ghc-hetmet.git
23 years ago[project @ 2001-04-19 15:58:56 by sewardj]
sewardj [Thu, 19 Apr 2001 15:58:56 +0000 (15:58 +0000)]
[project @ 2001-04-19 15:58:56 by sewardj]
merge revision 1.52.2.1

23 years ago[project @ 2001-04-18 11:58:02 by sewardj]
sewardj [Wed, 18 Apr 2001 11:58:02 +0000 (11:58 +0000)]
[project @ 2001-04-18 11:58:02 by sewardj]
include ../includes/config.h

23 years ago[project @ 2001-04-18 07:49:07 by michaelw]
michaelw [Wed, 18 Apr 2001 07:49:07 +0000 (07:49 +0000)]
[project @ 2001-04-18 07:49:07 by michaelw]
readded Debian's old SGML catalog path for backward compat with Debian-2.2

23 years ago[project @ 2001-04-18 01:49:22 by dsyme]
dsyme [Wed, 18 Apr 2001 01:49:23 +0000 (01:49 +0000)]
[project @ 2001-04-18 01:49:22 by dsyme]
More ILX backend fixes from Don Syme

23 years ago[project @ 2001-04-17 22:11:00 by lewie]
lewie [Tue, 17 Apr 2001 22:11:00 +0000 (22:11 +0000)]
[project @ 2001-04-17 22:11:00 by lewie]
Stomp some debug output.

23 years ago[project @ 2001-04-17 21:44:16 by lewie]
lewie [Tue, 17 Apr 2001 21:44:16 +0000 (21:44 +0000)]
[project @ 2001-04-17 21:44:16 by lewie]
This is the test I meant to put in ... ;-)

23 years ago[project @ 2001-04-17 21:41:29 by michaelw]
michaelw [Tue, 17 Apr 2001 21:41:29 +0000 (21:41 +0000)]
[project @ 2001-04-17 21:41:29 by michaelw]
* added autoconf macro that compares dot-separated version
  numbers
* adjusted happy version checking
* ./configure now happy with CVS-happy's [1.10]  ;)

23 years ago[project @ 2001-04-17 16:00:48 by rrt]
rrt [Tue, 17 Apr 2001 16:00:48 +0000 (16:00 +0000)]
[project @ 2001-04-17 16:00:48 by rrt]
Add m4 to the required packages list.

23 years ago[project @ 2001-04-17 15:55:33 by rrt]
rrt [Tue, 17 Apr 2001 15:55:33 +0000 (15:55 +0000)]
[project @ 2001-04-17 15:55:33 by rrt]
Update the rather old instructions on installing ancillary tools on Windows.

23 years ago[project @ 2001-04-17 06:40:07 by lewie]
lewie [Tue, 17 Apr 2001 06:40:07 +0000 (06:40 +0000)]
[project @ 2001-04-17 06:40:07 by lewie]
Woops.  Wrong example.

23 years ago[project @ 2001-04-17 06:14:59 by lewie]
lewie [Tue, 17 Apr 2001 06:14:59 +0000 (06:14 +0000)]
[project @ 2001-04-17 06:14:59 by lewie]
Added test for parallel lets in parallel list comprehensions.

23 years ago[project @ 2001-04-16 03:01:04 by chak]
chak [Mon, 16 Apr 2001 03:01:04 +0000 (03:01 +0000)]
[project @ 2001-04-16 03:01:04 by chak]
* Adapted to GHC 5.00.  By default an interactive system is built.

* Automagically runs autoconf if `configure' is missing from the source tar
  ball.

23 years ago[project @ 2001-04-16 01:30:16 by chak]
chak [Mon, 16 Apr 2001 01:30:16 +0000 (01:30 +0000)]
[project @ 2001-04-16 01:30:16 by chak]
$(bindir) has to be in place before creating symbolic links (otherwise, rpms
cannot be built with the BUILD_ROOT feature).

23 years ago[project @ 2001-04-14 22:32:14 by qrczak]
qrczak [Sat, 14 Apr 2001 22:32:14 +0000 (22:32 +0000)]
[project @ 2001-04-14 22:32:14 by qrczak]
{-# INLINE instance #-} support.

23 years ago[project @ 2001-04-14 22:29:16 by qrczak]
qrczak [Sat, 14 Apr 2001 22:29:16 +0000 (22:29 +0000)]
[project @ 2001-04-14 22:29:16 by qrczak]
Add stg_ap_1_upd_info to symbols known to the linker.

