ghc-hetmet.git
21 years ago[project @ 2002-07-16 14:56:08 by simonmar]
simonmar [Tue, 16 Jul 2002 14:56:11 +0000 (14:56 +0000)]
[project @ 2002-07-16 14:56:08 by simonmar]
Implement a primitive failsafe mechanism for protecting against
linking inconsistent object files.  The idea is that if object files
which were compiled in the wrong order (non-dependency order) or
compiled in different ways (eg. profiled vs. non-profiled) are linked
together, a link error will result.

This is achieved by adding the module version and the way to the
module init label.  For example, previously the init label for a
module Foo was named

__stginit_Foo

now it is named

__stginit_Foo_<version>_<way>

where <version> is the module version of Foo (same as the version in
the interface file), and <way> is the current way (or empty).

We also have to have a way to refer to the old plain init label, for
using as the argument to shutdownHaskell() in a program using foreign
exports.  So the old label now points to a jump instruction which
transfers control to the new init code.

21 years ago[project @ 2002-07-16 14:49:38 by simonmar]
simonmar [Tue, 16 Jul 2002 14:49:38 +0000 (14:49 +0000)]
[project @ 2002-07-16 14:49:38 by simonmar]
comment fix

21 years ago[project @ 2002-07-16 12:05:37 by simonmar]
simonmar [Tue, 16 Jul 2002 12:05:37 +0000 (12:05 +0000)]
[project @ 2002-07-16 12:05:37 by simonmar]
un-rot one transformation on x86: we normally transform

movl $_blah, %eax
jmp  *%eax

into simply

jmp _blah

but the pattern had rotted w.r.t. gcc so this was no longer applying.
Should reduce code size measurably.

MERGE TO STABLE

21 years ago[project @ 2002-07-16 10:58:16 by simonmar]
simonmar [Tue, 16 Jul 2002 10:58:16 +0000 (10:58 +0000)]
[project @ 2002-07-16 10:58:16 by simonmar]
Remove DLL_IMPORT from the two closure declarations in this file,
because otherwise the file is not standalone.

Doing the right thing doesn't seem easy, because we have to get the
right value of DONT_WANT_WIN32_DLLS from somewhere.  Anyway if/when
DLL support is revived we'll have to revisit this.

MERGE TO STABLE

21 years ago[project @ 2002-07-16 10:51:12 by simonmar]
simonmar [Tue, 16 Jul 2002 10:51:12 +0000 (10:51 +0000)]
[project @ 2002-07-16 10:51:12 by simonmar]
The restore instruction on Sparc apparently has arguments under GCC
3.1, which caused one of the patterns to break in the mangler.

I can now run simple programs compiled with GHC on Sparc using GCC
3.1, so hopefully this fixes it.

MERGE TO STABLE (and the previous one).

21 years ago[project @ 2002-07-16 10:17:37 by simonmar]
simonmar [Tue, 16 Jul 2002 10:17:37 +0000 (10:17 +0000)]
[project @ 2002-07-16 10:17:37 by simonmar]
GCC 3.1 on Sparc now uses '.long'.

21 years ago[project @ 2002-07-16 07:04:49 by stolz]
stolz [Tue, 16 Jul 2002 07:04:50 +0000 (07:04 +0000)]
[project @ 2002-07-16 07:04:49 by stolz]
The <dlfcn.h> constants RTLD_NOW, RTLD_GLOBAL, RTLD_LOCAL
are not defined in the OpenBSD implementation of the dl library.

dons@cse.unsw.edu.au (Donald Bruce Stewart)

21 years ago[project @ 2002-07-16 06:42:04 by sof]
sof [Tue, 16 Jul 2002 06:42:04 +0000 (06:42 +0000)]
[project @ 2002-07-16 06:42:04 by sof]
Treat ';' and '{' as word separators, so that from
input like like "{import Foo" and ";;;;import Foo",
the 'Foo' imports are identified.

21 years ago[project @ 2002-07-15 16:43:34 by panne]
panne [Mon, 15 Jul 2002 16:43:34 +0000 (16:43 +0000)]
[project @ 2002-07-15 16:43:34 by panne]
* Solaris' quite spartanic /bin/sh doesn't know about the '!' keyword,
  so let's simply reverse the operator in the test instead.

* The FPTOOLS_SEARCH_LIBS macro expects a function without any
  arguments, but glXChooseVisual *does* have arguments. glXWaitX is a
  better test for GLX libs.

