ghc.git
16 years ago[project @ 2003-08-22 06:29:42 by panne]
panne [Fri, 22 Aug 2003 06:29:42 +0000 (06:29 +0000)] 
[project @ 2003-08-22 06:29:42 by panne]
OpenBSD's sh doesn't like the parentheses in "foo='(-1)'" so let's
drop them. Thanks to Don for pointing this out.

16 years ago[project @ 2003-08-21 20:08:43 by panne]
panne [Thu, 21 Aug 2003 20:08:43 +0000 (20:08 +0000)] 
[project @ 2003-08-21 20:08:43 by panne]
Added a new test FP_HTML_COMMENT (better name?) and use it in configure.ac.

As an additional bonus, autoreconf doesn't call aclocal anymore
because we have an aclocal.m4 now and it obviously doesn't require
automake stuff. This fixes problems with older (< 1.4-p6) automakes,
which insist on "configure.in".

16 years ago[project @ 2003-08-20 20:51:15 by krc]
krc [Wed, 20 Aug 2003 20:51:15 +0000 (20:51 +0000)] 
[project @ 2003-08-20 20:51:15 by krc]
Oops. Should import Directory instead of System.Directory.

16 years ago[project @ 2003-08-20 18:48:20 by sof]
sof [Wed, 20 Aug 2003 18:48:20 +0000 (18:48 +0000)] 
[project @ 2003-08-20 18:48:20 by sof]
my_prefix_match ~> Util.maybePrefixMatch

16 years ago[project @ 2003-08-20 15:54:43 by panne]
panne [Wed, 20 Aug 2003 15:54:45 +0000 (15:54 +0000)] 
[project @ 2003-08-20 15:54:43 by panne]
Nuked FPTOOLS_O_BINARY test and use FP_CHECK_CONST instead, which is
shorter and better (e.g. cross compilation).

16 years ago[project @ 2003-08-20 15:44:18 by panne]
panne [Wed, 20 Aug 2003 15:44:19 +0000 (15:44 +0000)] 
[project @ 2003-08-20 15:44:18 by panne]
* Cleaned up FP_CHECK_ALIGNMENT test a bit.

* Nuked FPTOOLS_CHECK_CCONST test in favour of a more general one
  which handles cross compilation and languages different from C.

16 years ago[project @ 2003-08-20 15:16:43 by sof]
sof [Wed, 20 Aug 2003 15:16:43 +0000 (15:16 +0000)] 
[project @ 2003-08-20 15:16:43 by sof]
pprInstr{PPC}: insist on xori,ori,andi being fed non-negative immediate values.
               GAS is barfing on neg. values.

[ Don't have access to a PPC box right now to really test the change, so beware.]

16 years ago[project @ 2003-08-20 15:11:06 by simonmar]
simonmar [Wed, 20 Aug 2003 15:11:06 +0000 (15:11 +0000)] 
[project @ 2003-08-20 15:11:06 by simonmar]
After removing temp files, reset the list of temporary files to empty.

16 years ago[project @ 2003-08-20 15:10:22 by simonmar]
simonmar [Wed, 20 Aug 2003 15:10:22 +0000 (15:10 +0000)] 
[project @ 2003-08-20 15:10:22 by simonmar]
Don't drop {-# LINE #-} pragmas into the source, because it confuses
CPP's line counts, making them off-by-one.  GHC understands the
CPP-style line pragmas anyhow.

16 years ago[project @ 2003-08-20 15:07:56 by simonmar]
simonmar [Wed, 20 Aug 2003 15:07:58 +0000 (15:07 +0000)] 
[project @ 2003-08-20 15:07:56 by simonmar]
Use Util.maybePrefixMatch instead of DriverUtil.my_prefix_match, and
remove the latter.

16 years ago[project @ 2003-08-20 15:06:23 by simonmar]
simonmar [Wed, 20 Aug 2003 15:06:23 +0000 (15:06 +0000)] 
[project @ 2003-08-20 15:06:23 by simonmar]
Define maybePrefixMatch, which is like prefixMatch but returns the
rest of the String after the match.

16 years ago[project @ 2003-08-20 12:55:14 by simonmar]
simonmar [Wed, 20 Aug 2003 12:55:14 +0000 (12:55 +0000)] 
[project @ 2003-08-20 12:55:14 by simonmar]
Fix bug in hashSubStringBA, which it appears has been broken for
ever.  However, I suspect so far we've never tickled the bug in GHC.