23 years ago[project @ 2001-04-14 22:28:46 by qrczak]
qrczak [Sat, 14 Apr 2001 22:28:46 +0000 (22:28 +0000)]
[project @ 2001-04-14 22:28:46 by qrczak]
castCCharToChar can be more straightforward and more efficient now.

23 years ago[project @ 2001-04-14 22:28:22 by qrczak]
qrczak [Sat, 14 Apr 2001 22:28:22 +0000 (22:28 +0000)]
[project @ 2001-04-14 22:28:22 by qrczak]
Cosmetics.

23 years ago[project @ 2001-04-14 22:27:34 by qrczak]
qrczak [Sat, 14 Apr 2001 22:27:34 +0000 (22:27 +0000)]
[project @ 2001-04-14 22:27:34 by qrczak]
Don't rely so much on exports from ArrayBase.

23 years ago[project @ 2001-04-14 22:27:00 by qrczak]
qrczak [Sat, 14 Apr 2001 22:27:00 +0000 (22:27 +0000)]
[project @ 2001-04-14 22:27:00 by qrczak]
Implementation of arrays rewritten
----------------------------------

Bulk operations like listArray, elems, fmap/amap, (==), getElems,
getAssocs, freeze etc. no longer check whether indices which are
not provided by the programmer are in bounds (they always are), and
avoid unnecessary translation between Ix indices and Int indices.
Some operations are implemented more efficiently, most notably (==)
and compare.

This applies to all IArray and MArray instances, including Haskell
98 Arrays.

Old methods of IArray and MArray are now functions; this is the only
change in the interface. New methods are exported only by ArrayBase,
i.e. not officially exported. They work on Int indices and are unsafe:
they don't do bounds checks themselves. Public functions do checks
and index translation instead where necessary.

More is inlined, to ensure that anything worth specialization or list
fusion gets specialized and fused. Perhaps a bit too much is inlined.
If it was possible to say that a function should be instantiated in
other modules for each type but not inlined on each use, it would be
useful here.

Using UArray Int Char wrapped in a nice interface (not included
here) instead of PackedString should be reasonable. PackedStrings
are 10% faster than UArray in a sorting test (and don't support
Unicode). Standard Strings are 50% slower (and take up more memory),
even though other test versions convert input from standard Strings
and convert output to them. ByteArrays tuned by hand for the benchmark
are 15% faster. The same UArray test compiled with released ghc-5.00,
with compare defined in terms of assocs, is 7 times slower.

23 years ago[project @ 2001-04-14 22:25:24 by qrczak]
qrczak [Sat, 14 Apr 2001 22:25:24 +0000 (22:25 +0000)]
[project @ 2001-04-14 22:25:24 by qrczak]
Do more Word arithmetic on constants at compile time.

23 years ago[project @ 2001-04-14 22:24:24 by qrczak]
qrczak [Sat, 14 Apr 2001 22:24:24 +0000 (22:24 +0000)]
[project @ 2001-04-14 22:24:24 by qrczak]
Add {-# INLINE instance #-} pragma which ensures that the dictionary
function is inlined.

23 years ago[project @ 2001-04-14 07:36:58 by lewie]
lewie [Sat, 14 Apr 2001 07:36:58 +0000 (07:36 +0000)]
[project @ 2001-04-14 07:36:58 by lewie]
Fix a couple of bugs affecting implicit params.  With implicit params, you
can have a rho type w/o being a forall type.

23 years ago[project @ 2001-04-13 21:37:42 by panne]
panne [Fri, 13 Apr 2001 21:37:43 +0000 (21:37 +0000)]
[project @ 2001-04-13 21:37:42 by panne]
First steps toward a better typing of f.e.d. and friends: Make FunPtr
a fully-fledged data type, not a renaming for Ptr. This is necessary,
because the FFI "looks through" newtypes, which we don't want in this
case.

23 years ago[project @ 2001-04-13 17:49:23 by lewie]
lewie [Fri, 13 Apr 2001 17:49:23 +0000 (17:49 +0000)]
[project @ 2001-04-13 17:49:23 by lewie]
Add new test for IP.

