ghc.git
6 years agoadded -flate-float-leave-LNE, which prevents LNEs from being lifted late-lam-lift
Nicolas Frisby [Thu, 11 Apr 2013 11:07:45 +0000 (12:07 +0100)] 
added -flate-float-leave-LNE, which prevents LNEs from being lifted

  ALLOC WORSE                                      llf    llf -flate-float-leave-LNE
            boyer2              3901736           -7.0%           +0.0%
            puzzle            165864064          -19.1%           -2.9%

  TIME BETTER
    fannkuch-redux                 6.55          +11.7%           -0.1%
         --  382807736 ENT_LNE_ctr      <-- likely part of the improvement
             fasta                 1.11           +1.2%           +0.0%
              life                 0.43           +3.8%           -0.8%
               scs                 0.99           -0.3%           -4.7%
     spectral-norm                 3.25           +2.8%           +0.0%
              wang                 0.22           +0.0%           -3.0%

Also causes slowdowns, but I don't immediately see why

   TIME WORSE
      binary-trees                 1.67           +1.4%           +5.0%
            hidden                 0.77           -3.9%           +0.4%
           integer                 2.78           -4.7%           -0.7%
         typecheck                 0.36           -5.6%           +2.8%

I looked at typecheck's ticky and CorePrep changes and didn't see
anything indicative. Ugh.

6 years agofixed bug in fast-call detection logic; added: do not check for bad space if all...
Nicolas Frisby [Tue, 9 Apr 2013 18:27:35 +0000 (19:27 +0100)] 
fixed bug in fast-call detection logic; added: do not check for bad space if all lambdas are oneshot

6 years agoI think the CorePrep float approximation is correct and the cg is precise
Nicolas Frisby [Tue, 9 Apr 2013 15:09:55 +0000 (16:09 +0100)] 
I think the CorePrep float approximation is correct and the cg is precise

6 years agoadded -flate-float-singly-SAT-fallback and the corresponding once analysis; now worki...
Nicolas Frisby [Mon, 8 Apr 2013 16:20:41 +0000 (17:20 +0100)] 
added -flate-float-singly-SAT-fallback and the corresponding once analysis; now working on precise CG in lam

6 years agosimplifications; fixed bug in LNE detection; tracking one shot lambdas; fixed bug...
Nicolas Frisby [Mon, 8 Apr 2013 01:16:44 +0000 (02:16 +0100)] 
simplifications; fixed bug in LNE detection; tracking one shot lambdas; fixed bug in CG-in-lam calculation

6 years agoenriching the LLF to predict CorePrep floating
Nicolas Frisby [Sun, 7 Apr 2013 14:48:48 +0000 (15:48 +0100)] 
enriching the LLF to predict CorePrep floating

6 years agoMerge branch 'late-lambda-float'
Nicolas Frisby [Thu, 4 Apr 2013 21:21:58 +0000 (22:21 +0100)] 
Merge branch 'late-lambda-float'

6 years agoadded ticky counters for heap and stack checks
Nicolas Frisby [Thu, 4 Apr 2013 21:20:27 +0000 (22:20 +0100)] 
added ticky counters for heap and stack checks

6 years agoMerge branch 'master' of darcs.haskell.org:/home/darcs/ghc
Simon Peyton Jones [Wed, 3 Apr 2013 17:20:23 +0000 (18:20 +0100)] 
Merge branch 'master' of darcs.haskell.org:/home/darcs/ghc

6 years agoFix Trac #7804, about floating equalites
Simon Peyton Jones [Wed, 3 Apr 2013 17:20:07 +0000 (18:20 +0100)] 
Fix Trac #7804, about floating equalites

We float unsolved equalities from underneath a 'forall', to
help solve them, in TcSimplify.floatEqualities.

It's regrettably delicate though,as this bug shows. I'm not
happy with the new code; but there are copious notes; see
Note [Float equalities from under a skolem binding].

6 years agoFix Trac #7805: don't allow nested foralls in promoted types
Simon Peyton Jones [Wed, 3 Apr 2013 13:38:14 +0000 (14:38 +0100)] 
Fix Trac #7805: don't allow nested foralls in promoted types

6 years agoThis changes fixes a bad error in canonicalisation, concerning kind equality
Simon Peyton Jones [Wed, 3 Apr 2013 13:37:50 +0000 (14:37 +0100)] 
This changes fixes a bad error in canonicalisation, concerning kind equality

We care careful not to construct a canonical equality whose LHS and RHS
have incompatible kinds.  (This is one of the invariants of a canonical
equality.)  See Note [Equalities with incompatible kinds].  However,
what I had not dealt with is when LHS and RHS *look* as if they have
different kinds, but after zonking they become the same.  Bad!

