 
move computeFingerprint from MkIface to Binary

8 years agoAdd 64-bit signed and unsigned integer literals to HsSyn.
Simon Marlow [Fri, 8 Jul 2011 09:41:22 +0000 (10:41 +0100)] 
Add 64-bit signed and unsigned integer literals to HsSyn.
No concrete syntax yet, but I need to be able to use these in code
generated for derived Typeable instances.

8 years agoMake an extension for interruptible FFI calls
Ian Lynagh [Mon, 11 Jul 2011 17:24:22 +0000 (18:24 +0100)] 
Make an extension for interruptible FFI calls

8 years agoFix "./sync-all get"
Ian Lynagh [Fri, 8 Jul 2011 13:48:06 +0000 (14:48 +0100)] 
Fix "./sync-all get"

8 years agoFix the build
Ian Lynagh [Fri, 8 Jul 2011 12:31:51 +0000 (13:31 +0100)] 
Fix the build

The seq# case in the new codegen was being shadowed by a more general
case.

8 years agoRemove old ./quickcheck/ stuff
Ian Lynagh [Fri, 8 Jul 2011 00:51:18 +0000 (01:51 +0100)] 
Remove old ./quickcheck/ stuff

8 years agoBump Cabal deps
Ian Lynagh [Thu, 7 Jul 2011 19:28:02 +0000 (20:28 +0100)] 
Bump Cabal deps

8 years agoRandom is now only needed to build dph
Ian Lynagh [Wed, 6 Jul 2011 19:30:58 +0000 (20:30 +0100)] 
Random is now only needed to build dph

8 years agoRelax some dependencies
Ian Lynagh [Wed, 6 Jul 2011 19:25:28 +0000 (20:25 +0100)] 
Relax some dependencies

8 years agoRemove mention of -fno-method-sharing from comments and docs
Ian Lynagh [Wed, 6 Jul 2011 11:25:27 +0000 (12:25 +0100)] 
Remove mention of -fno-method-sharing from comments and docs

8 years agoPort 'Add two new primops seq# and spark#' (be54417) to new codegen.
Edward Z. Yang [Fri, 8 Jul 2011 00:48:48 +0000 (20:48 -0400)] 
Port 'Add two new primops seq# and spark#' (be54417) to new codegen.

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>
8 years agoDon't export the _info symbol for the data constructor worker bindings
Max Bolingbroke [Wed, 6 Jul 2011 17:01:02 +0000 (18:01 +0100)] 
Don't export the _info symbol for the data constructor worker bindings

This is safe because GHC never generates a fast call to a data constructor
worker: if the call is seen statically it will be eta-expanded and the
allocation of the data will be inlined. We still need to export the _closure
in case the constructor is used in an unapplied fashion.

8 years agoRefactoring: explicitly mark whether we have an info table in RawCmm
Max Bolingbroke [Wed, 6 Jul 2011 15:18:09 +0000 (16:18 +0100)] 
Refactoring: explicitly mark whether we have an info table in RawCmm

I introduced this to support explicitly recording the info table label
in RawCmm for another patch I am working on, but it turned out to lead
to significant simplification in those parts of the compiler that
consume RawCmm.

Now, instead of lots of tests for null [CmmStatic] we have a simple
test of a Maybe, and have reduced the number of guys that need to know
how to convert entry->info labels by a TON. There are only 3 callers
of that function now!

8 years agofix bug: check_out_flag should be checked_out_tree
Simon Marlow [Wed, 6 Jul 2011 08:47:02 +0000 (09:47 +0100)] 
fix bug: check_out_flag should be checked_out_tree

8 years agoNew functionality for sync-all to support bare repos, mirroring the main repo layout.
George Roldugin [Fri, 1 Jul 2011 06:07:54 +0000 (16:07 +1000)] 
New functionality for sync-all to support bare repos, mirroring the main repo layout.

Signed-off-by: George Roldugin <groldugin@cse.unsw.edu.au>
8 years agoTypo
Austin Seipp [Mon, 4 Jul 2011 18:21:40 +0000 (13:21 -0500)] 
Typo

8 years agoAdd documentation for compiler plugins
Austin Seipp [Mon, 4 Jul 2011 17:41:50 +0000 (12:41 -0500)] 
Add documentation for compiler plugins

8 years agoFix typo in Safe Haskell doc
David Terei [Wed, 6 Jul 2011 08:17:12 +0000 (01:17 -0700)] 
Fix typo in Safe Haskell doc

8 years agoSome general code cleaning in LLVM backend
David Terei [Wed, 6 Jul 2011 08:16:16 +0000 (01:16 -0700)] 
Some general code cleaning in LLVM backend

8 years agoFix printing of llvm IR to work with llvm-3.0
David Terei [Wed, 6 Jul 2011 06:55:10 +0000 (23:55 -0700)] 
Fix printing of llvm IR to work with llvm-3.0

8 years agoMerge branch 'master' of http://darcs.haskell.org/ghc
Ian Lynagh [Tue, 5 Jul 2011 19:41:09 +0000 (20:41 +0100)] 
Merge branch 'master' of darcs.haskell.org/ghc

8 years agoDon't export SRT labels
Max Bolingbroke [Mon, 4 Jul 2011 16:44:43 +0000 (17:44 +0100)] 
Don't export SRT labels

8 years agoRemove the unused CmmAlign and CmmDataLabel from CmmStatic
Max Bolingbroke [Tue, 5 Jul 2011 08:31:08 +0000 (09:31 +0100)] 
Remove the unused CmmAlign and CmmDataLabel from CmmStatic

8 years agoRefactoring: use a structured CmmStatics type rather than [CmmStatic]
Max Bolingbroke [Tue, 5 Jul 2011 08:23:58 +0000 (09:23 +0100)] 
Refactoring: use a structured CmmStatics type rather than [CmmStatic]

I observed that the [CmmStatics] within CmmData uses the list in a very stylised way.
The first item in the list is almost invariably a CmmDataLabel. Many parts of the
compiler pattern match on this list and fail if this is not true.