16 years ago[project @ 2003-08-20 07:07:59 by panne]
panne [Wed, 20 Aug 2003 07:07:59 +0000 (07:07 +0000)] 
[project @ 2003-08-20 07:07:59 by panne]
Nuked the broken FPTOOLS_CC_FLAG test and added a more general
FP_CHECK_FLAG test instead.

16 years ago[project @ 2003-08-19 22:09:09 by krc]
krc [Tue, 19 Aug 2003 22:09:10 +0000 (22:09 +0000)] 
[project @ 2003-08-19 22:09:09 by krc]
External Core programs sometimes contain newtype declarations with no
data constructors. GHC expected every newtype declaration to have at
least one data constructor, so it would previously fail with an "empty list"
error if you tried to compile such a program. These declarations are handled
properly now.

16 years ago[project @ 2003-08-19 22:04:54 by krc]
krc [Tue, 19 Aug 2003 22:04:54 +0000 (22:04 +0000)] 
[project @ 2003-08-19 22:04:54 by krc]
Added correct parsing of external calls. External calls are assumed to be
static C calls, but this assumption should always hold since compiling to
External Core will fail for programs that contain any other kind of
external call.

Added correct parsing of the null-pointer literal (0::GHCziPrim.Addrzh) --
this caused an "unknown integer literal type" error before.

16 years ago[project @ 2003-08-19 22:03:04 by krc]
krc [Tue, 19 Aug 2003 22:03:04 +0000 (22:03 +0000)] 
[project @ 2003-08-19 22:03:04 by krc]
Changed outputForeignStubs to check whether stub files from a previous
compilation still exist (in the case where no new stubs exist). This is
necessary to compile External Core programs that require these stubs.

16 years ago[project @ 2003-08-19 21:59:40 by krc]
krc [Tue, 19 Aug 2003 21:59:40 +0000 (21:59 +0000)] 
[project @ 2003-08-19 21:59:40 by krc]
Two issues:

1. According to the spec for External Core, datatype declarations are required
to have at least one data constructor. Previously, if you tried to generate
External Core for a program containing a datatype declaration with no
constructors, generating the Core file would succeed, but compiling it would
result in a parse error. Changed MkExternalCore to signal an error if such
a declaration is encountered while compiling to External Core.

2. Previously, MachLabel literals were translated into Externals when compiling
to External Core. This is wrong -- such literals are not foreign calls and
can't be handled in the same way (compiling any External Core code generated
from code containing literals resulting from "foreign label" declarations would
result in a strange error message). There doesn't seem to be any way to
correctly represent these labels in External Core, so MkExternalCore now
signals an error if one of these is encountered as well.

16 years ago[project @ 2003-08-19 16:32:23 by simonmar]
simonmar [Tue, 19 Aug 2003 16:32:23 +0000 (16:32 +0000)] 
[project @ 2003-08-19 16:32:23 by simonmar]
setlocale(LC_ALL,""), to enable the locale support in libc.

16 years ago[project @ 2003-08-19 16:31:40 by simonmar]
simonmar [Tue, 19 Aug 2003 16:31:40 +0000 (16:31 +0000)] 
[project @ 2003-08-19 16:31:40 by simonmar]
Check for locale.h and wctype.h

16 years ago[project @ 2003-08-19 15:41:19 by panne]
panne [Tue, 19 Aug 2003 15:41:19 +0000 (15:41 +0000)] 
[project @ 2003-08-19 15:41:19 by panne]
Nuked FPTOOLS_MSGHDR_MSG_ACCRIGHTS and FPTOOLS_MSGHDR_MSG_CONTROL
tests. Use autoconf's own macros for dealing with struct members.

16 years ago[project @ 2003-08-19 14:22:14 by panne]
panne [Tue, 19 Aug 2003 14:22:16 +0000 (14:22 +0000)] 
[project @ 2003-08-19 14:22:14 by panne]
Death to the acconfig.h hack!

16 years ago[project @ 2003-08-19 12:42:39 by panne]
panne [Tue, 19 Aug 2003 12:42:39 +0000 (12:42 +0000)] 
[project @ 2003-08-19 12:42:39 by panne]
Well, we really need autoconf 2.52...

16 years ago[project @ 2003-08-19 12:35:47 by panne]
panne [Tue, 19 Aug 2003 12:35:47 +0000 (12:35 +0000)] 
[project @ 2003-08-19 12:35:47 by panne]
Expand the (a bit faulty) definition of AC_MSG_FAILURE, which is only
available since autoconf 2.57. Thanks to Don for pointing this out.

