ghc-hetmet.git
21 years ago[project @ 2002-06-19 13:38:36 by simonmar]
simonmar [Wed, 19 Jun 2002 13:38:36 +0000 (13:38 +0000)]
[project @ 2002-06-19 13:38:36 by simonmar]
Include directories called 'tests' in source distributions

21 years ago[project @ 2002-06-19 12:01:28 by simonmar]
simonmar [Wed, 19 Jun 2002 12:01:28 +0000 (12:01 +0000)]
[project @ 2002-06-19 12:01:28 by simonmar]
Fix wibble in previous commit to Weak.c.  Should fix problems with
--enable-threaded-rts.

21 years ago[project @ 2002-06-18 15:13:32 by simonpj]
simonpj [Tue, 18 Jun 2002 15:13:32 +0000 (15:13 +0000)]
[project @ 2002-06-18 15:13:32 by simonpj]
clarification in multi-threading docs

21 years ago[project @ 2002-06-18 14:00:07 by simonpj]
simonpj [Tue, 18 Jun 2002 14:00:07 +0000 (14:00 +0000)]
[project @ 2002-06-18 14:00:07 by simonpj]
Comments only

21 years ago[project @ 2002-06-18 13:58:53 by simonpj]
simonpj [Tue, 18 Jun 2002 13:58:53 +0000 (13:58 +0000)]
[project @ 2002-06-18 13:58:53 by simonpj]
typo

21 years ago[project @ 2002-06-18 13:58:22 by simonpj]
simonpj [Tue, 18 Jun 2002 13:58:24 +0000 (13:58 +0000)]
[project @ 2002-06-18 13:58:22 by simonpj]
---------------------------------------
    Rehash the handling of SeqOp
---------------------------------------

See the comments in the commentary (Cunning Prelude Code).

* Expunge SeqOp altogether

* Add GHC.Base.lazy :: a -> a
  to GHC.Base

* Add GHC.Base.lazy
  to basicTypes/MkId.  The idea is that this defn will over-ride
  the info from GHC.Base.hi, thereby hiding strictness and
  unfolding

* Make stranal/WorkWrap do a "manual inlining" for GHC.Base.lazy
  This happens nicely after the strictness analyser has run.

* Expunge the SeqOp/ParOp magic in CorePrep

* Expunge the RULE for seq in PrelRules

* Change the defns of pseq/par in GHC.Conc to:

{-# INLINE pseq  #-}
        pseq :: a -> b -> b
        pseq  x y = x `seq` lazy y

        {-# INLINE par  #-}
        par :: a -> b -> b
        par  x y = case (par# x) of { _ -> lazy y }

21 years ago[project @ 2002-06-18 12:43:42 by simonpj]
simonpj [Tue, 18 Jun 2002 12:43:42 +0000 (12:43 +0000)]
[project @ 2002-06-18 12:43:42 by simonpj]
Notes about par/seq

21 years ago[project @ 2002-06-18 08:05:44 by simonpj]
simonpj [Tue, 18 Jun 2002 08:05:44 +0000 (08:05 +0000)]
[project @ 2002-06-18 08:05:44 by simonpj]
Wibble on printing outmost parens

21 years ago[project @ 2002-06-17 16:21:42 by simonpj]
simonpj [Mon, 17 Jun 2002 16:21:42 +0000 (16:21 +0000)]
[project @ 2002-06-17 16:21:42 by simonpj]
Ignore fewer type errors in tcSimplifyTop; fixes tc106

21 years ago[project @ 2002-06-17 15:53:42 by simonpj]
simonpj [Mon, 17 Jun 2002 15:53:42 +0000 (15:53 +0000)]
[project @ 2002-06-17 15:53:42 by simonpj]
Fix HnType parenthesisation; fixes rnfail020, tcfail057

21 years ago[project @ 2002-06-17 11:24:30 by simonpj]
simonpj [Mon, 17 Jun 2002 11:24:30 +0000 (11:24 +0000)]
[project @ 2002-06-17 11:24:30 by simonpj]
Record correct location for fixity errors

21 years ago[project @ 2002-06-16 16:10:29 by panne]
panne [Sun, 16 Jun 2002 16:10:29 +0000 (16:10 +0000)]
[project @ 2002-06-16 16:10:29 by panne]
* Use the encoded module name for the C stub of "foreign import
  wrapper", so e.g. the '.' in hierarchical module names gets mangled.
* Nuke unused parameter in mkFExportCBits

21 years ago[project @ 2002-06-15 20:23:12 by wolfgang]
wolfgang [Sat, 15 Jun 2002 20:23:13 +0000 (20:23 +0000)]
[project @ 2002-06-15 20:23:12 by wolfgang]
Documentation for Darwin/MacOS X-Specific options/packacke.conf entries
(-framework, -framework-path; framework_dirs, extra_frameworks)

21 years ago[project @ 2002-06-15 19:38:11 by wolfgang]
wolfgang [Sat, 15 Jun 2002 19:38:11 +0000 (19:38 +0000)]
[project @ 2002-06-15 19:38:11 by wolfgang]
remove RTS_DARWIN_ONLY_SYMBOLS, use RTS_EXTRA_SYMBOLS instead.

