ghc-hetmet.git
14 years agoImprove optimisation of cost centres
simonpj@microsoft.com [Fri, 11 Sep 2009 11:56:30 +0000 (11:56 +0000)]
Improve optimisation of cost centres

This patch fixes test failures for the profiling way for drv001.
The problem was that the arity of a function was decreasing during
"optimisation" because of interaction with SCC annotations.
In particular
      f = /\a. scc "f" (h x)    -- where h had arity 2
and h gets inlined, led to
      f = /\a. scc "f" let v = scc "f" x in \y. <blah>

Two main changes:

1.  exprIsTrivial now says True for (scc "f" x)
    See Note [SCCs are trivial] in CoreUtils

2.  The simplifier eliminates nested pushing of the same cost centre:
   scc "f" (...(scc "f" e)...)
   ==>  scc "f" (...e...)

14 years agoFix build with GHC <= 6.10.1
Simon Marlow [Fri, 11 Sep 2009 10:31:42 +0000 (10:31 +0000)]
Fix build with GHC <= 6.10.1

14 years agoloosen base dependency to allow base-3
Simon Marlow [Fri, 11 Sep 2009 09:31:39 +0000 (09:31 +0000)]
loosen base dependency to allow base-3

14 years agoremove superfluous trailing backslash
Simon Marlow [Fri, 11 Sep 2009 09:31:24 +0000 (09:31 +0000)]
remove superfluous trailing backslash

14 years agoWork around bug in old bytestring versions by reading the package DB strictly
Simon Marlow [Fri, 11 Sep 2009 09:31:03 +0000 (09:31 +0000)]
Work around bug in old bytestring versions by reading the package DB strictly

14 years agoAdd cmm-notes, describing Simon and John's work on Cmm pipeline
simonpj@microsoft.com [Fri, 11 Sep 2009 10:53:16 +0000 (10:53 +0000)]
Add cmm-notes, describing Simon and John's work on Cmm pipeline

14 years agoWibble to error message
simonpj@microsoft.com [Fri, 11 Sep 2009 09:08:19 +0000 (09:08 +0000)]
Wibble to error message

14 years agoadd comment: __thread is not supported by gcc on OS X yet
Simon Marlow [Thu, 10 Sep 2009 15:21:46 +0000 (15:21 +0000)]
add comment: __thread is not supported by gcc on OS X yet

14 years agoFix #3320: we forgot to save/restore the GC register variable
Simon Marlow [Thu, 10 Sep 2009 15:16:23 +0000 (15:16 +0000)]
Fix #3320: we forgot to save/restore the GC register variable

14 years agoput back F_, just in case the via-C codegen uses it
Simon Marlow [Thu, 10 Sep 2009 12:40:37 +0000 (12:40 +0000)]
put back F_, just in case the via-C codegen uses it

14 years agoremove debugging/testing code that got here by mistake
Simon Marlow [Thu, 10 Sep 2009 12:26:07 +0000 (12:26 +0000)]
remove debugging/testing code that got here by mistake

14 years agofix installation of header files (#3451)
Simon Marlow [Thu, 10 Sep 2009 12:18:31 +0000 (12:18 +0000)]
fix installation of header files (#3451)

14 years agoThree improvements to Template Haskell (fixes #3467)
simonpj@microsoft.com [Thu, 10 Sep 2009 12:58:48 +0000 (12:58 +0000)]
Three improvements to Template Haskell (fixes #3467)

This patch implements three significant improvements to Template Haskell.

Declaration-level splices with no "$"
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
This change simply allows you to omit the "$(...)" wrapper for
declaration-level TH splices.  An expression all by itself is
not legal, so we now treat it as a TH splice.  Thus you can now
say
data T = T1 | T2
  deriveMyStuff ''T

where deriveMyStuff :: Name -> Q [Dec]
This makes a much nicer interface for clients of libraries that use
TH: no scary $(deriveMyStuff ''T).