16 years ago[project @ 2003-08-19 09:39:48 by panne]
panne [Tue, 19 Aug 2003 09:39:48 +0000 (09:39 +0000)] 
[project @ 2003-08-19 09:39:48 by panne]
To ease autoheader's job, add description for USE_QUARTZ_OPENGL.

16 years ago[project @ 2003-08-19 09:25:34 by panne]
panne [Tue, 19 Aug 2003 09:25:35 +0000 (09:25 +0000)] 
[project @ 2003-08-19 09:25:34 by panne]
Renamed FP_ALTZONE to FP_DECL_ALTZONE

16 years ago[project @ 2003-08-19 09:23:08 by panne]
panne [Tue, 19 Aug 2003 09:23:09 +0000 (09:23 +0000)] 
[project @ 2003-08-19 09:23:08 by panne]
Revamped FPTOOLS_PROG_DIFF

16 years ago[project @ 2003-08-18 16:46:37 by panne]
panne [Mon, 18 Aug 2003 16:46:42 +0000 (16:46 +0000)] 
[project @ 2003-08-18 16:46:37 by panne]
Nuked AC_C_PROTOTYPES macro. We don't need the result and newer
autoconfs come with their own definition.

16 years ago[project @ 2003-08-18 16:34:05 by panne]
panne [Mon, 18 Aug 2003 16:34:05 +0000 (16:34 +0000)] 
[project @ 2003-08-18 16:34:05 by panne]
Nuked unused macro FPTOOLS_IN_SCOPE

16 years ago[project @ 2003-08-18 16:29:36 by panne]
panne [Mon, 18 Aug 2003 16:29:36 +0000 (16:29 +0000)] 
[project @ 2003-08-18 16:29:36 by panne]
* Revamped macro for alignment test
* Fixed a few tests for "long long"s

16 years ago[project @ 2003-08-18 15:53:51 by panne]
panne [Mon, 18 Aug 2003 15:53:51 +0000 (15:53 +0000)] 
[project @ 2003-08-18 15:53:51 by panne]
Nuked FPTOOLS_C_LONG_LONG, AC_CHECK_TYPES already does what we want.

16 years ago[project @ 2003-08-18 15:45:07 by panne]
panne [Mon, 18 Aug 2003 15:45:09 +0000 (15:45 +0000)] 
[project @ 2003-08-18 15:45:07 by panne]
Don't build ltx and sgmlverb, they are unused. With this change, the
test for flex/lex can be removed from the configuration process. The
make rules for flex/lex files are still in mk, but they could be moved
to glafp-utils/mk if someone feels the need for this.

Nuked ancient "msub" stuff and a CVS keyword on the way.

16 years ago[project @ 2003-08-18 14:54:40 by panne]
panne [Mon, 18 Aug 2003 14:54:44 +0000 (14:54 +0000)] 
[project @ 2003-08-18 14:54:40 by panne]
Ooops, forgot about a dozen references to configure.in... :-}

16 years ago[project @ 2003-08-18 14:35:26 by panne]
panne [Mon, 18 Aug 2003 14:35:27 +0000 (14:35 +0000)] 
[project @ 2003-08-18 14:35:26 by panne]
AC_PROG_LEX really checks for the existence of "lex" when "flex" is
not found, so there is no need for AC_PROG_LEX_STRICT.

Furthermore, flex/lex is not really needed for building GHC, so we
issue only a warning instead of dying when flex and lex are not found.

16 years ago[project @ 2003-08-18 13:54:04 by panne]
panne [Mon, 18 Aug 2003 13:54:06 +0000 (13:54 +0000)] 
[project @ 2003-08-18 13:54:04 by panne]
Nuked unused FPTOOLS_PROG_YACCY test and support for yacc/bison in
general. If it is ever needed again, it can easily be resurrected.

16 years ago[project @ 2003-08-18 13:15:53 by panne]
panne [Mon, 18 Aug 2003 13:15:56 +0000 (13:15 +0000)] 
[project @ 2003-08-18 13:15:53 by panne]
Nuked unused FPTOOLS_REGEX_IN_LIBC test

16 years ago[project @ 2003-08-18 12:46:20 by panne]
panne [Mon, 18 Aug 2003 12:46:21 +0000 (12:46 +0000)] 
[project @ 2003-08-18 12:46:20 by panne]
Revamped altzone detection. Note that we test only for successful
compilation now, not for successful linking, but this is what most
autoconf macros do.