MERGE TO STABLE

21 years ago[project @ 2002-07-15 14:49:17 by simonmar]
simonmar [Mon, 15 Jul 2002 14:49:17 +0000 (14:49 +0000)]
[project @ 2002-07-15 14:49:17 by simonmar]
Unbreak the registering of foreign exports as stable pointers, which
was brokwn in rev. 1.64.  This will eventually cause any program that
uses the same foreign export repeatedly to seg fault.  Dunno why we
don't have a test for this.

*sigh* the first bad bug in 5.04

MERGE TO STABLE

21 years ago[project @ 2002-07-15 08:41:46 by simonmar]
simonmar [Mon, 15 Jul 2002 08:41:46 +0000 (08:41 +0000)]
[project @ 2002-07-15 08:41:46 by simonmar]
Don't declare the StgClosures 'const' (they aren't declared 'const' in
generated code).

21 years ago[project @ 2002-07-15 08:41:02 by simonmar]
simonmar [Mon, 15 Jul 2002 08:41:02 +0000 (08:41 +0000)]
[project @ 2002-07-15 08:41:02 by simonmar]
Static closures are not declared 'const' in generated code, so don't
declare them 'const' in here.

Spotted-by: GCC 3.1

21 years ago[project @ 2002-07-15 07:24:58 by chak]
chak [Mon, 15 Jul 2002 07:24:58 +0000 (07:24 +0000)]
[project @ 2002-07-15 07:24:58 by chak]
Let `configure' know about the "m68k-unknown-openbsd" architecture.
Patch contributed by Donald Stewart <dons@cse.unsw.edu.au>.

21 years ago[project @ 2002-07-12 17:20:57 by sof]
sof [Fri, 12 Jul 2002 17:20:57 +0000 (17:20 +0000)]
[project @ 2002-07-12 17:20:57 by sof]
mkdll: in installed mode, invoke the 'dllwrap' that's distributed with GHC,
and feed it the right options.

Merge to STABLE.

21 years ago[project @ 2002-07-12 13:42:22 by simonpj]
simonpj [Fri, 12 Jul 2002 13:42:22 +0000 (13:42 +0000)]
[project @ 2002-07-12 13:42:22 by simonpj]
comments

21 years ago[project @ 2002-07-12 04:45:27 by sof]
sof [Fri, 12 Jul 2002 04:45:27 +0000 (04:45 +0000)]
[project @ 2002-07-12 04:45:27 by sof]
suggest compiling up the DLL wrapper .c file using 'ghc' rather than 'gcc'

21 years ago[project @ 2002-07-12 02:34:08 by chak]
chak [Fri, 12 Jul 2002 02:34:08 +0000 (02:34 +0000)]
[project @ 2002-07-12 02:34:08 by chak]
Updates for 5.04

21 years ago[project @ 2002-07-12 01:28:10 by chak]
chak [Fri, 12 Jul 2002 01:28:10 +0000 (01:28 +0000)]
[project @ 2002-07-12 01:28:10 by chak]
Added DocBook catalog location for OpenBSD as suggested by
Donald Stewart <dons@cse.unsw.edu.au>

21 years ago[project @ 2002-07-11 17:19:22 by sof]
sof [Thu, 11 Jul 2002 17:19:22 +0000 (17:19 +0000)]
[project @ 2002-07-11 17:19:22 by sof]
Use of InstallShield is a thing of the past

21 years ago[project @ 2002-07-11 11:24:31 by panne]
panne [Thu, 11 Jul 2002 11:24:31 +0000 (11:24 +0000)]
[project @ 2002-07-11 11:24:31 by panne]
Automagically insert links to the documentation of the OpenGL/GLUT
packages iff --enable-hopengl is used

21 years ago[project @ 2002-07-11 10:44:37 by panne]
panne [Thu, 11 Jul 2002 10:44:37 +0000 (10:44 +0000)]
[project @ 2002-07-11 10:44:37 by panne]
Corrected and normalized HTML

21 years ago[project @ 2002-07-11 09:44:41 by simonmar]
simonmar [Thu, 11 Jul 2002 09:44:41 +0000 (09:44 +0000)]
[project @ 2002-07-11 09:44:41 by simonmar]
Fix a couple of links in the announcement

21 years ago[project @ 2002-07-11 06:52:23 by ken]
ken [Thu, 11 Jul 2002 06:52:23 +0000 (06:52 +0000)]
[project @ 2002-07-11 06:52:23 by ken]
Perhaps defining _REENTRANT when gcc-compiling for the alpha would
help with RTS threading problems

21 years ago[project @ 2002-07-10 16:33:09 by sof]
sof [Wed, 10 Jul 2002 16:33:09 +0000 (16:33 +0000)]
[project @ 2002-07-10 16:33:09 by sof]
printDump,printErrs,printSDoc: flush stdout and stderr