21 years ago[project @ 2002-06-15 14:45:53 by sof]
sof [Sat, 15 Jun 2002 14:45:53 +0000 (14:45 +0000)]
[project @ 2002-06-15 14:45:53 by sof]
all: restore old behaviour of exiting upon 'boot' failure in a SUBDIR

21 years ago[project @ 2002-06-15 12:05:13 by sof]
sof [Sat, 15 Jun 2002 12:05:13 +0000 (12:05 +0000)]
[project @ 2002-06-15 12:05:13 by sof]
Accept empty package lists.

21 years ago[project @ 2002-06-14 16:19:04 by simonpj]
simonpj [Fri, 14 Jun 2002 16:19:04 +0000 (16:19 +0000)]
[project @ 2002-06-14 16:19:04 by simonpj]
Another HsParTy wibble; cures rn018 failure

21 years ago[project @ 2002-06-14 14:04:01 by simonpj]
simonpj [Fri, 14 Jun 2002 14:04:01 +0000 (14:04 +0000)]
[project @ 2002-06-14 14:04:01 by simonpj]
Make count_lines use the in-place unlit

21 years ago[project @ 2002-06-14 14:03:25 by simonpj]
simonpj [Fri, 14 Jun 2002 14:03:27 +0000 (14:03 +0000)]
[project @ 2002-06-14 14:03:25 by simonpj]
---------------------------------------
Utterly expunge the tyGenInfo field of
an IdInfo
---------------------------------------

tyGenInfo was a relic of a previous version of Keith's usage
analyser.  It's just dead code, so I've nuked it.

21 years ago[project @ 2002-06-14 14:00:49 by simonpj]
simonpj [Fri, 14 Jun 2002 14:00:49 +0000 (14:00 +0000)]
[project @ 2002-06-14 14:00:49 by simonpj]
Wibble in Generics; fixes str002

21 years ago[project @ 2002-06-14 08:23:57 by simonpj]
simonpj [Fri, 14 Jun 2002 08:23:57 +0000 (08:23 +0000)]
[project @ 2002-06-14 08:23:57 by simonpj]
---------------------------------------
Add a target "count" which counts lines
       in user source files
---------------------------------------

21 years ago[project @ 2002-06-13 23:31:55 by sof]
sof [Thu, 13 Jun 2002 23:31:55 +0000 (23:31 +0000)]
[project @ 2002-06-13 23:31:55 by sof]
Make it compile

21 years ago[project @ 2002-06-13 21:14:51 by wolfgang]
wolfgang [Thu, 13 Jun 2002 21:14:51 +0000 (21:14 +0000)]
[project @ 2002-06-13 21:14:51 by wolfgang]
added a clarifying comment to the PowerPC version

21 years ago[project @ 2002-06-13 11:41:15 by simonmar]
simonmar [Thu, 13 Jun 2002 11:41:15 +0000 (11:41 +0000)]
[project @ 2002-06-13 11:41:15 by simonmar]
Fix a logic bug which caused some dead weak pointers to not get
stubbed out with DEAD_WEAK.

21 years ago[project @ 2002-06-12 22:29:43 by wolfgang]
wolfgang [Wed, 12 Jun 2002 22:29:44 +0000 (22:29 +0000)]
[project @ 2002-06-12 22:29:43 by wolfgang]
Initial support for GHCi on MacOS X.
The code still has problems with global variables in compiled C code, but
it _seems_ to work OK for GHC-compiled code. Many error checks are still
missing.

21 years ago[project @ 2002-06-12 22:04:25 by wolfgang]
wolfgang [Wed, 12 Jun 2002 22:04:27 +0000 (22:04 +0000)]
[project @ 2002-06-12 22:04:25 by wolfgang]
Added support for Frameworks on MacOS X.
*) On MacOS X, two additional command-line options are supported:
-framework <FRAMEWORK>    link with framework, gets passed on to ld
-framework-path <PATH>    gets passed on to ld as "-F<PATH>". (-F is already taken for GHC).
*) Two corresponding additional options for package.conf files:
framework_dirs
extra_frameworks
These options are allowed on any platform. They are ignored everywhere except on MacOS X.
*) ghc-pkg no longer uses Read. Instead, it uses a Happy parser. ghc/utils/ghc-pkg/ParsePkgConfLite.y is basically a copy of ghc/compiler/main/ParsePkgConf.y. "Light" refers to the fact that it doesn't depend on other GHC modules and has less sophisticated error reporting.
*) Frameworks will need some additional work for GHCi.