Nested top-level splices
~~~~~~~~~~~~~~~~~~~~~~~~
Previously TH would reject this, saying that splices cannot be nested:
f x = $(g $(h 'x))
But there is no reason for this not to work.  First $(h 'x) is run,
yielding code <blah> that is spliced instead of the $(h 'x). Then (g
<blah>) is typechecked and run, yielding code that replaces the
$(g ...) splice.

So this simply lifts the restriction.

Fix Trac #3467: non-top-level type splices
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

It appears that when I added the ability to splice types in TH
programs, I failed to pay attention to non-top-level splices -- that
is, splices inside quotatation brackets.

This patch fixes the problem.  I had to modify HsType, so there's a
knock-on change to Haddock.

Its seems that a lot of lines of code has changed, but almost all the
new lines are comments!

General tidying up
~~~~~~~~~~~~~~~~~~
As a result of thinking all this out I re-jigged the data type ThStage,
which had far too many values before.  And I wrote a nice state transition
diagram to make it all precise;
   see Note [Template Haskell state diagram] in TcSplice

Lots more refactoring in TcSplice, resulting in significantly less code.
(A few more lines, but actually less code -- the rest is comments.)

I think the result is significantly cleaner.

14 years agoRefactor the parsing of data type declarations
simonpj@microsoft.com [Tue, 8 Sep 2009 19:23:27 +0000 (19:23 +0000)]
Refactor the parsing of data type declarations

This is a minor change to the parser that tidies it up a bit,
and allows us to parse

data T :: *
        data S :: * -> *

just like

data T
data S a

14 years agoRemove unused data constructor ITpragma in type Lexer.Token
simonpj@microsoft.com [Sat, 5 Sep 2009 17:15:11 +0000 (17:15 +0000)]
Remove unused data constructor ITpragma in type Lexer.Token

14 years agoComments in Cmm
simonpj@microsoft.com [Thu, 10 Sep 2009 12:29:28 +0000 (12:29 +0000)]
Comments in Cmm

14 years agofix warnings
Simon Marlow [Thu, 10 Sep 2009 11:11:56 +0000 (11:11 +0000)]
fix warnings

14 years agobinary has unused-import warnings
Simon Marlow [Thu, 10 Sep 2009 11:11:15 +0000 (11:11 +0000)]
binary has unused-import warnings

14 years agoChange the representation of the package database
Simon Marlow [Thu, 10 Sep 2009 10:27:03 +0000 (10:27 +0000)]
Change the representation of the package database

 - the package DB is a directory containing one file per package
   instance (#723)

 - there is a binary cache of the database (#593, #2089)

 - the binary package is now a boot package

 - there is a new package, bin-package-db, containing the Binary
   instance of InstalledPackageInfo for the binary cache.

Also included in this patch

 - Use colour in 'ghc-pkg list' to indicate broken or hidden packages

   Broken packages are red, hidden packages are

   Colour support comes from the terminfo package, and is only used when
    - not --simple-output
    - stdout is a TTY
    - the terminal type has colour capability

 - Fix the bug that 'ghc-pkg list --user' shows everything as broken

14 years agoImprove the user guide section on shared libs
Duncan Coutts [Tue, 8 Sep 2009 18:32:41 +0000 (18:32 +0000)]
Improve the user guide section on shared libs
Make it clear that Haskell code to be used by other Haskell code
must be built as a package.

14 years agoFIX #711 implement osFreeAllMBlocks for unix
Austin Seipp [Thu, 10 Sep 2009 08:46:30 +0000 (08:46 +0000)]
FIX #711 implement osFreeAllMBlocks for unix

14 years agoNO_CLEAN_GMP fixes
Simon Marlow [Wed, 9 Sep 2009 09:34:43 +0000 (09:34 +0000)]
NO_CLEAN_GMP fixes

14 years agoOmit visibility pragmas on Windows (fixes warnings/validate failures)
Simon Marlow [Wed, 9 Sep 2009 09:32:17 +0000 (09:32 +0000)]
Omit visibility pragmas on Windows (fixes warnings/validate failures)

14 years agoComments only, on Cmm data types
simonpj@microsoft.com [Thu, 10 Sep 2009 09:17:03 +0000 (09:17 +0000)]
Comments only, on Cmm data types

14 years agoFix up logic to decide whether shared libs are being built or not
Simon Marlow [Wed, 9 Sep 2009 11:57:57 +0000 (11:57 +0000)]
Fix up logic to decide whether shared libs are being built or not

Now, adding dyn to $(GhcLibWays) is how shared libs are enabled.
Everything else keys off that, rather than testing
$(BuildSharedLibs).