16 years ago[project @ 2003-08-18 11:29:58 by panne]
panne [Mon, 18 Aug 2003 11:29:58 +0000 (11:29 +0000)] 
[project @ 2003-08-18 11:29:58 by panne]
* Following the autoconf coding style recommendations, use "#" for
  comments, not "dnl".

* Local macros will get the prefix "FP_" instead of "FPTOOLS_". The
  latter is a bit heavyweight and too fptools-centric.

16 years ago[project @ 2003-08-18 11:16:33 by panne]
panne [Mon, 18 Aug 2003 11:16:34 +0000 (11:16 +0000)] 
[project @ 2003-08-18 11:16:33 by panne]
Nuked unused tests FPTOOLS_HAVE_TIMEZONE and FPTOOLS_TYPE_TIMEZONE

16 years ago[project @ 2003-08-18 10:57:54 by panne]
panne [Mon, 18 Aug 2003 10:57:54 +0000 (10:57 +0000)] 
[project @ 2003-08-18 10:57:54 by panne]
CVS keywords considered harmful

16 years ago[project @ 2003-08-18 10:55:31 by panne]
panne [Mon, 18 Aug 2003 10:55:31 +0000 (10:55 +0000)] 
[project @ 2003-08-18 10:55:31 by panne]
We require at least autoconf 2.50 (not 2.52 and/or 2.53?!)

16 years ago[project @ 2003-08-18 10:47:33 by panne]
panne [Mon, 18 Aug 2003 10:47:34 +0000 (10:47 +0000)] 
[project @ 2003-08-18 10:47:33 by panne]
The autoconf docs talk about "configure.ac" instead of the misnomer
"configure.in" for more than two years now, so let's do the same.

16 years ago[project @ 2003-08-18 10:33:39 by panne]
panne [Mon, 18 Aug 2003 10:33:41 +0000 (10:33 +0000)] 
[project @ 2003-08-18 10:33:39 by panne]
Updated to match latest versions from autoconf repository

16 years ago[project @ 2003-08-18 09:27:54 by dons]
dons [Mon, 18 Aug 2003 09:27:54 +0000 (09:27 +0000)] 
[project @ 2003-08-18 09:27:54 by dons]
x86 OpenBSD moves into ELF land
(and sparc, alpha, sparc64 were already there)

Also, it needs sligtly different header files.

And dlopen can only be called with the DL_LAZY argument on this
system.

16 years ago[project @ 2003-08-18 09:24:50 by dons]
dons [Mon, 18 Aug 2003 09:24:50 +0000 (09:24 +0000)] 
[project @ 2003-08-18 09:24:50 by dons]
OpenBSD has now moved to ELF on x86 architectures.
So mangle it.

16 years ago[project @ 2003-08-18 09:23:09 by dons]
dons [Mon, 18 Aug 2003 09:23:09 +0000 (09:23 +0000)] 
[project @ 2003-08-18 09:23:09 by dons]
OpenBSD x86 has finally moved to ELF.
So it no longer needs a leading underscore on symbols.
However, like FreeBSD, it tries to be friendly, and accept it
anyway. So we have to hard code it.

16 years ago[project @ 2003-08-18 00:14:10 by dons]
dons [Mon, 18 Aug 2003 00:14:10 +0000 (00:14 +0000)] 
[project @ 2003-08-18 00:14:10 by dons]
sparc-*-openbsd mangler

16 years ago[project @ 2003-08-18 00:01:07 by dons]
dons [Mon, 18 Aug 2003 00:01:07 +0000 (00:01 +0000)] 
[project @ 2003-08-18 00:01:07 by dons]
Add canonical form for sparc-*-openbsd*

16 years ago[project @ 2003-08-17 01:36:54 by sof]
sof [Sun, 17 Aug 2003 01:36:54 +0000 (01:36 +0000)] 
[project @ 2003-08-17 01:36:54 by sof]
expandEnvVars.expandString: don't drop prefix when locating variable
reference.

Merge to STABLE.

16 years ago[project @ 2003-08-16 11:19:07 by igloo]
igloo [Sat, 16 Aug 2003 11:19:08 +0000 (11:19 +0000)] 
[project @ 2003-08-16 11:19:07 by igloo]
Sparc Linux configure and mangler

16 years ago[project @ 2003-08-15 15:53:00 by igloo]
igloo [Fri, 15 Aug 2003 15:53:01 +0000 (15:53 +0000)] 
[project @ 2003-08-15 15:53:00 by igloo]
Add support for splicing in foreign exports.