21 years ago[project @ 2002-06-12 21:43:47 by wolfgang]
wolfgang [Wed, 12 Jun 2002 21:43:47 +0000 (21:43 +0000)]
[project @ 2002-06-12 21:43:47 by wolfgang]
(powerpc-darwin) Add a missing newline to the prologue.

21 years ago[project @ 2002-06-11 08:06:33 by matthewc]
matthewc [Tue, 11 Jun 2002 08:06:33 +0000 (08:06 +0000)]
[project @ 2002-06-11 08:06:33 by matthewc]
#ifdef use of EM_IA_64

21 years ago[project @ 2002-06-10 02:13:03 by matthewc]
matthewc [Mon, 10 Jun 2002 02:13:03 +0000 (02:13 +0000)]
[project @ 2002-06-10 02:13:03 by matthewc]
Fix nuking of warnings, and nuke another one.

21 years ago[project @ 2002-06-09 19:27:16 by panne]
panne [Sun, 9 Jun 2002 19:27:16 +0000 (19:27 +0000)]
[project @ 2002-06-09 19:27:16 by panne]
Un-break previous commit and nuke a few warnings

21 years ago[project @ 2002-06-09 14:01:27 by matthewc]
matthewc [Sun, 9 Jun 2002 14:01:27 +0000 (14:01 +0000)]
[project @ 2002-06-09 14:01:27 by matthewc]
Build cbits (not necessarily RTS) with -G0 on IA64, so GHCI doesn't have
to deal with sdata sections

21 years ago[project @ 2002-06-09 13:44:30 by matthewc]
matthewc [Sun, 9 Jun 2002 13:44:30 +0000 (13:44 +0000)]
[project @ 2002-06-09 13:44:30 by matthewc]
Don't SplitObjs on IA64 (not supported yet).

21 years ago[project @ 2002-06-09 13:40:53 by matthewc]
matthewc [Sun, 9 Jun 2002 13:40:53 +0000 (13:40 +0000)]
[project @ 2002-06-09 13:40:53 by matthewc]
IA64 doesn't have TABLES_NEXT_TO_CODE (despite being registerised).

21 years ago[project @ 2002-06-09 13:39:14 by matthewc]
matthewc [Sun, 9 Jun 2002 13:39:14 +0000 (13:39 +0000)]
[project @ 2002-06-09 13:39:14 by matthewc]
Recognise .subsection directive (IA64).

21 years ago[project @ 2002-06-09 13:37:41 by matthewc]
matthewc [Sun, 9 Jun 2002 13:37:44 +0000 (13:37 +0000)]
[project @ 2002-06-09 13:37:41 by matthewc]
Initial GHCI support on IA64.

21 years ago[project @ 2002-06-07 09:40:05 by matthewc]
matthewc [Fri, 7 Jun 2002 09:40:10 +0000 (09:40 +0000)]
[project @ 2002-06-07 09:40:05 by matthewc]
STG to machine register mapping for IA64.

21 years ago[project @ 2002-06-07 07:16:04 by chak]
chak [Fri, 7 Jun 2002 07:16:06 +0000 (07:16 +0000)]
[project @ 2002-06-07 07:16:04 by chak]
Fixed handling of infix operators in types:
- Pretty printing didn't take nested infix operators into account
- Explicit parenthesis were ignored in the fixity parser:
  * I added a constructor `HsParTy' to `HsType' (in the spirit of `HsPar' in
    `HsExpr'), which tracks the use of explicit parenthesis
  * Occurences of `HsParTy' in type-ish things that are not quite types (like
    context predicates) are removed in `ParseUtils'; all other occurences of
    `HsParTy' are removed during type checking (just as it works with `HsPar')

21 years ago[project @ 2002-06-06 07:48:47 by simonpj]
simonpj [Thu, 6 Jun 2002 07:48:48 +0000 (07:48 +0000)]
[project @ 2002-06-06 07:48:47 by simonpj]
Fix bogon in rebindable syntax implementation

21 years ago[project @ 2002-06-05 14:39:27 by simonpj]
simonpj [Wed, 5 Jun 2002 14:39:31 +0000 (14:39 +0000)]
[project @ 2002-06-05 14:39:27 by simonpj]
---------------------------------------
Add rebindable syntax for do-notation
(this time, on the HEAD)
---------------------------------------

Make do-notation use rebindable syntax, so that -fno-implicit-prelude
makes do-notation use whatever (>>=), (>>), return, fail are in scope,
rather than the Prelude versions.