23 years ago[project @ 2001-04-13 13:37:24 by panne]
panne [Fri, 13 Apr 2001 13:37:24 +0000 (13:37 +0000)]
[project @ 2001-04-13 13:37:24 by panne]
To keep people debugging GHC sane, disable CSE in *every* module using
GLOBAL_VARs. This solves the problem with the strange -M output, where
some global IORefs were commoned up (again). CSE seems to be really
broken, but a comment in ghc/Makefile promises a fix.  Anybody out
there with this fix on his/her hard disk: Please commit soon!

23 years ago[project @ 2001-04-13 03:50:52 by dsyme]
dsyme [Fri, 13 Apr 2001 03:50:52 +0000 (03:50 +0000)]
[project @ 2001-04-13 03:50:52 by dsyme]
More changes to the ILX code generator, currently only relevant to Don

23 years ago[project @ 2001-04-12 21:29:43 by lewie]
lewie [Thu, 12 Apr 2001 21:29:43 +0000 (21:29 +0000)]
[project @ 2001-04-12 21:29:43 by lewie]
Don't use the same simplify code for both restricted and unrestricted
bindings.  In particular, a restricted binding shouldn't try to capture
implicit params.

23 years ago[project @ 2001-04-12 21:20:28 by lewie]
lewie [Thu, 12 Apr 2001 21:20:28 +0000 (21:20 +0000)]
[project @ 2001-04-12 21:20:28 by lewie]
Export a useful predicate.

23 years ago[project @ 2001-04-12 15:17:17 by michaelw]
michaelw [Thu, 12 Apr 2001 15:17:17 +0000 (15:17 +0000)]
[project @ 2001-04-12 15:17:17 by michaelw]
clean-up Main.hi

23 years ago[project @ 2001-04-12 15:16:02 by michaelw]
michaelw [Thu, 12 Apr 2001 15:16:02 +0000 (15:16 +0000)]
[project @ 2001-04-12 15:16:02 by michaelw]
Debian updates

23 years ago[project @ 2001-04-11 18:42:07 by lewie]
lewie [Wed, 11 Apr 2001 18:42:07 +0000 (18:42 +0000)]
[project @ 2001-04-11 18:42:07 by lewie]
I guess it's supposed to be done *this* way...

23 years ago[project @ 2001-04-11 18:23:32 by lewie]
lewie [Wed, 11 Apr 2001 18:23:32 +0000 (18:23 +0000)]
[project @ 2001-04-11 18:23:32 by lewie]
Another implicit param test.

23 years ago[project @ 2001-04-11 18:10:16 by lewie]
lewie [Wed, 11 Apr 2001 18:10:16 +0000 (18:10 +0000)]
[project @ 2001-04-11 18:10:16 by lewie]
Add implicit param test.

23 years ago[project @ 2001-04-11 18:09:59 by lewie]
lewie [Wed, 11 Apr 2001 18:09:59 +0000 (18:09 +0000)]
[project @ 2001-04-11 18:09:59 by lewie]
Add `-w' to OPTIONS so the test will pass.

23 years ago[project @ 2001-04-11 10:41:46 by sewardj]
sewardj [Wed, 11 Apr 2001 10:41:46 +0000 (10:41 +0000)]
[project @ 2001-04-11 10:41:46 by sewardj]
GhcBinDistLinks should include ghc-pkg, not ghci-pkg.

23 years ago[project @ 2001-04-11 10:24:38 by sewardj]
sewardj [Wed, 11 Apr 2001 10:24:38 +0000 (10:24 +0000)]
[project @ 2001-04-11 10:24:38 by sewardj]
ghci/ByteCodeLink_HC_OPTS = -fvia-C    -- sparc NCG seems to need it.

23 years ago[project @ 2001-04-10 22:34:47 by lewie]
lewie [Tue, 10 Apr 2001 22:34:47 +0000 (22:34 +0000)]
[project @ 2001-04-10 22:34:47 by lewie]
Don't use foldr/build if we've got a parallel list comp.  We'd need to
have/generate a family of functions foldr{2,3,...} to do anything nicer here.

23 years ago[project @ 2001-04-10 16:30:46 by sewardj]
sewardj [Tue, 10 Apr 2001 16:30:46 +0000 (16:30 +0000)]
[project @ 2001-04-10 16:30:46 by sewardj]
Fix bogon which gives incorrect setting for GhcWithInterpreter on solaris2.

23 years ago[project @ 2001-04-10 15:54:31 by qrczak]
qrczak [Tue, 10 Apr 2001 15:54:31 +0000 (15:54 +0000)]
[project @ 2001-04-10 15:54:31 by qrczak]
Typo.