(This led to an ASSERT failure in the test for Trac #7176.)

6 years agoIn build.mk.sample, include v in GhcLibWays even if DYNAMIC_GHC_PROGRAMS is set
Ian Lynagh [Wed, 3 Apr 2013 12:01:24 +0000 (13:01 +0100)] 
In build.mk.sample, include v in GhcLibWays even if DYNAMIC_GHC_PROGRAMS is set

Although it's not technically needed, it's less confusing if the vanilla
way works when people use the sample build.mk.

6 years agoFix installation
Ian Lynagh [Wed, 3 Apr 2013 11:47:36 +0000 (12:47 +0100)] 
Fix installation

The build system thought that the RTS built more library files than
it actually did, and installation failed when we tried to 'strip'
one of these non-existant files.

6 years agoMake sure README.md mentions that make can build in parallel.
Austin Seipp [Fri, 29 Mar 2013 19:47:42 +0000 (14:47 -0500)] 
Make sure README.md mentions that make can build in parallel.

Signed-off-by: Austin Seipp <aseipp@pobox.com>
6 years agoRevert "Make sure README.md mentions that make can build in parallel."
Austin Seipp [Fri, 29 Mar 2013 19:44:56 +0000 (14:44 -0500)] 
Revert "Make sure README.md mentions that make can build in parallel."

This reverts commit 0b72347dc468ba081d67385a077238bcaa9eb4d5.

6 years agoRevert "Wibble."
Austin Seipp [Fri, 29 Mar 2013 19:44:45 +0000 (14:44 -0500)] 
Revert "Wibble."

This reverts commit e7ca132466e3c2e71f7c8fdb2b96ec87c24353be.

6 years agoWibble.
Austin Seipp [Fri, 29 Mar 2013 19:38:40 +0000 (14:38 -0500)] 
Wibble.

Signed-off-by: Austin Seipp <aseipp@pobox.com>
6 years agoMake sure README.md mentions that make can build in parallel.
Austin Seipp [Fri, 29 Mar 2013 19:37:31 +0000 (14:37 -0500)] 
Make sure README.md mentions that make can build in parallel.

Signed-off-by: Austin Seipp <aseipp@pobox.com>
6 years agoticky enhancements
Nicolas Frisby [Wed, 6 Mar 2013 21:46:14 +0000 (21:46 +0000)] 
ticky enhancements

  * the new StgCmmArgRep module breaks a dependency cycle; I also
    untabified it, but made no real changes

  * updated the documentation in the wiki and change the user guide to
    point there

  * moved the allocation enters for ticky and CCS to after the heap check

    * I left LDV where it was, which was before the heap check at least
      once, since I have no idea what it is

  * standardized all (active?) ticky alloc totals to bytes

  * in order to avoid double counting StgCmmLayout.adjustHpBackwards
    no longer bumps ALLOC_HEAP_ctr

  * I resurrected the SLOW_CALL counters

    * the new module StgCmmArgRep breaks cyclic dependency between
      Layout and Ticky (which the SLOW_CALL counters cause)

    * renamed them SLOW_CALL_fast_<pattern> and VERY_SLOW_CALL

  * added ALLOC_RTS_ctr and _tot ticky counters

    * eg allocation by Storage.c:allocate or a BUILD_PAP in stg_ap_*_info

    * resurrected ticky counters for ALLOC_THK, ALLOC_PAP, and
      ALLOC_PRIM

    * added -ticky and -DTICKY_TICKY in ways.mk for debug ways

  * added a ticky counter for total LNE entries

  * new flags for ticky: -ticky-allocd -ticky-dyn-thunk -ticky-LNE

    * all off by default

    * -ticky-allocd: tracks allocation *of* closure in addition to
       allocation *by* that closure

    * -ticky-dyn-thunk tracks dynamic thunks as if they were functions

    * -ticky-LNE tracks LNEs as if they were functions

  * updated the ticky report format, including making the argument
    categories (more?) accurate again

  * the printed name for things in the report include the unique of
    their ticky parent as well as if they are not top-level

6 years agoimprove dead code elimination in CorePrep (fixes #7796)
Nicolas Frisby [Wed, 27 Mar 2013 20:25:28 +0000 (20:25 +0000)] 
improve dead code elimination in CorePrep (fixes #7796)

6 years agoFix non-termination of SpecConstr (see #5550).
Amos Robinson [Thu, 28 Mar 2013 01:37:42 +0000 (12:37 +1100)] 
Fix non-termination of SpecConstr (see #5550).
ForceSpecConstr will now only specialise recursive types a finite number of times.
There is a new option -fspec-constr-recursive, with a default value of 3.