On the way, combine HsDo and HsDoOut into one constructor in HsSyn,
and tidy up type checking of HsDo.

21 years ago[project @ 2002-06-05 14:08:23 by simonpj]
simonpj [Wed, 5 Jun 2002 14:08:25 +0000 (14:08 +0000)]
[project @ 2002-06-05 14:08:23 by simonpj]
------------------------------------------------
Fix the (new) lexer, and make the derived read
and show code work according to the new H98 report
------------------------------------------------

The new lexer, based on Koen's cunning parser (Text.ParserCombinators.ReadP)
wasn't quite right.  It's all very cool now.

In particular:

* The H98 "lex" function should return the exact string parsed, and it
now does, aided by the new combinator ReadP.gather.

* As a result the Text.Read.Lex Lexeme type is much simpler than before
    data Lexeme
      = Char   Char -- Quotes removed,
      | String String --  escapes interpreted
      | Punc   String  -- Punctuation, eg "(", "::"
      | Ident  String -- Haskell identifiers, e.g. foo, baz
      | Symbol String -- Haskell symbols, e.g. >>, %
      | Int Integer
      | Rat Rational
      | EOF
     deriving (Eq,Show)

* Multi-character punctuation, like "::" was getting lexed as a Symbol,
but it should be a Punc.

* Parsing numbers wasn't quite right.  "1..n" got it confused because it
got committed to a decimal point and then found a second '.'.

* The new H98 spec for Show is there, which ignores associativity.

21 years ago[project @ 2002-06-05 14:07:47 by simonpj]
simonpj [Wed, 5 Jun 2002 14:07:47 +0000 (14:07 +0000)]
[project @ 2002-06-05 14:07:47 by simonpj]
Add comment

21 years ago[project @ 2002-06-05 13:30:52 by simonpj]
simonpj [Wed, 5 Jun 2002 13:30:53 +0000 (13:30 +0000)]
[project @ 2002-06-05 13:30:52 by simonpj]
Documentation for rebindable do-notation

21 years ago[project @ 2002-06-04 19:21:28 by sof]
sof [Tue, 4 Jun 2002 19:21:28 +0000 (19:21 +0000)]
[project @ 2002-06-04 19:21:28 by sof]
cygwin support.

21 years ago[project @ 2002-06-04 19:17:57 by sof]
sof [Tue, 4 Jun 2002 19:17:57 +0000 (19:17 +0000)]
[project @ 2002-06-04 19:17:57 by sof]
Three way split of HSbase + adjust pointed fingers in the direction
of the real culprit here (the GNU linker, not the file format.)

21 years ago[project @ 2002-06-04 18:17:50 by sof]
sof [Tue, 4 Jun 2002 18:17:50 +0000 (18:17 +0000)]
[project @ 2002-06-04 18:17:50 by sof]
GhcWithInterpreter: thumbs up for cygwin

21 years ago[project @ 2002-06-04 18:09:00 by sof]
sof [Tue, 4 Jun 2002 18:09:00 +0000 (18:09 +0000)]
[project @ 2002-06-04 18:09:00 by sof]
loadDynamic: both mingw and cygwin slurp DLLs

21 years ago[project @ 2002-06-04 16:13:53 by sof]
sof [Tue, 4 Jun 2002 16:13:53 +0000 (16:13 +0000)]
[project @ 2002-06-04 16:13:53 by sof]
Given how CPP works, the prev. commit does not cut the mustard.

Reduce the risk of name clashes by giving the #defines some more
obscure names instead.

21 years ago[project @ 2002-06-04 12:01:38 by simonpj]
simonpj [Tue, 4 Jun 2002 12:01:38 +0000 (12:01 +0000)]
[project @ 2002-06-04 12:01:38 by simonpj]
Make the type-var-default warning more informative

21 years ago[project @ 2002-06-03 23:36:40 by sof]
sof [Mon, 3 Jun 2002 23:36:40 +0000 (23:36 +0000)]
[project @ 2002-06-03 23:36:40 by sof]
whether or not to quote -D and -U opts is a property of the host, not target OS

21 years ago[project @ 2002-06-03 13:08:37 by matthewc]
matthewc [Mon, 3 Jun 2002 13:08:41 +0000 (13:08 +0000)]
[project @ 2002-06-03 13:08:37 by matthewc]
Initial mangling and tailcalls support for IA64.