23 years ago[project @ 2001-04-10 13:52:31 by sewardj]
sewardj [Tue, 10 Apr 2001 13:52:31 +0000 (13:52 +0000)]
[project @ 2001-04-10 13:52:31 by sewardj]
Make installation of documentation in binary dists work properly.
Also install documentation by default for make install.

23 years ago[project @ 2001-04-09 23:22:42 by lewie]
lewie [Mon, 9 Apr 2001 23:22:42 +0000 (23:22 +0000)]
[project @ 2001-04-09 23:22:42 by lewie]
Tag id of implicit param has having a PredType.

23 years ago[project @ 2001-04-09 09:22:00 by sewardj]
sewardj [Mon, 9 Apr 2001 09:22:00 +0000 (09:22 +0000)]
[project @ 2001-04-09 09:22:00 by sewardj]
Mention fundeps.

23 years ago[project @ 2001-04-08 18:56:31 by panne]
panne [Sun, 8 Apr 2001 18:56:32 +0000 (18:56 +0000)]
[project @ 2001-04-08 18:56:31 by panne]
Updated my email address

23 years ago[project @ 2001-04-07 22:31:11 by qrczak]
qrczak [Sat, 7 Apr 2001 22:31:11 +0000 (22:31 +0000)]
[project @ 2001-04-07 22:31:11 by qrczak]
Mention {Diff,Storable}Array and Monad{List,Cont}.

Aren't working fundeps new in ghc-5.00? They are not mentioned in
release notes.

23 years ago[project @ 2001-04-07 22:30:01 by qrczak]
qrczak [Sat, 7 Apr 2001 22:30:01 +0000 (22:30 +0000)]
[project @ 2001-04-07 22:30:01 by qrczak]
Append final newline to package.conf.

23 years ago[project @ 2001-04-07 14:58:44 by lewie]
lewie [Sat, 7 Apr 2001 14:58:44 +0000 (14:58 +0000)]
[project @ 2001-04-07 14:58:44 by lewie]
Sigh... the moving of the structs..  `struct tm' is only in <time.h> on
RedHat 7.1/whatever-version-of-glibc-or-whatever-that-implies.  The following
patch works on RH 6.2, RH 7.0 and RH 7.1.  Your mileage shouldn't vary,
but you know how these things go..

23 years ago[project @ 2001-04-07 09:13:55 by qrczak]
qrczak [Sat, 7 Apr 2001 09:13:55 +0000 (09:13 +0000)]
[project @ 2001-04-07 09:13:55 by qrczak]
Fix typos and change version to 5.00 in banners.

23 years ago[project @ 2001-04-06 16:15:39 by sewardj]
sewardj [Fri, 6 Apr 2001 16:15:39 +0000 (16:15 +0000)]
[project @ 2001-04-06 16:15:39 by sewardj]
Commit the following change:

> > Why the default libdir is /usr/local/lib
> > and not /usr/local/lib/ghc-<version>?
>
> Great question.  I end up running config with the likes of
> `--libdir=/usr/local/lib/ghc-5.0' every time, which gets to
> be annoying ;-)

I've been meaning to fix this for a while, but couldn't see a good way
to do it.  I found a (mildly-hacky) way to do it today: in
fptools/ghc, we override $(libdir) to be
$(libdir)/$(ProjectNameShort)-$(ProjectVersion), so everything inside
fptools/ghc will be installed in the subdirectory.  fptools/hslibs is
a bit more of a hack.

23 years ago[project @ 2001-04-06 10:53:08 by sewardj]
sewardj [Fri, 6 Apr 2001 10:53:08 +0000 (10:53 +0000)]
[project @ 2001-04-06 10:53:08 by sewardj]
wibble

23 years ago[project @ 2001-04-06 04:28:53 by lewie]
lewie [Fri, 6 Apr 2001 04:28:53 +0000 (04:28 +0000)]
[project @ 2001-04-06 04:28:53 by lewie]
More Version Blues.

23 years ago[project @ 2001-04-06 01:07:23 by lewie]
lewie [Fri, 6 Apr 2001 01:07:23 +0000 (01:07 +0000)]
[project @ 2001-04-06 01:07:23 by lewie]
Fix a bad case of the Version Change Blues ;-)

23 years ago[project @ 2001-04-05 14:45:29 by sewardj]
sewardj [Thu, 5 Apr 2001 14:45:29 +0000 (14:45 +0000)]
[project @ 2001-04-05 14:45:29 by sewardj]
Denouncements for 5.0.