14 years agoremove a ToDo that is Done
Simon Marlow [Tue, 8 Sep 2009 15:44:11 +0000 (15:44 +0000)]
remove a ToDo that is Done

14 years agocomment fix
Simon Marlow [Tue, 8 Sep 2009 15:03:51 +0000 (15:03 +0000)]
comment fix

14 years agorefactor *_OPTS settings in the build system; no functional changes
Simon Marlow [Wed, 9 Sep 2009 09:44:08 +0000 (09:44 +0000)]
refactor *_OPTS settings in the build system; no functional changes

14 years agoadd $(CONF_*_OPTS) for options that come from ./configure (fixes #3426)
Simon Marlow [Tue, 8 Sep 2009 15:19:08 +0000 (15:19 +0000)]
add $(CONF_*_OPTS) for options that come from ./configure (fixes #3426)

14 years agoAdd -package-id, improve package shadowing behaviour and error messages
Simon Marlow [Tue, 8 Sep 2009 14:09:06 +0000 (14:09 +0000)]
Add -package-id, improve package shadowing behaviour and error messages
A few changes bundled together here:

 - Add -package-id flag, like -package but takes an
   InstalledPackageId as an argument (mainly for Cabal)
   (docs to follow)

 - Fix the shadowing behaviour: if we end up with
   two packages with the same name/version that came from
   different package databases, then one will shadow the
   other.  This may mean we have to disable other packages
   that depended on the now-shadowed package.  Lots of
   refactoring to ensure that we can get reasonable diagnostics when
   this happens

<command line>: cannot satisfy -package shadowdep:
    shadowdep-1-XXX is unusable due to missing or recursive dependencies:
      shadow-1-XXX
    (use -v for more information)

14 years agoremove --enable-hc-boot-unregisterised (implied by --enable-hc-boot now)
Simon Marlow [Tue, 8 Sep 2009 14:19:19 +0000 (14:19 +0000)]
remove --enable-hc-boot-unregisterised (implied by --enable-hc-boot now)
and fix up related cruft

14 years agofix a warning
Simon Marlow [Tue, 8 Sep 2009 13:15:51 +0000 (13:15 +0000)]
fix a warning

14 years agofix warnings
Simon Marlow [Tue, 8 Sep 2009 13:08:54 +0000 (13:08 +0000)]
fix warnings

14 years agofix warnings
Simon Marlow [Tue, 8 Sep 2009 11:42:00 +0000 (11:42 +0000)]
fix warnings

14 years agoremove Haddock-lexing/parsing/renaming from GHC
Isaac Dupree [Wed, 26 Aug 2009 14:59:24 +0000 (14:59 +0000)]
remove Haddock-lexing/parsing/renaming from GHC

14 years agoFix Trac #3403: interaction of CPR and pattern-match failure
simonpj@microsoft.com [Tue, 8 Sep 2009 13:14:00 +0000 (13:14 +0000)]
Fix Trac #3403: interaction of CPR and pattern-match failure

A fine bug report (#3403) demonstrated that we were losing the tail
call property when a complicated pattern match was involved.  After
a bit of investigation I discovered that the culprit was the failure
join-point introduced by the pattern matcher.  It was a zero-argument
thunk, which is not very CPR-friendly, and that interacted badly with
CPR worker/wrapper.

It's easy to fix, the same way that we fix other join points, by supplying
a dummy argument (that is not really passed at runtime.

14 years agoFix Trac #3468: improve checking for hs-boot interfaces
simonpj@microsoft.com [Tue, 8 Sep 2009 13:03:50 +0000 (13:03 +0000)]
Fix Trac #3468: improve checking for hs-boot interfaces

When checking the interface exported by a hs-boot file against the
Real Thing, I'd failed to check the kind of a type constructor.  If you
get it wrong, the inconsistency leads to all manner of mischief, as
'wkahl' reports in #3468.

This patch should do the job.