Function prologues and epilogues are deleted and we use a single register
stack frame throughout (with a little register renaming in the mangler...)
Dropthrough from fast to slow entry point is also implemented.
Tailcalls are marked and converted into jumps at mangle time.

21 years ago[project @ 2002-06-03 12:56:41 by matthewc]
matthewc [Mon, 3 Jun 2002 12:56:41 +0000 (12:56 +0000)]
[project @ 2002-06-03 12:56:41 by matthewc]
Remove pointless macro INIT_VECTOR

21 years ago[project @ 2002-06-03 11:31:55 by simonmar]
simonmar [Mon, 3 Jun 2002 11:31:55 +0000 (11:31 +0000)]
[project @ 2002-06-03 11:31:55 by simonmar]
#undef C & R after using them, we don't want to pollute C's namespace
with one-character #defines...

21 years ago[project @ 2002-06-03 10:27:11 by simonmar]
simonmar [Mon, 3 Jun 2002 10:27:11 +0000 (10:27 +0000)]
[project @ 2002-06-03 10:27:11 by simonmar]
Fix buglet with --force, spotted by George Russell

21 years ago[project @ 2002-05-31 21:12:28 by sof]
sof [Fri, 31 May 2002 21:12:29 +0000 (21:12 +0000)]
[project @ 2002-05-31 21:12:28 by sof]
added struct msghdr feature tests

21 years ago[project @ 2002-05-29 22:39:39 by sof]
sof [Wed, 29 May 2002 22:39:39 +0000 (22:39 +0000)]
[project @ 2002-05-29 22:39:39 by sof]
- nuke ghc_{501,502}_at_least defns, no longer used.
- mingw/cygwin cross-compilation tweak.

21 years ago[project @ 2002-05-29 22:32:49 by sof]
sof [Wed, 29 May 2002 22:32:50 +0000 (22:32 +0000)]
[project @ 2002-05-29 22:32:49 by sof]
Use HOSTPLATFORM / *_HOST_OS to examine properties of the environment/compiler
we're using.

21 years ago[project @ 2002-05-29 22:11:58 by sof]
sof [Wed, 29 May 2002 22:11:59 +0000 (22:11 +0000)]
[project @ 2002-05-29 22:11:58 by sof]
Use HOSTPLATFORM / *_HOST_OS (instead of TARGET equivalents), as we're
dealing with host-specific properties, not that of the target.

21 years ago[project @ 2002-05-29 13:44:18 by simonmar]
simonmar [Wed, 29 May 2002 13:44:18 +0000 (13:44 +0000)]
[project @ 2002-05-29 13:44:18 by simonmar]
gcc 3.1 broke the mangler again...  this time it seems gcc is adding
spurious writes to the stack in the prologue, triggered perhaps by
inline functions.  The code generated for uses of ASSIGN_DBL() and
ASSIGN_INT64() is really terrible - we should really submit a bug
report to the gcc folks for this one.

Anyway, this patch should get us going again; we now toss the spurious
instructions back into the main part of the code (just in case they
happened to be there for a good reason).

21 years ago[project @ 2002-05-29 13:09:24 by simonpj]
simonpj [Wed, 29 May 2002 13:09:24 +0000 (13:09 +0000)]
[project @ 2002-05-29 13:09:24 by simonpj]
Wibble to previous commit on negative literals in patterns (fixes crash in GHC/Real.lhs)

21 years ago[project @ 2002-05-29 13:06:02 by simonpj]
simonpj [Wed, 29 May 2002 13:06:02 +0000 (13:06 +0000)]
[project @ 2002-05-29 13:06:02 by simonpj]
Document GHC-mingw vs GHC-cygwin

21 years ago[project @ 2002-05-28 21:52:06 by sof]
sof [Tue, 28 May 2002 21:52:06 +0000 (21:52 +0000)]
[project @ 2002-05-28 21:52:06 by sof]
setMode: permit (redundant) setting of same mode flag

21 years ago[project @ 2002-05-28 09:22:08 by wolfgang]
wolfgang [Tue, 28 May 2002 09:22:08 +0000 (09:22 +0000)]
[project @ 2002-05-28 09:22:08 by wolfgang]
Preliminary support for GCC 3.1 on MacOS X

21 years ago[project @ 2002-05-28 08:38:19 by simonpj]
simonpj [Tue, 28 May 2002 08:38:19 +0000 (08:38 +0000)]
[project @ 2002-05-28 08:38:19 by simonpj]
Fix assert