23 years ago[project @ 2001-04-05 13:48:22 by sewardj]
sewardj [Thu, 5 Apr 2001 13:48:22 +0000 (13:48 +0000)]
[project @ 2001-04-05 13:48:22 by sewardj]
Now 5.00.

23 years ago[project @ 2001-04-05 13:47:27 by sewardj]
sewardj [Thu, 5 Apr 2001 13:47:27 +0000 (13:47 +0000)]
[project @ 2001-04-05 13:47:27 by sewardj]
Advance the version number to 5.00.  Wahey!  Not that this means we're
official _at_ 5.0 yet, tho.

23 years ago[project @ 2001-04-05 11:54:37 by simonpj]
simonpj [Thu, 5 Apr 2001 11:54:37 +0000 (11:54 +0000)]
[project @ 2001-04-05 11:54:37 by simonpj]
Make type synonyms work right in H98

23 years ago[project @ 2001-04-05 11:48:59 by simonpj]
simonpj [Thu, 5 Apr 2001 11:48:59 +0000 (11:48 +0000)]
[project @ 2001-04-05 11:48:59 by simonpj]
Add tc120

23 years ago[project @ 2001-04-05 11:48:26 by simonpj]
simonpj [Thu, 5 Apr 2001 11:48:26 +0000 (11:48 +0000)]
[project @ 2001-04-05 11:48:26 by simonpj]
Remove tcfail081; it is moving to should_compile

23 years ago[project @ 2001-04-05 11:31:26 by simonpj]
simonpj [Thu, 5 Apr 2001 11:31:26 +0000 (11:31 +0000)]
[project @ 2001-04-05 11:31:26 by simonpj]
--------------------------------
Better grouping for ty/cls decls
--------------------------------

When finding mutually-recursive groups of type and class decls,
we shouldn't include classes mentioned in a deriving clause as
edges. E.g.

class Eq a where ...

data Bool = True | False deriving( Eq )

Eq and Bool are not mutually recursive.

The edges are computed by RnHsSyn.tyClDeclFVs, so we remove the
derivings from there.

There a consequential fix in RnSource.rnSourceDecl

23 years ago[project @ 2001-04-05 11:28:53 by simonpj]
simonpj [Thu, 5 Apr 2001 11:28:53 +0000 (11:28 +0000)]
[project @ 2001-04-05 11:28:53 by simonpj]
Improve error reporting

23 years ago[project @ 2001-04-05 11:28:36 by simonpj]
simonpj [Thu, 5 Apr 2001 11:28:36 +0000 (11:28 +0000)]
[project @ 2001-04-05 11:28:36 by simonpj]
------------------
Better arity stuff
------------------

* CoreToStg now has a local function, predictArity, to
  predict the code-gen arity of a function. Better not to
  use CoreUtils.exprArity, because predictArity is a very
  local thing

* CoreUtils.exprArity is freed to do a better job.  Comments
  below.

exprArity is a cheap-and-cheerful version of exprEtaExpandArity.
It tells how many things the expression can be applied to before doing
any work.  It doesn't look inside cases, lets, etc.  The idea is that
exprEtaExpandArity will do the hard work, leaving something that's easy
for exprArity to grapple with.  In particular, Simplify uses exprArity to
compute the ArityInfo for the Id.

Originally I thought that it was enough just to look for top-level lambdas, but
it isn't.  I've seen this

foo = PrelBase.timesInt

We want foo to get arity 2 even though the eta-expander will leave it
unchanged, in the expectation that it'll be inlined.  But occasionally it
isn't, because foo is blacklisted (used in a rule).

Similarly, see the ok_note check in exprEtaExpandArity.  So
f = __inline_me (\x -> e)
won't be eta-expanded.

And in any case it seems more robust to have exprArity be a bit more intelligent.

23 years ago[project @ 2001-04-05 09:17:15 by simonmar]
simonmar [Thu, 5 Apr 2001 09:17:15 +0000 (09:17 +0000)]
[project @ 2001-04-05 09:17:15 by simonmar]
oops, pass the filename rather than the contents to findDependency for
use in error messages.