14 years agoAllow TH/annotations to be used with -dynamic
Simon Marlow [Tue, 8 Sep 2009 10:16:44 +0000 (10:16 +0000)]
Allow TH/annotations to be used with -dynamic

14 years agoFix warnings on 64-bit platforms; fixes validate on x86-64
Simon Marlow [Tue, 8 Sep 2009 09:26:39 +0000 (09:26 +0000)]
Fix warnings on 64-bit platforms; fixes validate on x86-64

14 years agofix -u flags (after changes for #3310)
Simon Marlow [Thu, 3 Sep 2009 07:23:18 +0000 (07:23 +0000)]
fix -u flags (after changes for #3310)

14 years agoUnify event logging and debug tracing.
Simon Marlow [Sat, 29 Aug 2009 09:47:27 +0000 (09:47 +0000)]
Unify event logging and debug tracing.

  - tracing facilities are now enabled with -DTRACING, and -DDEBUG
    additionally enables debug-tracing.  -DEVENTLOG has been
    removed.

  - -debug now implies -eventlog

  - events can be printed to stderr instead of being sent to the
    binary .eventlog file by adding +RTS -v (which is implied by the
    +RTS -Dx options).

  - -Dx debug messages can be sent to the binary .eventlog file
    by adding +RTS -l.  This should help debugging by reducing
    the impact of debug tracing on execution time.

  - Various debug messages that duplicated the information in events
    have been removed.

14 years agowaitForReturnCapability: fix logic bug
Simon Marlow [Mon, 31 Aug 2009 06:49:08 +0000 (06:49 +0000)]
waitForReturnCapability: fix logic bug
The check for whether a Capability was free was inverted, which harmed
performance for callbacks.

14 years agoHandle renames from #3310
Simon Marlow [Sun, 30 Aug 2009 15:31:23 +0000 (15:31 +0000)]
Handle renames from #3310

Also add a panic for resurrecting a thread blocked on an exception,
since it should never happen.

14 years agoadd RTS_PRIVATE attribute
Simon Marlow [Sat, 29 Aug 2009 17:17:52 +0000 (17:17 +0000)]
add RTS_PRIVATE attribute

14 years agoFix incorrectly hidden RTS symbols
Simon Marlow [Sat, 29 Aug 2009 13:28:14 +0000 (13:28 +0000)]
Fix incorrectly hidden RTS symbols

14 years agoMaintain Task/Capability invariants in performPendingThrowTos
Simon Marlow [Sat, 29 Aug 2009 09:26:07 +0000 (09:26 +0000)]
Maintain Task/Capability invariants in performPendingThrowTos
Fixes an ASSERTION failure with concprog001, -threaded -debug, +RTS -N2

14 years agoDeclare RTS-private prototypes with __attribute__((visibility("hidden")))
Simon Marlow [Wed, 5 Aug 2009 10:21:59 +0000 (10:21 +0000)]
Declare RTS-private prototypes with __attribute__((visibility("hidden")))
This has no effect with static libraries, but when the RTS is in a
shared library it does two things:

 - it prevents the function from being exposed by the shared library

 - internal calls to the function can use the faster non-PLT calls,
   because the function cannot be overriden at link time.

14 years agoConfigurable iconv header and library locations
Matthias Kilian [Wed, 26 Aug 2009 15:59:13 +0000 (15:59 +0000)]
Configurable iconv header and library locations
Should help to fix the build on OpenBSD (together with a corresponding
patch to libraries/base).

14 years agoUse -W, not -Werror, for gcc older than 3.4
Matthias Kilian [Wed, 26 Aug 2009 21:00:44 +0000 (21:00 +0000)]
Use -W, not -Werror, for gcc older than 3.4

14 years agoDon't mess up absolute INCLUDE_DIRS
Matthias Kilian [Wed, 26 Aug 2009 15:33:40 +0000 (15:33 +0000)]
Don't mess up absolute INCLUDE_DIRS
When there are some absolute include dirs, don't create arguments
like -Ilibraries/base//usr/local/include.

I need this fix for injecting the iconv include dir (which is
/usr/local/include on OpenBSD) into the build of libraries/base.

14 years agoFix #3461: protect the use of keepCAFs with #ifdef DYNAMIC
Simon Marlow [Fri, 28 Aug 2009 12:58:02 +0000 (12:58 +0000)]
Fix #3461: protect the use of keepCAFs with #ifdef DYNAMIC

14 years agofix 'darcs-all rec' (amongst other things)
Simon Marlow [Thu, 27 Aug 2009 14:15:54 +0000 (14:15 +0000)]
fix 'darcs-all rec' (amongst other things)

14 years agoREDO: Add -r option to darcs-all, and remove push-all (#3375)
Simon Marlow [Thu, 27 Aug 2009 13:57:17 +0000 (13:57 +0000)]
REDO: Add -r option to darcs-all, and remove push-all (#3375)

rolling back:

Mon Aug  3 11:44:13 BST 2009  Simon Marlow <marlowsd@gmail.com>
  UNDO: Add -r option to darcs-all, and remove push-all (#3375)
  Contributed by: seliopou@gmail.com

  This patch modifies darcs-all to have feature parity with push-all by
  recognizing two new options.