21 years ago[project @ 2002-05-27 16:13:42 by simonpj]
simonpj [Mon, 27 May 2002 16:13:43 +0000 (16:13 +0000)]
[project @ 2002-05-27 16:13:42 by simonpj]
Make negative literals work in patterns

The issue here is that

f (-1) = True
f x    = False

should generate

f x = x == negate (fromInteger 1)

rather than

f x = x == fromInteger (-1)

21 years ago[project @ 2002-05-27 16:13:32 by simonmar]
simonmar [Mon, 27 May 2002 16:13:32 +0000 (16:13 +0000)]
[project @ 2002-05-27 16:13:32 by simonmar]
Remove u2i, it was identical to getKey.

21 years ago[project @ 2002-05-27 16:12:35 by simonmar]
simonmar [Mon, 27 May 2002 16:13:05 +0000 (16:13 +0000)]
[project @ 2002-05-27 16:12:35 by simonmar]
Use Unique.getKey instead of Unique.u2i (they are the same function).

21 years ago[project @ 2002-05-27 16:11:42 by simonmar]
simonmar [Mon, 27 May 2002 16:11:42 +0000 (16:11 +0000)]
[project @ 2002-05-27 16:11:42 by simonmar]
Update copyright

21 years ago[project @ 2002-05-27 15:28:07 by simonpj]
simonpj [Mon, 27 May 2002 15:28:09 +0000 (15:28 +0000)]
[project @ 2002-05-27 15:28:07 by simonpj]
Allow infix type constructors

This commit adds infix type constructors (but not yet class constructors).
The documentation describes what should be the case.  Lots of tiresome
changes, but nothing exciting.

Allows infix type constructors everwhere a type can occur, and in a data
or type synonym decl.  E.g.

data a :*: b = ....

You can give fixity decls for type constructors, but the fixity decl
applies both to the tycon and the corresponding data con.

21 years ago[project @ 2002-05-27 10:20:20 by simonpj]
simonpj [Mon, 27 May 2002 10:20:20 +0000 (10:20 +0000)]
[project @ 2002-05-27 10:20:20 by simonpj]
Build TyDecl and ConDecl with the RdrHsSyn constructors, so the SysNames are right

21 years ago[project @ 2002-05-24 09:52:12 by simonpj]
simonpj [Fri, 24 May 2002 09:52:12 +0000 (09:52 +0000)]
[project @ 2002-05-24 09:52:12 by simonpj]
speakNth works for 11,12,13

21 years ago[project @ 2002-05-23 15:51:26 by simonpj]
simonpj [Thu, 23 May 2002 15:51:26 +0000 (15:51 +0000)]
[project @ 2002-05-23 15:51:26 by simonpj]
Don't report ambiguity errors
if other type errors have happened

This saves a gratuitous error cascade when the type checker
recovers from one error by giving f type (forall a.a), and
then find an ambiguity problem as a direct result.

21 years ago[project @ 2002-05-23 15:49:07 by simonpj]
simonpj [Thu, 23 May 2002 15:49:07 +0000 (15:49 +0000)]
[project @ 2002-05-23 15:49:07 by simonpj]
Bug-fix for validity checking of type synonyms

21 years ago[project @ 2002-05-23 15:37:32 by simonpj]
simonpj [Thu, 23 May 2002 15:37:32 +0000 (15:37 +0000)]
[project @ 2002-05-23 15:37:32 by simonpj]
Reject type arguments gracefully

21 years ago[project @ 2002-05-22 18:36:19 by sof]
sof [Wed, 22 May 2002 18:36:19 +0000 (18:36 +0000)]
[project @ 2002-05-22 18:36:19 by sof]
HC_SPLIT_POST: reduce the risk of exceeding command line length limits.

[As was, constructing the relocatable ".o"s from split object files
failed with stuff inside of libraries/haskell-src/Language/Haskell/
under cygwin(&bash).]

21 years ago[project @ 2002-05-22 09:02:40 by simonmar]
simonmar [Wed, 22 May 2002 09:02:41 +0000 (09:02 +0000)]
[project @ 2002-05-22 09:02:40 by simonmar]
We don't pass --old-ghc to Happy any more.  Also, wire $(HappyOpts) to
-agc in config.mk rather than passing it through from configure, since
it was always the same anyhow.

21 years ago[project @ 2002-05-21 14:58:49 by wolfgang]
wolfgang [Tue, 21 May 2002 14:58:51 +0000 (14:58 +0000)]
[project @ 2002-05-21 14:58:49 by wolfgang]
Bug-fixes for PowerPC
*) Rewrote StgRun in assembler (that seems to be the cleanest way to get it to work).
*) Fixed a small typo in createAdjustor.