23 years ago[project @ 2001-04-05 09:11:15 by simonmar]
simonmar [Thu, 5 Apr 2001 09:11:15 +0000 (09:11 +0000)]
[project @ 2001-04-05 09:11:15 by simonmar]
so much for "Release note for version 5.00 (February 2001)" :)

23 years ago[project @ 2001-04-05 09:10:35 by simonmar]
simonmar [Thu, 5 Apr 2001 09:10:35 +0000 (09:10 +0000)]
[project @ 2001-04-05 09:10:35 by simonmar]
Add notes on parallel list comprehensions from Jeff Lewis.

23 years ago[project @ 2001-04-04 23:53:37 by dsyme]
dsyme [Wed, 4 Apr 2001 23:53:37 +0000 (23:53 +0000)]
[project @ 2001-04-04 23:53:37 by dsyme]
Some work on the ILX backend by Don Syme.  Currently only being compiled by Don as it needs various makefile settings in order for this stuff to be compiled at all.

23 years ago[project @ 2001-04-04 15:50:30 by sewardj]
sewardj [Wed, 4 Apr 2001 15:50:30 +0000 (15:50 +0000)]
[project @ 2001-04-04 15:50:30 by sewardj]
Just build ghc/utils/unlit rather than all of ghc/utils, since the latter
requires a haskell compiler :)

23 years ago[project @ 2001-04-04 14:51:05 by qrczak]
qrczak [Wed, 4 Apr 2001 14:51:05 +0000 (14:51 +0000)]
[project @ 2001-04-04 14:51:05 by qrczak]
Fix escaping of >

23 years ago[project @ 2001-04-04 14:50:01 by sewardj]
sewardj [Wed, 4 Apr 2001 14:50:01 +0000 (14:50 +0000)]
[project @ 2001-04-04 14:50:01 by sewardj]
Make first glafp-utils and ghc/utils before starting on the .hc's.

23 years ago[project @ 2001-04-04 14:31:30 by rrt]
rrt [Wed, 4 Apr 2001 14:31:30 +0000 (14:31 +0000)]
[project @ 2001-04-04 14:31:30 by rrt]
Only pass --target to gmp's configure under mingwin.

23 years ago[project @ 2001-04-04 14:25:49 by sewardj]
sewardj [Wed, 4 Apr 2001 14:25:49 +0000 (14:25 +0000)]
[project @ 2001-04-04 14:25:49 by sewardj]
Don't forget to copy mk/bootstrap.mk into the source distrib tree.

23 years ago[project @ 2001-04-04 13:56:09 by simonpj]
simonpj [Wed, 4 Apr 2001 13:56:09 +0000 (13:56 +0000)]
[project @ 2001-04-04 13:56:09 by simonpj]
Document orphan modules

23 years ago[project @ 2001-04-04 13:39:09 by simonmar]
simonmar [Wed, 4 Apr 2001 13:39:10 +0000 (13:39 +0000)]
[project @ 2001-04-04 13:39:09 by simonmar]
Automatically re-generate and re-run configure as necessary.  This
seems to be a reasonable thing to do, and it should eliminate a large
class of "I ran cvs update and now I get this strange error" type of
reports.

If anyone can remember why this might not be a good idea (I vaguely
remember there being some reason but it escapes me now), please pipe
up.

23 years ago[project @ 2001-04-04 11:21:10 by rrt]
rrt [Wed, 4 Apr 2001 11:21:10 +0000 (11:21 +0000)]
[project @ 2001-04-04 11:21:10 by rrt]
Make FPTOOLS_TOP_ABS in a neater way.

23 years ago[project @ 2001-04-04 06:51:46 by qrczak]
qrczak [Wed, 4 Apr 2001 06:51:46 +0000 (06:51 +0000)]
[project @ 2001-04-04 06:51:46 by qrczak]
Fix filterM type in the export list comment.
Cosmetics.

23 years ago[project @ 2001-04-03 16:50:11 by rrt]
rrt [Tue, 3 Apr 2001 16:50:11 +0000 (16:50 +0000)]
[project @ 2001-04-03 16:50:11 by rrt]
Cleanup.

23 years ago[project @ 2001-04-03 16:43:28 by rrt]
rrt [Tue, 3 Apr 2001 16:43:28 +0000 (16:43 +0000)]
[project @ 2001-04-03 16:43:28 by rrt]
Move Windows options to The Right Place.

23 years ago[project @ 2001-04-03 16:35:12 by sewardj]
sewardj [Tue, 3 Apr 2001 16:35:12 +0000 (16:35 +0000)]
[project @ 2001-04-03 16:35:12 by sewardj]
Disable recursive selector-thunk machinery until we figure out why it
sometimes makes programs loop forever.