      * -i, equivalent to --ignore-failure in push-all
      * -r <repo>, specifies the remote repository darcs commands will use

  Some example commands:

  Get the libraries from a repository of your choosing. This is useful
  when working with a git mirror:

      $ ./darcs-all -r http://darcs.haskell.org get

  Pull changes. Used to be:

      $ ./push-all --pull http://darcs.haskell.org

  Is now:

      $ ./darcs-all -r http://darcs.haskell.org pull

  Or to use the default remote of the ghc repository:

      $ ./darcs-all pull

    M ./darcs-all -79 +33
    A ./push-all

14 years agoAdd a link to hp2any from the profiling section.
Simon Marlow [Thu, 27 Aug 2009 11:40:50 +0000 (11:40 +0000)]
Add a link to hp2any from the profiling section.

14 years agoFix "make install"
Ian Lynagh [Wed, 26 Aug 2009 15:56:38 +0000 (15:56 +0000)]
Fix "make install"
We need to change a dependency on
    pkg-inplace
to
    pkg-$abihash
when installing

14 years ago"ghc-cabal install" now needs to know where GHC is, to get the ABI hash
Ian Lynagh [Wed, 26 Aug 2009 13:33:59 +0000 (13:33 +0000)]
"ghc-cabal install" now needs to know where GHC is, to get the ABI hash

14 years agoFix bindist creation
Ian Lynagh [Wed, 26 Aug 2009 12:29:53 +0000 (12:29 +0000)]
Fix bindist creation
We were running into problems like:
    for f in  LICENSE configure config.sub config.guess [...]
    make[2]: execvp: /bin/sh: Argument list too long
This patch moves the loop into make, rather than the shell.

14 years agoUNDO: fix the inplace runghc (it broke the installed runghc)
Simon Marlow [Wed, 26 Aug 2009 10:49:37 +0000 (10:49 +0000)]
UNDO: fix the inplace runghc (it broke the installed runghc)

14 years agoFix part of #3398: pretty-printing always goes via the I/O library encoding
Simon Marlow [Wed, 26 Aug 2009 10:40:52 +0000 (10:40 +0000)]
Fix part of #3398: pretty-printing always goes via the I/O library encoding
That is, unless we're printing in LeftMode, where we bypass encoding
for speed.  This is safe, because LeftMode is used for outputting C or
asm, where everyting is Z-encoded and hence ASCII.

Error messages and other compiler output containing Unicode will now
appear correctly according to the locale settings.

14 years agoTidy up file headers and copyrights; point to the wiki for docs
Simon Marlow [Tue, 25 Aug 2009 09:50:47 +0000 (09:50 +0000)]
Tidy up file headers and copyrights; point to the wiki for docs

I've updated the wiki page about the RTS headers
  http://hackage.haskell.org/trac/ghc/wiki/Commentary/SourceTree/Includes
to reflect the new layout and explain some of the rationale.  All the
header files now point to this page.

14 years agoFollow changes in Cabal: package -> sourcePackageId
Simon Marlow [Mon, 24 Aug 2009 16:00:20 +0000 (16:00 +0000)]
Follow changes in Cabal: package -> sourcePackageId

14 years agofix an unused-import warning
Simon Marlow [Mon, 24 Aug 2009 15:36:57 +0000 (15:36 +0000)]
fix an unused-import warning

14 years agoAdd unique package identifiers (InstalledPackageId) in the package DB
Simon Marlow [Thu, 20 Aug 2009 11:09:20 +0000 (11:09 +0000)]
Add unique package identifiers (InstalledPackageId) in the package DB
See commentary at
http://hackage.haskell.org/trac/ghc/wiki/Commentary/Packages

14 years agoMove the standalone-deriving flag test from parser to renamer
simonpj@microsoft.com [Tue, 25 Aug 2009 07:33:24 +0000 (07:33 +0000)]
Move the standalone-deriving flag test from parser to renamer

This is just a tiny refactoring.  In general, we're trying to
get rid of parser errors in favour of later, more civlised, errors.