6 years agoImprove comments about dead code (thanks to Nick Frisby)
Simon Peyton Jones [Wed, 27 Mar 2013 14:24:51 +0000 (14:24 +0000)] 
Improve comments about dead code (thanks to Nick Frisby)

6 years agoComment only
Simon Peyton Jones [Wed, 27 Mar 2013 14:22:35 +0000 (14:22 +0000)] 
Comment only

6 years agoComment typo
Simon Peyton Jones [Wed, 13 Mar 2013 14:09:11 +0000 (14:09 +0000)] 
Comment typo

6 years agoHeap profiling: flush .hp file at the end of each frame
Takano Akio [Fri, 15 Feb 2013 11:06:38 +0000 (20:06 +0900)] 
Heap profiling: flush .hp file at the end of each frame

6 years agoAllow to bootstrap with a compiler from the FreeBSD Ports Collection
Gabor Pali [Sun, 24 Mar 2013 21:29:30 +0000 (22:29 +0100)] 
Allow to bootstrap with a compiler from the FreeBSD Ports Collection

6 years agomissed a bunch of places in my flag changes
Nicolas Frisby [Sun, 24 Mar 2013 18:49:19 +0000 (18:49 +0000)] 
missed a bunch of places in my flag changes

6 years agowas not ignoring tyco arguments when determing argument strictness
Nicolas Frisby [Thu, 21 Mar 2013 18:52:25 +0000 (18:52 +0000)] 
was not ignoring tyco arguments when determing argument strictness

6 years agofixed conflict in simpleCore pipeline
Nicolas Frisby [Wed, 20 Mar 2013 16:56:45 +0000 (16:56 +0000)] 
fixed conflict in simpleCore pipeline

6 years agoadded flags for the LLF's use of strictness info and its escape analysis
Nicolas Frisby [Sun, 17 Mar 2013 02:15:53 +0000 (02:15 +0000)] 
added flags for the LLF's use of strictness info and its escape analysis

6 years agoadded -fno-late-float-abstract-LNE-var flag to disallow abstracting over LNE variables
Nicolas Frisby [Sun, 17 Mar 2013 01:27:31 +0000 (01:27 +0000)] 
added -fno-late-float-abstract-LNE-var flag to disallow abstracting over LNE variables

6 years agonow identifying LNEs and not considering them as closures (ie captors)
Nicolas Frisby [Sun, 17 Mar 2013 00:38:21 +0000 (00:38 +0000)] 
now identifying LNEs and not considering them as closures (ie captors)

6 years agopreparation in SetLevels for identifying LNEs -- confirmed that it did not change...
Nicolas Frisby [Sat, 16 Mar 2013 15:36:56 +0000 (15:36 +0000)] 
preparation in SetLevels for identifying LNEs -- confirmed that it did not change allocation wrt last night's baseline nor cg0pin

6 years agocleaned-up flags to stop confusing myself
Nicolas Frisby [Thu, 14 Mar 2013 15:18:56 +0000 (15:18 +0000)] 
cleaned-up flags to stop confusing myself

SetLevels now uses strictess information to determine if arguments
will become closures, this enables the -8% in cichelli and helps a bit
in anna too

I think the captors-interpretation plan is very nearly precise except
for LNEs now

I added the late-float-retry flag, it enables a rough draft in
SetLevels that attempts to unpin vacuous cycles involving closure
growth and fast call preservation

6 years agototal rewrite of the in-closure analysis: using the interpreted Captors data type...
Nicolas Frisby [Wed, 13 Mar 2013 22:27:34 +0000 (22:27 +0000)] 
total rewrite of the in-closure analysis: using the interpreted Captors data type for now

6 years agoclean-up in SetLevels, better debug output, tracking the names of thunks that a bindi...
Nicolas Frisby [Wed, 13 Mar 2013 15:12:37 +0000 (15:12 +0000)] 
clean-up in SetLevels, better debug output, tracking the names of thunks that a binding occurs in; next: do not penalize for those thunks if they have been floated!