23 years ago[project @ 2001-04-03 16:30:43 by rrt]
rrt [Tue, 3 Apr 2001 16:30:43 +0000 (16:30 +0000)]
[project @ 2001-04-03 16:30:43 by rrt]
Move mingwin gcc options to opts.mk (put in CC_OPTS always).

23 years ago[project @ 2001-04-03 16:28:11 by sewardj]
sewardj [Tue, 3 Apr 2001 16:28:11 +0000 (16:28 +0000)]
[project @ 2001-04-03 16:28:11 by sewardj]
A --make test.  Unfortunately not really connected to the build
system yet.

23 years ago[project @ 2001-04-03 15:57:30 by dsyme]
dsyme [Tue, 3 Apr 2001 15:57:30 +0000 (15:57 +0000)]
[project @ 2001-04-03 15:57:30 by dsyme]
Second half of fix to makefile options to get the right options passed across when GHC is used as a C compiler (and also an assembler).  Tested by building the RTS and compiler using the new options.

23 years ago[project @ 2001-04-03 15:51:54 by simonmar]
simonmar [Tue, 3 Apr 2001 15:51:54 +0000 (15:51 +0000)]
[project @ 2001-04-03 15:51:54 by simonmar]
add missing file

23 years ago[project @ 2001-04-03 15:38:51 by simonmar]
simonmar [Tue, 3 Apr 2001 15:38:51 +0000 (15:38 +0000)]
[project @ 2001-04-03 15:38:51 by simonmar]
set SUBDIRS from ProjectsToBuild (as suggested by Marcin).

23 years ago[project @ 2001-04-03 15:36:44 by sewardj]
sewardj [Tue, 3 Apr 2001 15:36:44 +0000 (15:36 +0000)]
[project @ 2001-04-03 15:36:44 by sewardj]
Add headers to generated .hc files in --make mode too.

23 years ago[project @ 2001-04-03 15:30:03 by dsyme]
dsyme [Tue, 3 Apr 2001 15:30:03 +0000 (15:30 +0000)]
[project @ 2001-04-03 15:30:03 by dsyme]
Fix by Reuben Thomas on Don's machine.  CC optsions passed to GHC now have -optc added at the point where we choose to use GHC as the CC compiler, i.e. at "CC=..." in ghc/rts/Makefile and ghc/compiler/Makefile.

23 years ago[project @ 2001-04-03 15:08:35 by sewardj]
sewardj [Tue, 3 Apr 2001 15:08:35 +0000 (15:08 +0000)]
[project @ 2001-04-03 15:08:35 by sewardj]
Fix up make-inplace for binary distribs.

23 years ago[project @ 2001-04-03 15:05:52 by simonmar]
simonmar [Tue, 3 Apr 2001 15:05:53 +0000 (15:05 +0000)]
[project @ 2001-04-03 15:05:52 by simonmar]
Move the RULES for intToInt32# and wordToWord32# to PrelBase, so that
PrelInt and PrelWord are no longer orphan modules.

23 years ago[project @ 2001-04-03 13:18:34 by sewardj]
sewardj [Tue, 3 Apr 2001 13:18:35 +0000 (13:18 +0000)]
[project @ 2001-04-03 13:18:34 by sewardj]
Clean up the post-install-script mechanism a bit, so that the script
for project is <project>/mk/post-install-script.

For project ghc, don't install pkgconf unless BIN_DIST=1.

23 years ago[project @ 2001-04-03 12:58:54 by rrt]
rrt [Tue, 3 Apr 2001 12:58:54 +0000 (12:58 +0000)]
[project @ 2001-04-03 12:58:54 by rrt]
Make ghc-inplace work on Windows by adding FPTOOLS_TOP_ABS_UNIX to be always
the UNIX-style version of the path (for the Cygwin tools), and make CC_OPTS
be set even if there is no C_PROG, so that -mno-cygwin and -mwin32 are
always passed on Windows (and must be filtered out where not needed, e.g. in
lndir.c).

23 years ago[project @ 2001-04-03 12:55:01 by rrt]
rrt [Tue, 3 Apr 2001 12:55:01 +0000 (12:55 +0000)]
[project @ 2001-04-03 12:55:01 by rrt]
Use UNIX-style path in ghc-inplace shell script, even on a Windows system.