16 years ago[project @ 2003-08-15 12:43:57 by simonmar]
simonmar [Fri, 15 Aug 2003 12:43:57 +0000 (12:43 +0000)] 
[project @ 2003-08-15 12:43:57 by simonmar]
Fix a bug in forkProcess: the current thread should not be on the run
queue while it is running.

Merged from STABLE (1.169.2.1).

16 years ago[project @ 2003-08-15 11:31:02 by simonmar]
simonmar [Fri, 15 Aug 2003 11:31:02 +0000 (11:31 +0000)] 
[project @ 2003-08-15 11:31:02 by simonmar]
Fix a cause of non-linearity when compiling code with large amounts of
static data, such as long lists of integers.

The problem was that CorePrep was calling allLazy on the list of
floated bindings at each level of a nested let, and allLazy runs over
all the bindings floated so far, hence O(n^2) performance.  The fix is
to cache the result of allLazy along with the list of floated
bindings.

16 years ago[project @ 2003-08-14 15:43:22 by simonmar]
simonmar [Thu, 14 Aug 2003 15:43:22 +0000 (15:43 +0000)] 
[project @ 2003-08-14 15:43:22 by simonmar]
Don't require Alex for building GHC (yet).

16 years ago[project @ 2003-08-14 15:36:13 by simonmar]
simonmar [Thu, 14 Aug 2003 15:36:13 +0000 (15:36 +0000)] 
[project @ 2003-08-14 15:36:13 by simonmar]
Reset g0s0->n_to_blocks when using the two-space collector.  Fixes
ever-increasing live data count with +RTS -Sstderr -G1.

16 years ago[project @ 2003-08-14 10:43:10 by simonmar]
simonmar [Thu, 14 Aug 2003 10:43:10 +0000 (10:43 +0000)] 
[project @ 2003-08-14 10:43:10 by simonmar]
Fix terrible bug in isNonPtrRep, which was ignoring the possibily of
the 64-bit reps.  The result was that functions with 2 64-bit args
were getting the wrong argument descriptor, leading to crashes.

Thanks to Peter Thiemann for a great bug report.

16 years ago[project @ 2003-08-13 13:35:07 by simonmar]
simonmar [Wed, 13 Aug 2003 13:35:07 +0000 (13:35 +0000)] 
[project @ 2003-08-13 13:35:07 by simonmar]
Fix the Alex-version-detecting machinery

16 years ago[project @ 2003-08-13 12:51:01 by simonmar]
simonmar [Wed, 13 Aug 2003 12:51:01 +0000 (12:51 +0000)] 
[project @ 2003-08-13 12:51:01 by simonmar]
Add alex to AllProjects

16 years ago[project @ 2003-08-13 12:43:13 by simonmar]
simonmar [Wed, 13 Aug 2003 12:43:13 +0000 (12:43 +0000)] 
[project @ 2003-08-13 12:43:13 by simonmar]
Add DERIVED_ALEX_SRCS to maintainer-clean

16 years ago[project @ 2003-08-13 12:34:45 by simonmar]
simonmar [Wed, 13 Aug 2003 12:34:46 +0000 (12:34 +0000)] 
[project @ 2003-08-13 12:34:45 by simonmar]
Add Alex support.

16 years ago[project @ 2003-08-13 12:34:04 by simonmar]
simonmar [Wed, 13 Aug 2003 12:34:05 +0000 (12:34 +0000)] 
[project @ 2003-08-13 12:34:04 by simonmar]
Detect Alex, and add --use-src-tree-alex option to configure.

16 years ago[project @ 2003-08-11 08:53:39 by simonmar]
simonmar [Mon, 11 Aug 2003 08:53:39 +0000 (08:53 +0000)] 
[project @ 2003-08-11 08:53:39 by simonmar]
Revert part of previous commit that sneaked in by accident.

16 years ago[project @ 2003-08-08 16:46:51 by simonmar]
simonmar [Fri, 8 Aug 2003 16:46:51 +0000 (16:46 +0000)] 
[project @ 2003-08-08 16:46:51 by simonmar]
Fix string gap.