6 years agoFix the names of the libffi archives
Ian Lynagh [Sun, 24 Mar 2013 17:29:30 +0000 (17:29 +0000)] 
Fix the names of the libffi archives

Fixes a bug which meant that they weren't getting installed.

6 years agoChange how we handle libffi
Ian Lynagh [Sat, 23 Mar 2013 23:55:56 +0000 (23:55 +0000)] 
Change how we handle libffi

I think overall the new approach is simpler. Rather than unpacking
the libffi.a and putting the .o files into our libHSrts.a, we just
use the libffi.a.

This change also means that when compiling programs for the dyn
way, they get explicitly linked against libffi.so (rather than
relying on librts.so being linked against it). This might
fix a problem on FreeBSD, where programs cannot find libffi.so.

6 years agoSmall configure.ac refactoring
Ian Lynagh [Sat, 23 Mar 2013 21:54:53 +0000 (21:54 +0000)] 
Small configure.ac refactoring

6 years agoTweak the _HC_OPTS variables
Ian Lynagh [Sat, 23 Mar 2013 11:56:46 +0000 (11:56 +0000)] 
Tweak the _HC_OPTS variables

We no longer pass -hisuf/-osuf flags to "ghc -M". Doing so didn't really
make sense with the way the -dep-suffix flags now work.

6 years agoFix the handling of ways, and in particular Opt_Static and Opt_SplitObjs
Ian Lynagh [Fri, 22 Mar 2013 17:15:51 +0000 (17:15 +0000)] 
Fix the handling of ways, and in particular Opt_Static and Opt_SplitObjs

6 years agoTurn object splitting off when we use the dynamic way
Ian Lynagh [Fri, 22 Mar 2013 14:08:55 +0000 (14:08 +0000)] 
Turn object splitting off when we use the dynamic way

There's no point splitting objects when we're going to be dynamically
linking. Plus it breaks compilation on OSX x86.

6 years agoRemove some more old, unused code
Ian Lynagh [Thu, 21 Mar 2013 21:25:42 +0000 (21:25 +0000)] 
Remove some more old, unused code

6 years agoTrack subrepos Cabal
Gabor Greif [Wed, 20 Mar 2013 21:37:43 +0000 (22:37 +0100)] 
Track subrepos Cabal

6 years agoFix build with non-Linux ELF OSes
Ian Lynagh [Wed, 20 Mar 2013 19:25:27 +0000 (19:25 +0000)] 
Fix build with non-Linux ELF OSes

We were only setting an RPATH for the RTS DLL on Linux, but as far
as I can see we should be doing it for all ELF OSes. Hopefully this
will fix the problem where the installed ghc-pkg can't find libffi.dll
on FreeBSD.

6 years agoRemove some commented out build system code
Ian Lynagh [Tue, 19 Mar 2013 21:47:55 +0000 (21:47 +0000)] 
Remove some commented out build system code

6 years agoFix building when $(CC_STAGE0) contains spaces
Ian Lynagh [Tue, 19 Mar 2013 15:34:50 +0000 (15:34 +0000)] 
Fix building when $(CC_STAGE0) contains spaces

Spotted by Bill Tutt

6 years agoBe more optimistic about which platforms support shared libraries
Ian Lynagh [Tue, 19 Mar 2013 15:25:17 +0000 (15:25 +0000)] 
Be more optimistic about which platforms support shared libraries

Pretty much everything does now, so list those that don't instead
instead of listing those that do.

6 years agoOptimistically assume that LD_LIBRARY_PATH works for all platforms
Ian Lynagh [Tue, 19 Mar 2013 00:46:15 +0000 (00:46 +0000)] 
Optimistically assume that LD_LIBRARY_PATH works for all platforms

other than Windows and OS X.

It's known to work on Linux and FreeBSD.

6 years agoMerge branch 'master' of darcs.haskell.org:/srv/darcs//ghc
Ian Lynagh [Tue, 19 Mar 2013 00:41:22 +0000 (00:41 +0000)] 
Merge branch 'master' of darcs.haskell.org:/srv/darcs//ghc

6 years agoMerge branch 'master' of mac:ghc/git/val64/.
Ian Lynagh [Tue, 19 Mar 2013 00:37:04 +0000 (00:37 +0000)] 
Merge branch 'master' of mac:ghc/git/val64/.