23 years ago[project @ 2001-04-03 11:31:38 by sewardj]
sewardj [Tue, 3 Apr 2001 11:31:39 +0000 (11:31 +0000)]
[project @ 2001-04-03 11:31:38 by sewardj]
Make binary-dist installations work, by generating a proper package.conf
file.  This is done by allowing fptools packages to supply an
arbitrary post-install shell script, which the bin-dist 'make install'
runs last of all.  In Project=Ghc's case, the post-install script
runs pkgconf with appropriate params to create the package.conf
file in the right place.

23 years ago[project @ 2001-04-03 10:09:23 by rrt]
rrt [Tue, 3 Apr 2001 10:09:23 +0000 (10:09 +0000)]
[project @ 2001-04-03 10:09:23 by rrt]
Removed spurious bracket that was causing a parse error.

23 years ago[project @ 2001-04-02 21:24:44 by panne]
panne [Mon, 2 Apr 2001 21:24:44 +0000 (21:24 +0000)]
[project @ 2001-04-02 21:24:44 by panne]
Don't use deprecated exception functions

23 years ago[project @ 2001-04-02 21:20:07 by panne]
panne [Mon, 2 Apr 2001 21:20:07 +0000 (21:20 +0000)]
[project @ 2001-04-02 21:20:07 by panne]
fromInt/toInt is dead, long live fromIntegral!

23 years ago[project @ 2001-04-02 21:11:14 by panne]
panne [Mon, 2 Apr 2001 21:11:14 +0000 (21:11 +0000)]
[project @ 2001-04-02 21:11:14 by panne]
Take 4-letter timezones like "CEST" into account. Nevertheless, these
test are bogus IMHO...

23 years ago[project @ 2001-04-02 21:10:04 by panne]
panne [Mon, 2 Apr 2001 21:10:04 +0000 (21:10 +0000)]
[project @ 2001-04-02 21:10:04 by panne]
Made this test a little bit less fragile (my 1GHz Athlon was too fast
for the expected output :-)

23 years ago[project @ 2001-04-02 16:10:32 by rrt]
rrt [Mon, 2 Apr 2001 16:10:33 +0000 (16:10 +0000)]
[project @ 2001-04-02 16:10:32 by rrt]
Remove old mingwin kludgery; using the latest version of mingwin from CVS,
things now work.

23 years ago[project @ 2001-04-02 16:10:05 by rrt]
rrt [Mon, 2 Apr 2001 16:10:05 +0000 (16:10 +0000)]
[project @ 2001-04-02 16:10:05 by rrt]
Add .exe to progName on Windows.

23 years ago[project @ 2001-04-02 14:51:57 by simonmar]
simonmar [Mon, 2 Apr 2001 14:51:57 +0000 (14:51 +0000)]
[project @ 2001-04-02 14:51:57 by simonmar]
- make Printer.c a little more useful when profiling is on,
  by taking advantage of the extra information in the info table.

- enhance findPtr to follow chains of references backwards through the
  heap.  This is a cunning little tool for tracking down space leaks
  (if you're comfortable using gdb as the user-interface to your
  profiler :-)

23 years ago[project @ 2001-04-02 14:47:25 by simonmar]
simonmar [Mon, 2 Apr 2001 14:47:25 +0000 (14:47 +0000)]
[project @ 2001-04-02 14:47:25 by simonmar]
NOINLINE fixTc.  This fixes a space leak, believe it or not.  From the
comment:

-- aargh!  Not inlining fixTc alleviates a space leak problem.
-- Normally fixTc is used with a lazy tuple match: if the optimiser is
-- shown the definition of fixTc, it occasionally transforms the code
-- in such a way that the code generator doesn't spot the selector
-- thunks.  Sigh.

(BTW, NOINLINE pragmas quite often aren't honoured within a single
module, which looks like a bug).

23 years ago[project @ 2001-04-02 14:18:05 by simonmar]
simonmar [Mon, 2 Apr 2001 14:18:05 +0000 (14:18 +0000)]
[project @ 2001-04-02 14:18:05 by simonmar]
Recursively evacuate THUNK_SELETORs down to a bounded depth.

23 years ago[project @ 2001-04-02 07:24:12 by qrczak]
qrczak [Mon, 2 Apr 2001 07:24:12 +0000 (07:24 +0000)]
[project @ 2001-04-02 07:24:12 by qrczak]
Don't cast pointers to ints to perform pointer arithmetic.