14 years agoError message wibble
simonpj@microsoft.com [Tue, 25 Aug 2009 07:32:59 +0000 (07:32 +0000)]
Error message wibble

14 years agoFix Trac #3406 (albeit not very satisfactorily): scoped type variables
simonpj@microsoft.com [Tue, 25 Aug 2009 07:30:59 +0000 (07:30 +0000)]
Fix Trac #3406 (albeit not very satisfactorily): scoped type variables

The issue here is this:

  type ItemColID a b = Int  -- Discards a,b

  get :: ItemColID a b -> a -> ItemColID a b
  get (x :: ItemColID a b) = x :: ItemColID a b

The pattern signature for 'x' doesn't actually rigidly bind a,b.
This crashed GHC 6.10 with a 'readFilledBox' panic.  Now we fail
with an erroe message

With the new outside-in algorithm we'll be able to accept this program.

14 years agoMake FastString thread-safe.
Thomas Schilling [Mon, 24 Aug 2009 18:22:52 +0000 (18:22 +0000)]
Make FastString thread-safe.

This is needed both for per-session parallelism and for allowing
multiple concurrent sessions in the same process.  With the help of
atomicModifyIORef and unsafePerformIO it is also quite fast--an MVar
would most likely be slower.  On a full compilation of Cabal's head
branch it was about 1-2 percent slower, but then overall compilation
times varied by about 4 percent, so I think it's worth it.

14 years agoImprove docs on -XFlexibleContexts
simonpj@microsoft.com [Mon, 24 Aug 2009 11:37:22 +0000 (11:37 +0000)]
Improve docs on -XFlexibleContexts

14 years agofix do-notation warnings
Simon Marlow [Mon, 24 Aug 2009 08:04:00 +0000 (08:04 +0000)]
fix do-notation warnings

14 years agoUse explicit Word32/Int32 in place of Int in the on-disk .hi file
Simon Marlow [Fri, 21 Aug 2009 15:50:28 +0000 (15:50 +0000)]
Use explicit Word32/Int32 in place of Int in the on-disk .hi file
For: FastStrings, Names, and Bin values.  This makes .hi files smaller
on 64-bit platforms, while also making the format a bit more robust.

14 years agoFix the interface-file incompatibility crash (#3435)
Simon Marlow [Fri, 21 Aug 2009 15:47:37 +0000 (15:47 +0000)]
Fix the interface-file incompatibility crash (#3435)

We now have a dummy 32/64-bit field near the beginning of a .hi file
for backward-compatibility reasons; see comments (Note [dummy iface
field]) in BinIface.hs.

14 years agotrim Data.Sequence import, in preparation for expanding its API
Ross Paterson [Mon, 24 Aug 2009 08:05:33 +0000 (08:05 +0000)]
trim Data.Sequence import, in preparation for expanding its API

14 years agoRemove bitrotted IA64 code in Linker.c
Ian Lynagh [Sun, 23 Aug 2009 15:16:00 +0000 (15:16 +0000)]
Remove bitrotted IA64 code in Linker.c
It breaks the unregisterised build on IA64.

14 years agoMake consIORef atomic.
Thomas Schilling [Sat, 22 Aug 2009 23:09:37 +0000 (23:09 +0000)]
Make consIORef atomic.

14 years agoMake updates to external package state idempotent.
Thomas Schilling [Fri, 21 Aug 2009 23:53:55 +0000 (23:53 +0000)]
Make updates to external package state idempotent.

Without this, concurrent updates to the EPS could introduce
overlapping instances (even though they came from the same module).

14 years agoFix Trac #3423: missed instantiation for newtype-derived instances
simonpj@microsoft.com [Fri, 21 Aug 2009 21:07:00 +0000 (21:07 +0000)]
Fix Trac #3423: missed instantiation for newtype-derived instances

Somehow I'd forgotten to instantiate the coercion that is stored in a
'NewtypeDerived' constructor in an InstInfo.  The necessary code is
in TcInstDcls.tc_inst_decl2.