6 years agoAdd support for OSX ld's -filelist flag
Ian Lynagh [Mon, 18 Mar 2013 22:48:43 +0000 (22:48 +0000)] 
Add support for OSX ld's -filelist flag

Without it, when linking the split objects for Language.Haskell.TH.Syntax,
the commandline was too long when listing all the files directly.

6 years agoFix stage2 build on ARM.
Austin Seipp [Mon, 18 Mar 2013 22:19:59 +0000 (22:19 +0000)] 
Fix stage2 build on ARM.

LLVM supports PIC on ARM from what I can see.

Signed-off-by: Austin Seipp <aseipp@pobox.com>
6 years agoRemove base 3 support from bin-package-db
Ian Lynagh [Mon, 18 Mar 2013 12:33:10 +0000 (12:33 +0000)] 
Remove base 3 support from bin-package-db

6 years agoOnly steal ~# as a reserved operator when MagicHash is enabled; fixes #7776
Ian Lynagh [Mon, 18 Mar 2013 10:12:57 +0000 (10:12 +0000)] 
Only steal ~# as a reserved operator when MagicHash is enabled; fixes #7776

6 years agoAdd a kludgy dependency to fix compiling modules that use annotations
Ian Lynagh [Sun, 17 Mar 2013 17:23:10 +0000 (17:23 +0000)] 
Add a kludgy dependency to fix compiling modules that use annotations

We now make the stage 2 compiler depend on GHC.Desugar.

6 years agoFix ohi-sanity-check when we have ways other than v and dyn enabled
Ian Lynagh [Sun, 17 Mar 2013 15:42:20 +0000 (15:42 +0000)] 
Fix ohi-sanity-check when we have ways other than v and dyn enabled

It thought that /all/ ways should create the dyn o/hi files, but only
the v way does.

6 years agoRemove a couple of unused make variables
Ian Lynagh [Sun, 17 Mar 2013 14:03:07 +0000 (14:03 +0000)] 
Remove a couple of unused make variables

6 years agoRemove some directories that used to be used by GUM
Ian Lynagh [Sun, 17 Mar 2013 13:56:27 +0000 (13:56 +0000)] 
Remove some directories that used to be used by GUM

This hasn't been used for some time

6 years agoAlso build the v way when DYNAMIC_GHC_PROGRAMS is YES
Ian Lynagh [Sun, 17 Mar 2013 13:53:42 +0000 (13:53 +0000)] 
Also build the v way when DYNAMIC_GHC_PROGRAMS is YES

Technically we don't need the v way, but with -dynamic-too it's cheap,
and having it makes life easier.

6 years agoEnable Opt_PIC for the dyn way regardless of OSes.
PHO [Wed, 13 Mar 2013 15:22:34 +0000 (00:22 +0900)] 
Enable Opt_PIC for the dyn way regardless of OSes.

We were previously enabling Opt_PIC on certain OSes namely Windows,
Darwin and Linux, but it should always be enabled for the dyn way
because it makes no sense to try to generate non-PIC dynamic
libraries, which is usually impossible.

6 years agoRemove a bitrotted hack for OpenBSD and NetBSD regarding the dyn way.
PHO [Thu, 14 Mar 2013 07:13:45 +0000 (16:13 +0900)] 
Remove a bitrotted hack for OpenBSD and NetBSD regarding the dyn way.

On OpenBSD and NetBSD, we were accidentally passing "-optl-pthread" to
the gcc to build dynamic libraries because of a bitrotted hack for
those OSes, which resulted in a weird situation where every dynamic
library is named "ptl-pthread" and placed at "$(TOP)".

The hack in question were to work around a linkage problem with
libHSffi, but we no longer build it and just use libffi these days so
the hack can safely be removed.

6 years agoRemove readIface's unused argument
Ian Lynagh [Sun, 17 Mar 2013 01:28:49 +0000 (01:28 +0000)] 
Remove readIface's unused argument

6 years agoUpdate Cabal
Ian Lynagh [Sun, 17 Mar 2013 00:37:32 +0000 (00:37 +0000)] 
Update Cabal

6 years agoMerge branch 'master' of darcs.haskell.org:/srv/darcs//ghc
Ian Lynagh [Sat, 16 Mar 2013 22:58:02 +0000 (22:58 +0000)] 
Merge branch 'master' of darcs.haskell.org:/srv/darcs//ghc

6 years agoFix the handling of Opt_Static
Ian Lynagh [Sat, 16 Mar 2013 21:35:29 +0000 (21:35 +0000)] 
Fix the handling of Opt_Static

There were some cases where we weren't unsetting it when turning the
Dyn way on.