21 years ago[project @ 2002-05-21 13:43:59 by simonpj]
simonpj [Tue, 21 May 2002 13:43:59 +0000 (13:43 +0000)]
[project @ 2002-05-21 13:43:59 by simonpj]
Parse External Core correctly

Amazingly, recursive bindings with only one binding in the
group were being parsed as non-recursive.

21 years ago[project @ 2002-05-18 05:28:14 by ken]
ken [Sat, 18 May 2002 05:28:15 +0000 (05:28 +0000)]
[project @ 2002-05-18 05:28:14 by ken]
Add missing va_end's to va_start's.

21 years ago[project @ 2002-05-17 09:03:13 by simonmar]
simonmar [Fri, 17 May 2002 09:03:13 +0000 (09:03 +0000)]
[project @ 2002-05-17 09:03:13 by simonmar]
If we have a Happy version 1.13 or later, turn on the --old-ghc option
for now.

21 years ago[project @ 2002-05-17 08:21:51 by simonpj]
simonpj [Fri, 17 May 2002 08:21:51 +0000 (08:21 +0000)]
[project @ 2002-05-17 08:21:51 by simonpj]
Wibble

21 years ago[project @ 2002-05-17 08:13:08 by simonpj]
simonpj [Fri, 17 May 2002 08:13:08 +0000 (08:13 +0000)]
[project @ 2002-05-17 08:13:08 by simonpj]
Win32 updates from Simon

21 years ago[project @ 2002-05-15 20:25:10 by sof]
sof [Wed, 15 May 2002 20:25:10 +0000 (20:25 +0000)]
[project @ 2002-05-15 20:25:10 by sof]
make prev commit work

21 years ago[project @ 2002-05-15 19:46:11 by sof]
sof [Wed, 15 May 2002 19:46:11 +0000 (19:46 +0000)]
[project @ 2002-05-15 19:46:11 by sof]
Assume the presence of a Bits Int32 instance only

21 years ago[project @ 2002-05-15 18:20:28 by sof]
sof [Wed, 15 May 2002 18:20:28 +0000 (18:20 +0000)]
[project @ 2002-05-15 18:20:28 by sof]
add Main.hi to CLEAN_FILES to cope with the fact that HaskTags.hs is rooming the Main module

21 years ago[project @ 2002-05-15 08:59:58 by chak]
chak [Wed, 15 May 2002 09:00:00 +0000 (09:00 +0000)]
[project @ 2002-05-15 08:59:58 by chak]
* Added the options `-ffi' and `-fffi', which switch on FFI support
  (`-fglasgow-exts' implies `-fffi').

  NB: Strictly speaking `-fffi' is the correct name; however, `-ffi' is
      supported as an alternative spelling.

* `-fglasgow-exts' no longer enables the `with' keyword for implicit
  parameters.  To use `with' as a keyword, the additional option `-fwith' has
  to be supplied.  (Note that SimonM recently enabled the use of `let' instead
  of `with'.)

  NB: This might prompt some makefile tweaks in libraries or regression tests.

* Renamed `Foreign.Marshal.Utils.withObject' to `Foreign.Marshal.Utils.with'
  as required by the FFI Addendum.  (The old name is still available, but
  marked as deprecated.)

* Added `realloc' to `Foreign.Marshal.Alloc' (tracking RC4 of the FFI
  Addendum).