16 years ago[project @ 2003-08-08 13:22:54 by simonmar]
simonmar [Fri, 8 Aug 2003 13:22:56 +0000 (13:22 +0000)] 
[project @ 2003-08-08 13:22:54 by simonmar]
Moved the contents of fptools/happy/alex -r simonm-hackery-branch to
here.  This will become Alex version 2.0 when I've hacked around with
the build and documentation a bit.

16 years ago[project @ 2003-08-08 08:08:47 by simonmar]
simonmar [Fri, 8 Aug 2003 08:08:47 +0000 (08:08 +0000)] 
[project @ 2003-08-08 08:08:47 by simonmar]
- Add -main-is to the flag table
- Sort entries in the linker section alphabetically

16 years ago[project @ 2003-08-06 15:26:02 by simonpj]
simonpj [Wed, 6 Aug 2003 15:26:02 +0000 (15:26 +0000)] 
[project @ 2003-08-06 15:26:02 by simonpj]
Add a note about readline on Win32

16 years ago[project @ 2003-08-05 15:52:01 by ross]
ross [Tue, 5 Aug 2003 15:52:01 +0000 (15:52 +0000)] 
[project @ 2003-08-05 15:52:01 by ross]
punctuation

16 years ago[project @ 2003-08-05 14:01:34 by simonpj]
simonpj [Tue, 5 Aug 2003 14:01:34 +0000 (14:01 +0000)] 
[project @ 2003-08-05 14:01:34 by simonpj]
Add another comment

16 years ago[project @ 2003-08-05 12:41:44 by panne]
panne [Tue, 5 Aug 2003 12:41:44 +0000 (12:41 +0000)] 
[project @ 2003-08-05 12:41:44 by panne]
* Added QuickCheck package to docs
* Mention OpenGL/GLUT package for Hugs, too

16 years ago[project @ 2003-08-05 12:00:03 by simonmar]
simonmar [Tue, 5 Aug 2003 12:00:03 +0000 (12:00 +0000)] 
[project @ 2003-08-05 12:00:03 by simonmar]
Remove special mingw32 case for configuring GMP, since it appears to
be wrong.

16 years ago[project @ 2003-08-05 07:25:56 by panne]
panne [Tue, 5 Aug 2003 07:25:56 +0000 (07:25 +0000)] 
[project @ 2003-08-05 07:25:56 by panne]
* Unbreak bootstrapping with GHC >= 6.1
* Nuked CVS keywords, you'll hate them when merging...

16 years ago[project @ 2003-08-04 14:38:59 by simonmar]
simonmar [Mon, 4 Aug 2003 14:38:59 +0000 (14:38 +0000)] 
[project @ 2003-08-04 14:38:59 by simonmar]
A couple of small updates that were sitting in my tree..

16 years ago[project @ 2003-08-04 10:06:31 by ross]
ross [Mon, 4 Aug 2003 10:06:31 +0000 (10:06 +0000)] 
[project @ 2003-08-04 10:06:31 by ross]
minor rearrangement

16 years ago[project @ 2003-08-01 15:38:41 by panne]
panne [Fri, 1 Aug 2003 15:38:41 +0000 (15:38 +0000)] 
[project @ 2003-08-01 15:38:41 by panne]
Compile with -Wall

16 years ago[project @ 2003-08-01 15:10:34 by panne]
panne [Fri, 1 Aug 2003 15:10:34 +0000 (15:10 +0000)] 
[project @ 2003-08-01 15:10:34 by panne]
Even more warning police:
* Help GCC a bit with "unitialized" variables.
* Added #include for islower
* Nuked variable `syncs', which always contained the empty string.
  GCC complained about an empty format string, BTW.

16 years ago[project @ 2003-08-01 14:58:47 by panne]
panne [Fri, 1 Aug 2003 14:58:48 +0000 (14:58 +0000)] 
[project @ 2003-08-01 14:58:47 by panne]
Warning police: Nuked unused variables.

16 years ago[project @ 2003-08-01 14:50:48 by panne]
panne [Fri, 1 Aug 2003 14:50:50 +0000 (14:50 +0000)] 
[project @ 2003-08-01 14:50:48 by panne]
Warning police: GCC complains about built-in function `log' being used
as a non-function, so let's rename it to logFile.