6 years agoFix searching for object files when doing TH
Ian Lynagh [Sat, 16 Mar 2013 20:03:25 +0000 (20:03 +0000)] 
Fix searching for object files when doing TH

We were finding vanilla object files when TH needed dynamic object files.

6 years agoA little polishing
Ian Lynagh [Sat, 16 Mar 2013 19:15:05 +0000 (19:15 +0000)] 
A little polishing

6 years agoDisable DYNAMIC_GHC_PROGRAMS on Windows
Ian Lynagh [Fri, 15 Mar 2013 23:11:59 +0000 (00:11 +0100)] 
Disable DYNAMIC_GHC_PROGRAMS on Windows

It doesn't work yet

6 years agoPartial support for dynamic ghc on Windows
Ian Lynagh [Fri, 15 Mar 2013 23:07:50 +0000 (00:07 +0100)] 
Partial support for dynamic ghc on Windows

6 years agoImplement type family instance support for ":info" (#4175)
Patrick Palka [Thu, 7 Mar 2013 16:39:10 +0000 (11:39 -0500)] 
Implement type family instance support for ":info" (#4175)

v2: added a couple of comments

6 years agoSet the way to 'dynamic' when running GHCi if GHCi is dynamically linked
Ian Lynagh [Thu, 14 Mar 2013 20:05:42 +0000 (20:05 +0000)] 
Set the way to 'dynamic' when running GHCi if GHCi is dynamically linked

6 years agoBy default, use the dynamic way for programs in the GHC tree
Ian Lynagh [Wed, 13 Mar 2013 21:29:27 +0000 (21:29 +0000)] 
By default, use the dynamic way for programs in the GHC tree

In particular, this means that GHCi will use DLLs, rather than loading
object files itself.

6 years agoImplement nullary type classes (#7642)
Simon Peyton Jones [Wed, 13 Mar 2013 21:10:20 +0000 (21:10 +0000)] 
Implement nullary type classes (#7642)

This is a slightly refined version of a patch by shachaf,
done by Krzysztof Gogolewski <krz.gogolewski@gmail.com>.

6 years agoRemove tabs and trailing whitespace from TcTyClsDecls
Krzysztof Gogolewski [Sun, 10 Mar 2013 21:53:33 +0000 (22:53 +0100)] 
Remove tabs and trailing whitespace from TcTyClsDecls

6 years agoFinish adding support for 2 DLLs in the ghc package; fixes #5987
Ian Lynagh [Tue, 12 Mar 2013 21:12:38 +0000 (22:12 +0100)] 
Finish adding support for 2 DLLs in the ghc package; fixes #5987

6 years agoFollow changes in Cabal
Ian Lynagh [Tue, 12 Mar 2013 18:21:55 +0000 (18:21 +0000)] 
Follow changes in Cabal

6 years agoMerge branch 'master' of win:c:/ghc/git/cygwin/.
Ian Lynagh [Tue, 12 Mar 2013 18:19:22 +0000 (18:19 +0000)] 
Merge branch 'master' of win:c:/ghc/git/cygwin/.

6 years agothe late lambda float
Nicolas Frisby [Thu, 7 Mar 2013 20:14:52 +0000 (20:14 +0000)] 
the late lambda float

added the -fno-LNE flag, added a DEBUG message for the oddity of a zero arity LNE

6 years agofixed the sizeExpr typo bug
Nicolas Frisby [Tue, 12 Mar 2013 18:10:46 +0000 (18:10 +0000)] 
fixed the sizeExpr typo bug

6 years agoMerge branch 'master' of darcs.haskell.org:/srv/darcs//ghc
Ian Lynagh [Tue, 12 Mar 2013 00:41:41 +0000 (00:41 +0000)] 
Merge branch 'master' of darcs.haskell.org:/srv/darcs//ghc

6 years agoTypo-fix for panic.
Edward Z. Yang [Tue, 12 Mar 2013 00:22:24 +0000 (17:22 -0700)] 
Typo-fix for panic.