Docu
~~~~
* Added `-ffi', `-fffi', and `-fwith' to the flag reference and the section
  describing options for Glasgow extensions

* Removed most of the FFI section in the User's Guide in favour of a reference
  to the Addendum plus a brief description of additional features supported by
  GHC.

  NB: The old material is still available at fptools/docs/, I merely removed
  the reference in the User's Guide.

21 years ago[project @ 2002-05-14 08:30:34 by matthewc]
matthewc [Tue, 14 May 2002 08:30:34 +0000 (08:30 +0000)]
[project @ 2002-05-14 08:30:34 by matthewc]
* Note that ia64-linux works unregisterised and a registerised port is in
  progress.
* Recommend bootstrapping from a machine of the same word size :)

21 years ago[project @ 2002-05-14 08:27:11 by matthewc]
matthewc [Tue, 14 May 2002 08:27:11 +0000 (08:27 +0000)]
[project @ 2002-05-14 08:27:11 by matthewc]
Make adjustor unavailability on a particular platform a runtime barf
instead of an #error (to allow the remainder of ghc to function without
architecture-specific support)

21 years ago[project @ 2002-05-14 08:25:46 by matthewc]
matthewc [Tue, 14 May 2002 08:25:46 +0000 (08:25 +0000)]
[project @ 2002-05-14 08:25:46 by matthewc]
Bootstrapping fixes:
* Correct includes and libraries to use new libraries/ tree
* Check UseGhcForCc for "YES" instead of ""

21 years ago[project @ 2002-05-14 08:23:11 by matthewc]
matthewc [Tue, 14 May 2002 08:23:11 +0000 (08:23 +0000)]
[project @ 2002-05-14 08:23:11 by matthewc]
Need #include <stdlib.h> for malloc (otherwise the C compiler will
assume the return value is an int, with bad results)

21 years ago[project @ 2002-05-14 08:17:38 by matthewc]
matthewc [Tue, 14 May 2002 08:17:38 +0000 (08:17 +0000)]
[project @ 2002-05-14 08:17:38 by matthewc]
Round up GMP allocation requests to a whole number of StgWords
(GMP limbs are not necessarily an StgWord in size)

21 years ago[project @ 2002-05-14 08:15:49 by matthewc]
matthewc [Tue, 14 May 2002 08:15:49 +0000 (08:15 +0000)]
[project @ 2002-05-14 08:15:49 by matthewc]
IA64 only changes:
* Place HEAP_BASE in an appropriate place
* On IA64, use GCC "section" attribute to coax compiler into putting info
  tables in .text

21 years ago[project @ 2002-05-14 08:09:07 by matthewc]
matthewc [Tue, 14 May 2002 08:09:07 +0000 (08:09 +0000)]
[project @ 2002-05-14 08:09:07 by matthewc]
* Add ia64-*-linux* as a recognised system type
* Standardise capitalisation of YES and NO in HaveLibGmp
* Error out if user tries to use in-tree GMP on IA64 (results in hangs
  and segfaults - it probably needs an upgrade sometime)

21 years ago[project @ 2002-05-11 13:58:18 by sof]
sof [Sat, 11 May 2002 13:58:18 +0000 (13:58 +0000)]
[project @ 2002-05-11 13:58:18 by sof]
Have createThread() use a separate mutex to ensure unique
allocation of ThreadIds. A less-than-light solution, but
cleaner than trying to reuse sched_mutex for this purpose.

21 years ago[project @ 2002-05-11 10:02:21 by panne]
panne [Sat, 11 May 2002 10:02:22 +0000 (10:02 +0000)]
[project @ 2002-05-11 10:02:21 by panne]
Include hschooks.h via a global option to get ghc_strlen's
prototype. It's too omnipresent for per-file OPTIONS.

21 years ago[project @ 2002-05-11 00:16:11 by sof]
sof [Sat, 11 May 2002 00:16:12 +0000 (00:16 +0000)]
[project @ 2002-05-11 00:16:11 by sof]
As a result of calling exitScheduler(), enter into a
'shutting-down' state, so that the subsequent (sequential)
execution of finalisers won't get stuck inside the
Scheduler waiting for more work.

Cleanest way I could think of solving this problem on a
Friday afternoon.

21 years ago[project @ 2002-05-10 22:02:02 by sof]
sof [Fri, 10 May 2002 22:02:02 +0000 (22:02 +0000)]
[project @ 2002-05-10 22:02:02 by sof]
stopTaskManager(): no seppuku, please.

21 years ago[project @ 2002-05-10 20:44:29 by panne]
panne [Fri, 10 May 2002 20:44:39 +0000 (20:44 +0000)]
[project @ 2002-05-10 20:44:29 by panne]
Re-enable bootstrapping: More Ptr trouble, now that it's (almost) abstract...

21 years ago[project @ 2002-05-10 13:34:18 by simonpj]
simonpj [Fri, 10 May 2002 13:34:18 +0000 (13:34 +0000)]
[project @ 2002-05-10 13:34:18 by simonpj]
Remove {| |} syntax from expressions; never fully implemented

21 years ago[project @ 2002-05-10 12:43:02 by simonpj]
simonpj [Fri, 10 May 2002 12:43:02 +0000 (12:43 +0000)]
[project @ 2002-05-10 12:43:02 by simonpj]
If a type signature mentions a type variable that doesn't appear in
the type, GHC was dying horribly. Example (from happy -s):

type T a = () -> ()

f :: T a
f () = ()

This commit fixes the problem.

22 years ago[project @ 2002-05-08 07:26:36 by sof]
sof [Wed, 8 May 2002 07:26:36 +0000 (07:26 +0000)]
[project @ 2002-05-08 07:26:36 by sof]
summarise{File}: filter out GHC.Prim from import lists