The result was ghc: panic! (the 'impossible' happened)
   (GHC version 6.10.3 for x86_64-unknown-linux):
   No match in record selector Var.tcTyVarDetails
because we were looking at an (uninstantiated) TyVar instead of
an (instantiated) TcTyVar.

14 years agoPut "dl" back in rts/package.conf if HAVE_DL is defined
Ian Lynagh [Fri, 21 Aug 2009 14:54:23 +0000 (14:54 +0000)]
Put "dl" back in rts/package.conf if HAVE_DL is defined
Fixes linking with -dynamic

14 years agoLink CMM objects into dynamic libraries
Ian Lynagh [Fri, 21 Aug 2009 13:28:09 +0000 (13:28 +0000)]
Link CMM objects into dynamic libraries
This fixes linking hello world with -dynamic. I've also added some more
variables, so there is less duplication between the different ways of
linking.

14 years ago-fPIC -fvia-C issues a warning and ignores -fvia-C
Simon Marlow [Fri, 21 Aug 2009 14:45:44 +0000 (14:45 +0000)]
-fPIC -fvia-C issues a warning and ignores -fvia-C
Also, -fPIC causes an error if the target is registerised and has no
native code generator.

14 years agoUse allocateLocal() rather than allocate() in the interpreter
Simon Marlow [Thu, 20 Aug 2009 15:23:25 +0000 (15:23 +0000)]
Use allocateLocal() rather than allocate() in the interpreter
This gives about a 15% performance boost in GHCi for me.  nice!

14 years agoAnother tiny tidy-up to RnPat
simonpj@microsoft.com [Fri, 21 Aug 2009 10:08:26 +0000 (10:08 +0000)]
Another tiny tidy-up to RnPat

14 years agoFix Trac #3437: strictness of specialised functions
simonpj@microsoft.com [Fri, 21 Aug 2009 09:52:51 +0000 (09:52 +0000)]
Fix Trac #3437: strictness of specialised functions

'lilac' helpful pin-pointed a space leak that was due to a specialised
function being insufficiently strict.  Here's the new comment in SpecConstr:

Note [Transfer strictness]
~~~~~~~~~~~~~~~~~~~~~~~~~~
We must transfer strictness information from the original function to
the specialised one.  Suppose, for example

  f has strictness     SS
        and a RULE     f (a:as) b = f_spec a as b

Now we want f_spec to have strictess  LLS, otherwise we'll use call-by-need
when calling f_spec instead of call-by-value.  And that can result in
unbounded worsening in space (cf the classic foldl vs foldl')

See Trac #3437 for a good example.

The function calcSpecStrictness performs the calculation.

14 years agoWibbles to field-label puns
simonpj@microsoft.com [Fri, 21 Aug 2009 09:06:37 +0000 (09:06 +0000)]
Wibbles to field-label puns

14 years agoFix library installation; fixes #3374
Ian Lynagh [Thu, 20 Aug 2009 17:37:07 +0000 (17:37 +0000)]
Fix library installation; fixes #3374
When configuring packages, enable library profiling and shared libraries
based on the ways in GhcLibWays.

14 years agoEscape some $s in makefiles for consistency
Ian Lynagh [Thu, 20 Aug 2009 16:50:33 +0000 (16:50 +0000)]
Escape some $s in makefiles for consistency

14 years agoWibble to RnPat refactoring
simonpj@microsoft.com [Thu, 20 Aug 2009 16:15:20 +0000 (16:15 +0000)]
Wibble to RnPat refactoring

14 years agoRollback: use cas() to claim the closure in copyPart(), to match copy_tag()
Simon Marlow [Thu, 20 Aug 2009 15:58:33 +0000 (15:58 +0000)]
Rollback: use cas() to claim the closure in copyPart(), to match copy_tag()

rolling back:

  * use cas() to claim the closure in copyPart(), to match copy_tag()
  * rename whitehole_spin to evac_collision, and update it properly

This introduced a new failure in parallel GC.  I'll rollback for now
until I've fixed it.

14 years agoRelax the assumption that all objects fit in a single block (#3424)
Simon Marlow [Thu, 20 Aug 2009 14:43:08 +0000 (14:43 +0000)]
Relax the assumption that all objects fit in a single block (#3424)

It is possible for the program to allocate single object larger than a
block, without going through the normal large-object mechanisms that
we have for arrays and threads and so on.

The GC was assuming that no object was larger than a block, but #3424
contains a program that breaks the assumption.  This patch removes the
assumption.  The objects in question will still be copied, that is
they don't get the normal large-object treatment, but this case is
unlikely to occur often in practice.

In the future we may improve things by generating code to allocate
them as large objects in the first place.

14 years agoremove a bogus assertion
Simon Marlow [Thu, 20 Aug 2009 14:33:05 +0000 (14:33 +0000)]
remove a bogus assertion

14 years agoRestore the GHC version check
Simon Marlow [Thu, 20 Aug 2009 13:42:34 +0000 (13:42 +0000)]
Restore the GHC version check
I'm not sure what happened here, but the ordering of tests was messed
up, with the result that the GHC version check was being omitted.

14 years agofix the inplace runghc
Simon Marlow [Thu, 20 Aug 2009 12:25:38 +0000 (12:25 +0000)]
fix the inplace runghc

It was invoking $(TOP)/inplace/bin/ghc rather than
$(TOP)/inplace/bin/ghc-stage2

14 years agoAdd a case for IND (and a comment). Fixes #3424, perhaps only partially.
Simon Marlow [Thu, 20 Aug 2009 13:15:37 +0000 (13:15 +0000)]
Add a case for IND (and a comment).  Fixes #3424, perhaps only partially.

14 years agoImprovements to record puns, wildcards
simonpj@microsoft.com [Thu, 20 Aug 2009 12:34:43 +0000 (12:34 +0000)]
Improvements to record puns, wildcards

* Make C { A.a } work with punning, expanding to C { A.a = a }

* Make it so that, with -fwarn-unused-matches,
        f (C {..}) = x
  does not complain about the bindings introduced by the "..".

* Make -XRecordWildCards implies -XDisambiguateRecordFields.

* Overall refactoring of RnPat, which had become very crufty.
  In particular, there is now a monad, CpsRn, private to RnPat,
  which deals with the cps-style plumbing.  This is why so many
  lines of RnPat have changed.

* Refactor the treatment of renaming of record fields into two passes
- rnHsRecFields1, used both for patterns and expressions,
     which expands puns, wild-cards
   - a local renamer in RnPat for fields in patterns
- a local renamer in RnExpr for fields in construction and update
  This make it all MUCH easier to understand

* Improve documentation of record puns, wildcards, and disambiguation

14 years agoMake -dppr-debug print locations in HsSyn
simonpj@microsoft.com [Thu, 20 Aug 2009 12:32:45 +0000 (12:32 +0000)]
Make -dppr-debug print locations in HsSyn

Show SrcSpans for Located things might be overkill, but it's sometimes
useful.

I also added
  ppWhen, ppUnless :: Bool -> SDoc -> SDoc
to Outputable

14 years agoMake -dynamic a proper way, so we read the .dyn_hi files
Simon Marlow [Thu, 20 Aug 2009 12:12:08 +0000 (12:12 +0000)]
Make -dynamic a proper way, so we read the .dyn_hi files
Also, I cleaned up some of the way-related infrastructure, removing
two global variables.

There's more that could be done here, but it's a start.  The way flags
probably don't need to be static any more.

14 years agofix warning
Simon Marlow [Thu, 20 Aug 2009 11:05:32 +0000 (11:05 +0000)]
fix warning

14 years agogeneralise the type of "on"
Simon Marlow [Thu, 20 Aug 2009 09:45:16 +0000 (09:45 +0000)]
generalise the type of "on"

14 years agoneed to check $(HADDOCK_DOCS) around contents/index generation
Simon Marlow [Thu, 20 Aug 2009 09:22:20 +0000 (09:22 +0000)]
need to check $(HADDOCK_DOCS) around contents/index generation

14 years agoWAY_dyn_LIB_TARGET is not used anywhere; kill it
Simon Marlow [Wed, 19 Aug 2009 15:59:52 +0000 (15:59 +0000)]
WAY_dyn_LIB_TARGET is not used anywhere; kill it

14 years agoRolling back: Add a kludge to fix building shared libs
Simon Marlow [Wed, 19 Aug 2009 14:55:07 +0000 (14:55 +0000)]
Rolling back: Add a kludge to fix building shared libs