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>
6 years agoUpdate Cabal
Ian Lynagh [Mon, 11 Mar 2013 20:12:36 +0000 (20:12 +0000)] 
Update Cabal

6 years agoMore advice about profiling CAFs.
Edward Z. Yang [Mon, 11 Mar 2013 20:12:25 +0000 (13:12 -0700)] 
More advice about profiling CAFs.

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>
6 years agoFix building the GHC package DLL on Windows
Ian Lynagh [Sun, 10 Mar 2013 16:57:23 +0000 (17:57 +0100)] 
Fix building the GHC package DLL on Windows

We now put a handful of modules in a separate DLL.
For now the list is hand-written, but we could automate it in the
future.

6 years agoMerge branch 'master' of http://darcs.haskell.org/ghc
David Waern [Sun, 10 Mar 2013 19:24:42 +0000 (12:24 -0700)] 
Merge branch 'master' of darcs.haskell.org/ghc

6 years agoFix construction of ALL_ALEX_OPTS and ALL_AS_OPTS
Max Bolingbroke [Sun, 3 Feb 2013 18:05:25 +0000 (18:05 +0000)] 
Fix construction of ALL_ALEX_OPTS and ALL_AS_OPTS

Signed-off-by: David Waern <david.waern@gmail.com>
6 years agoRemove some commented out SPECIALIZE pragmas
Ian Lynagh [Sun, 10 Mar 2013 01:27:34 +0000 (01:27 +0000)] 
Remove some commented out SPECIALIZE pragmas

As far as I can see, they've never been enabled

6 years agoBuilding with -dynamic-too now works, so enable it by default
Ian Lynagh [Sat, 9 Mar 2013 23:11:56 +0000 (23:11 +0000)] 
Building with -dynamic-too now works, so enable it by default

It's only actually used if both the v and dyn ways are built

6 years agoRemove unnecessary warnings suppressions, fixes ticket #7756; thanks monoidal for...
Edward Z. Yang [Sun, 10 Mar 2013 04:16:58 +0000 (20:16 -0800)] 
Remove unnecessary warnings suppressions, fixes ticket #7756; thanks monoidal for submitting.

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>
6 years agoFix -dynamic-too with -boot files
Ian Lynagh [Sat, 9 Mar 2013 20:49:28 +0000 (20:49 +0000)] 
Fix -dynamic-too with -boot files

It was looking for Foo.dyn_hi rather than Foo.dyn_hi-boot, and so
falling back to the slow way

6 years agoMerge branch 'master' of darcs.haskell.org:/srv/darcs//ghc
Ian Lynagh [Sat, 9 Mar 2013 20:00:33 +0000 (20:00 +0000)] 
Merge branch 'master' of darcs.haskell.org:/srv/darcs//ghc

6 years agoSlight tidy-up to Pedro's work on Typeable
Simon Peyton Jones [Sat, 9 Mar 2013 16:34:03 +0000 (16:34 +0000)] 
Slight tidy-up to Pedro's work on Typeable

6 years agoRemove unnecessary DynFlags arg to mkCgIdInfo
Simon Peyton Jones [Sat, 9 Mar 2013 16:33:32 +0000 (16:33 +0000)] 
Remove unnecessary DynFlags arg to mkCgIdInfo

6 years agoRemove stale, commented-out code about heap checks
Simon Peyton Jones [Sat, 9 Mar 2013 16:32:48 +0000 (16:32 +0000)] 
Remove stale, commented-out code about heap checks

6 years agoRemove warning-suppression (not needed)
Simon Peyton Jones [Sat, 9 Mar 2013 16:32:08 +0000 (16:32 +0000)] 
Remove warning-suppression (not needed)

6 years agoDon't pass -dynamic-too to GHC when compiling C files
Ian Lynagh [Sat, 9 Mar 2013 13:28:03 +0000 (13:28 +0000)] 
Don't pass -dynamic-too to GHC when compiling C files

6 years agoSimplify away some old -dynamic-too stuff from the previous approach
Ian Lynagh [Fri, 8 Mar 2013 22:45:28 +0000 (22:45 +0000)] 
Simplify away some old -dynamic-too stuff from the previous approach

6 years agoRe-enable a flag-consistency check
Ian Lynagh [Fri, 8 Mar 2013 22:05:27 +0000 (22:05 +0000)] 
Re-enable a flag-consistency check

I'm not sure why it caused problems before, but it's fine now.