16 years ago[project @ 2003-07-31 23:16:13 by sof]
sof [Thu, 31 Jul 2003 23:16:13 +0000 (23:16 +0000)] 
[project @ 2003-07-31 23:16:13 by sof]
disable --enable-win32-dlls, feature not currently on offer

16 years ago[project @ 2003-07-31 17:47:18 by ross]
ross [Thu, 31 Jul 2003 17:47:18 +0000 (17:47 +0000)] 
[project @ 2003-07-31 17:47:18 by ross]
add link to parsec package

16 years ago[project @ 2003-07-31 10:07:02 by simonmar]
simonmar [Thu, 31 Jul 2003 10:07:02 +0000 (10:07 +0000)] 
[project @ 2003-07-31 10:07:02 by simonmar]
Add comment about the LDV code in UPD_DB_UPDATABLE when
EAGER_BLACKHOLING is no - it looks like it can't possibly work to me.

16 years ago[project @ 2003-07-30 17:09:12 by simonmar]
simonmar [Wed, 30 Jul 2003 17:09:12 +0000 (17:09 +0000)] 
[project @ 2003-07-30 17:09:12 by simonmar]
Fix problem with MARK.

Patch submitted by Nicholas Nethercote <njn25@cam.ac.uk>.

16 years ago[project @ 2003-07-30 10:38:42 by simonmar]
simonmar [Wed, 30 Jul 2003 10:38:42 +0000 (10:38 +0000)] 
[project @ 2003-07-30 10:38:42 by simonmar]
TEXT_BEFORE_HEAP is not used any more.

16 years ago[project @ 2003-07-30 10:01:46 by simonmar]
simonmar [Wed, 30 Jul 2003 10:01:46 +0000 (10:01 +0000)] 
[project @ 2003-07-30 10:01:46 by simonmar]
Throw away an SCC on a single variable.  This seems to be justified:
there can be no work to attribute to the current cost centre when
evaluating a lone variable, other than the act of entering the
closure, and possibly returning immediately if it is a variable.

This also fixes a bug in profiling, which showed up as incorrect
transformations made by the simplifier resulting in extra strictness.
The simplifier assumes (in Simplify.simplLazyBind) that (let x = e in
x) will have been turned into x, but this isn't true if there's an SCC
around the x.

16 years ago[project @ 2003-07-30 09:21:12 by wolfgang]
wolfgang [Wed, 30 Jul 2003 09:21:12 +0000 (09:21 +0000)] 
[project @ 2003-07-30 09:21:12 by wolfgang]
OpenGL support for Mac OS X:

In FPTOOLS_CHECK_HTYPE, be ready to include Apple's non-standard
<OpenGL/gl.h>
instead of the usual <GL/gl.h>.

16 years ago[project @ 2003-07-30 09:20:41 by wolfgang]
wolfgang [Wed, 30 Jul 2003 09:20:42 +0000 (09:20 +0000)] 
[project @ 2003-07-30 09:20:41 by wolfgang]
OpenGL support for Mac OS X:

"./configure --enable-hopengl" on Mac OS X now uses the OpenGL libs that come
with Apple's "Quartz" display system.
"./configure --enable-hopengl=x11" looks for an X11 implementation of OpenGL
(which is an optional install on Mac OS X).

Added a corresponding "USE_QUARTZ_OPENGL" flag to config.h

16 years ago[project @ 2003-07-29 19:08:40 by panne]
panne [Tue, 29 Jul 2003 19:08:40 +0000 (19:08 +0000)] 
[project @ 2003-07-29 19:08:40 by panne]
Revert to previous commit, i.e. ProjectVersion is now a plain "6.1"
again, not "6.1.20030727". It looked like an accidental commit and
broke my build scripts.

16 years ago[project @ 2003-07-29 16:31:39 by igloo]
igloo [Tue, 29 Jul 2003 16:31:39 +0000 (16:31 +0000)] 
[project @ 2003-07-29 16:31:39 by igloo]
gcc 3.3's cpp fix on sparc.

Could be merged to stable if it's not too late.

16 years ago[project @ 2003-07-29 10:15:31 by simonpj]
simonpj [Tue, 29 Jul 2003 10:15:31 +0000 (10:15 +0000)] 
[project @ 2003-07-29 10:15:31 by simonpj]
Print HsExprs a bit deeper before ... ellipsis

16 years ago[project @ 2003-07-29 10:14:57 by simonpj]
simonpj [Tue, 29 Jul 2003 10:15:10 +0000 (10:15 +0000)] 
[project @ 2003-07-29 10:14:57 by simonpj]
Comments only

16 years ago[project @ 2003-07-28 16:16:07 by moran]
moran [Mon, 28 Jul 2003 16:16:07 +0000 (16:16 +0000)] 
[project @ 2003-07-28 16:16:07 by moran]
Fix multi-slurp protection (#define was missing).  (No I don't go around
looking at header files for minor problems like this; I was actually trying to
remind myself how to do it, and this was the first place I looked :-)

16 years ago[project @ 2003-07-28 16:05:30 by simonmar]
simonmar [Mon, 28 Jul 2003 16:05:39 +0000 (16:05 +0000)] 
[project @ 2003-07-28 16:05:30 by simonmar]
Disable update-in-place.  In its current form, it has a serious bug:
if the thunk being updated happens to have turned into a BLACKHOLE_BQ,
then the mutable list will be corrupted by the update.