This patch makes the invariant explicit by introducing a structured type CmmStatics
that holds the label and the list of remaining [CmmStatic].

There is one wrinkle: the x86 backend sometimes wants to output an alignment directive just
before the label. However, this can be easily fixed up by parameterising the native codegen
over the type of CmmStatics (though the GenCmmTop parameterisation) and using a pair
(Alignment, CmmStatics) there instead.

As a result, I think we will be able to remove CmmAlign and CmmDataLabel from the CmmStatic
data type, thus nuking a lot of code and failing pattern matches. This change will come as part
of my next patch.

8 years agoRemove -fmethod-sharing
Ian Lynagh [Sun, 3 Jul 2011 20:07:31 +0000 (21:07 +0100)] 
Remove -fmethod-sharing

It was already deprecated, and marked for removal in 7.2.

8 years agoRemove the Opt_Generics constructor
Ian Lynagh [Sun, 3 Jul 2011 18:53:04 +0000 (19:53 +0100)] 
Remove the Opt_Generics constructor

We no longer advertise that we support the Generics extension,
although we still do accept -XGenerics and -XNoGenerics flags
(but warn that we are ignoring them).

I also remove the even older -fgenerics and -fno-generics flags.

8 years ago-optm is now deprecated
Ian Lynagh [Sun, 3 Jul 2011 15:58:03 +0000 (16:58 +0100)] 
-optm is now deprecated

8 years agoFix typos
Ian Lynagh [Sun, 3 Jul 2011 15:52:21 +0000 (16:52 +0100)] 
Fix typos

8 years agodefaultErrorHandler now only takes LogAction
Ian Lynagh [Sun, 3 Jul 2011 02:11:32 +0000 (03:11 +0100)] 
defaultErrorHandler now only takes LogAction

It used to take a whole DynFlags, but that meant we had to
create a DynFlags with (panic "No settings") for settings, as
we didn't have any real settings.

Now we just pass the LogAction, which is all that it actually needed.
The default is exported from DynFlags as defaultLogAction.

8 years agoFix building docs
Ian Lynagh [Sun, 3 Jul 2011 02:11:13 +0000 (03:11 +0100)] 
Fix building docs

8 years agoFix a copy-and-paste-o
Ian Lynagh [Sat, 2 Jul 2011 20:04:43 +0000 (21:04 +0100)] 
Fix a copy-and-paste-o

8 years agoRemove old release note files
Ian Lynagh [Sat, 2 Jul 2011 17:08:56 +0000 (18:08 +0100)] 
Remove old release note files

8 years agoFlip fplugin options to restore hasarg parsing
Max Bolingbroke [Thu, 30 Jun 2011 11:33:55 +0000 (12:33 +0100)] 
Flip fplugin options to restore hasarg parsing

8 years agoBump version number to 7.3
Ian Lynagh [Fri, 1 Jul 2011 21:38:19 +0000 (22:38 +0100)] 
Bump version number to 7.3

8 years agoRemove outdated paragraph from docs
Ian Lynagh [Fri, 1 Jul 2011 14:25:14 +0000 (15:25 +0100)] 
Remove outdated paragraph from docs

8 years agoRemove some conditional CPP from DriverPipeline
Ian Lynagh [Thu, 30 Jun 2011 01:30:01 +0000 (02:30 +0100)] 
Remove some conditional CPP from DriverPipeline

8 years agoRemove conditional CPP in DsForeign
Ian Lynagh [Thu, 30 Jun 2011 01:13:12 +0000 (02:13 +0100)] 
Remove conditional CPP in DsForeign

8 years agoupdate freeHaskellFuncationPtr following changes to x86 adjustors
Simon Marlow [Thu, 30 Jun 2011 14:50:19 +0000 (15:50 +0100)] 
update freeHaskellFuncationPtr following changes to x86 adjustors
(fixes various ffi test failures on x86.  This change was supposed to
be part of 9f61598ce7b0cb3448e8f0c3d627c0ca47b7f55f, but somehow it
got lost).

8 years agosupport using local mirrors (created by git clone --mirror)
Simon Marlow [Thu, 30 Jun 2011 14:30:21 +0000 (15:30 +0100)] 
support using local mirrors (created by git clone --mirror)

8 years agodoc tweak: mention that UNPACK doesn't work without -O
Simon Marlow [Tue, 28 Jun 2011 12:42:00 +0000 (13:42 +0100)] 
doc tweak: mention that UNPACK doesn't work without -O

8 years agoBetter heap profiling for pinned objects (e.g. ByteStrings).
Simon Marlow [Wed, 15 Jun 2011 15:40:16 +0000 (16:40 +0100)] 
Better heap profiling for pinned objects (e.g. ByteStrings).
Previously we were completely ignoring these, due to the difficulties
of traversing the pinned blocks (the objects are not necessarily
end-to-end, we can't tell how large the gaps are).  Now just count the
whole block as a big ARR_WORDS, so at least we're accounting for the
memory and it has the right type.