Disabling update-in-place has some performance implications: many
programs are not affected, but one program in nofib (nucleic2) goes
about 20% slower.  However, I can get it to go 300% faster by adding a
few strictness annotations and compiling with -funbox-strict-fields.

16 years ago[project @ 2003-07-28 15:59:09 by simonmar]
simonmar [Mon, 28 Jul 2003 15:59:09 +0000 (15:59 +0000)] 
[project @ 2003-07-28 15:59:09 by simonmar]
Comments only: the TICK_UPD_*_IN_PLACE aren't relevant any more.

16 years ago[project @ 2003-07-28 15:57:40 by simonmar]
simonmar [Mon, 28 Jul 2003 15:57:40 +0000 (15:57 +0000)] 
[project @ 2003-07-28 15:57:40 by simonmar]
TICK_UPD_CON_IN_NEW() in the update code is wrong now, because it
could be a PAP, not just a CON.

16 years ago[project @ 2003-07-28 13:31:37 by simonpj]
simonpj [Mon, 28 Jul 2003 13:31:37 +0000 (13:31 +0000)] 
[project @ 2003-07-28 13:31:37 by simonpj]
Reorganise the type-system-extension part of GlaExts docs

16 years ago[project @ 2003-07-28 12:04:27 by simonpj]
simonpj [Mon, 28 Jul 2003 12:04:27 +0000 (12:04 +0000)] 
[project @ 2003-07-28 12:04:27 by simonpj]
--------------------------
  Fix an obscure but long-standing bug in Type.applyTys
   --------------------------

The interesting case, which previously killed GHC 6.0, is this
applyTys (forall a.a) [forall b.b, Int]
This really can happen, via dressing up polymorphic types with newtype
clothing.  Here's an example:
newtype R = R (forall a. a->a)
foo = case undefined :: R of

Test simplCore/should_compile/simpl0009 uses this as a test case.

16 years ago[project @ 2003-07-28 10:22:58 by simonpj]
simonpj [Mon, 28 Jul 2003 10:22:58 +0000 (10:22 +0000)] 
[project @ 2003-07-28 10:22:58 by simonpj]
Comments only

16 years ago[project @ 2003-07-28 09:01:58 by simonmar]
simonmar [Mon, 28 Jul 2003 09:01:58 +0000 (09:01 +0000)] 
[project @ 2003-07-28 09:01:58 by simonmar]
Fix previous patch (backslashes required in multi-line macro defns).

16 years ago[project @ 2003-07-25 14:39:17 by simonmar]
simonmar [Fri, 25 Jul 2003 14:39:17 +0000 (14:39 +0000)] 
[project @ 2003-07-25 14:39:17 by simonmar]
Further updates to the hc-file-bundle target: ghc/compiler/utils is
required, and we should also keep ghc/rts/AutoApply.hc.

16 years ago[project @ 2003-07-25 10:22:25 by simonmar]
simonmar [Fri, 25 Jul 2003 10:22:25 +0000 (10:22 +0000)] 
[project @ 2003-07-25 10:22:25 by simonmar]
Some more revisions; it nearly works on my unregisterised setup now.

16 years ago[project @ 2003-07-25 08:57:34 by simonmar]
simonmar [Fri, 25 Jul 2003 08:57:34 +0000 (08:57 +0000)] 
[project @ 2003-07-25 08:57:34 by simonmar]
Remove multi-line strings in macros (CPP is stricter in GCC 3.3).

From: Oliver Braun.

16 years ago[project @ 2003-07-24 16:31:42 by simonmar]
simonmar [Thu, 24 Jul 2003 16:31:42 +0000 (16:31 +0000)] 
[project @ 2003-07-24 16:31:42 by simonmar]
Build genapply when BootingFromHc too - it'll help when compiling the
RTS in the second stage (after turning off BootingFromHc).