21 years ago[project @ 2002-07-10 13:07:42 by simonpj]
simonpj [Wed, 10 Jul 2002 13:07:42 +0000 (13:07 +0000)]
[project @ 2002-07-10 13:07:42 by simonpj]
Fix trivial bug in WorkWrap that killed all INLINE pragmas!  Merge to STABLE

21 years ago[project @ 2002-07-10 13:06:35 by simonpj]
simonpj [Wed, 10 Jul 2002 13:06:35 +0000 (13:06 +0000)]
[project @ 2002-07-10 13:06:35 by simonpj]
Improve error message; merge to STABLE

21 years ago[project @ 2002-07-10 09:28:54 by simonmar]
simonmar [Wed, 10 Jul 2002 09:28:56 +0000 (09:28 +0000)]
[project @ 2002-07-10 09:28:54 by simonmar]
Fix a GC bug.  In a "large block", only the block descriptor for the
head of the block has the fields step, gen_no and flags set.  So
whenever we want one of these fields in the descriptor for a random
object anywhere in the large block, we have to check whether it is in
the head block, and if not follow the link to the head block
descriptor.

evacuate() was doing this correctly, but isAlive() wasn't (and
goodness knows what other places are broken in this way - I identified
several other possible cases of the same bug).

So to try to make things more robust, when we allocate a large block
we now initialise the step, gen_no, and flags fields in the descriptor
for *every* sub-block, not just the first one.  Now, as long as you
only want one of these fields from the descriptor, there's no need to
try to find the block head.  evacuate() gets minutely faster, and
hopefully multiple obscure bugs are fixed by this.

21 years ago[project @ 2002-07-10 08:39:00 by simonmar]
simonmar [Wed, 10 Jul 2002 08:39:00 +0000 (08:39 +0000)]
[project @ 2002-07-10 08:39:00 by simonmar]
change #ifdef USING_LIBBFD to #if defined(DEBUG) && defined(HAVE_LIBBFD).

21 years ago[project @ 2002-07-10 08:34:19 by simonmar]
simonmar [Wed, 10 Jul 2002 08:34:19 +0000 (08:34 +0000)]
[project @ 2002-07-10 08:34:19 by simonmar]
Fix a bug in the handling of libbfd - we should be checking for DEBUG
in $(GhcRtsHcOpts) not $(GhcHcOpts), and we can get HAVE_LIBBFD from
config.h so there's no need to unconditionally define USING_LIBBFD.

21 years ago[project @ 2002-07-09 20:44:24 by sof]
sof [Tue, 9 Jul 2002 20:44:24 +0000 (20:44 +0000)]
[project @ 2002-07-09 20:44:24 by sof]
awaitEvent: better handling of EBADFs, i.e., don't
unconditionally barf() and exit if select() reports
an EBADF. See source code comments for details, but
in short, we attempt to unblock all threads to handle
the error condition before bailing out.

If only select() would indicate which file descriptor
that was the bad one.

(There's no good reason why select() errors other
than EBADF could also be handled this way, but let's
focus on it for now..)

21 years ago[project @ 2002-07-09 18:02:48 by sof]
sof [Tue, 9 Jul 2002 18:02:48 +0000 (18:02 +0000)]
[project @ 2002-07-09 18:02:48 by sof]
gcc versionitis wrt gcc -E -MM .. usage:

2.95.<wurble> is silent on stdout if -MM is used, 2.96 ain't
(even when -o is used..hmm.) 2.95 (on mingw32) even fails if
you try to redirect the output.

So, be more accepting in first trying to run CPP with a redirect,
and if that fails, try without.

21 years ago[project @ 2002-07-09 15:57:09 by sof]
sof [Tue, 9 Jul 2002 15:57:09 +0000 (15:57 +0000)]
[project @ 2002-07-09 15:57:09 by sof]
merge rev. 1.1.6.1

21 years ago[project @ 2002-07-09 08:19:14 by simonpj]
simonpj [Tue, 9 Jul 2002 08:19:14 +0000 (08:19 +0000)]
[project @ 2002-07-09 08:19:14 by simonpj]
Improve error message

*** MERGE TO STABLE BRANCH ***

21 years ago[project @ 2002-07-09 06:16:14 by panne]
panne [Tue, 9 Jul 2002 06:16:15 +0000 (06:16 +0000)]
[project @ 2002-07-09 06:16:14 by panne]
Fixed my email address

21 years ago[project @ 2002-07-08 14:38:26 by simonpj]
simonpj [Mon, 8 Jul 2002 14:38:26 +0000 (14:38 +0000)]
[project @ 2002-07-08 14:38:26 by simonpj]
Remove duplicate line in Linker.c

*** MERGE TO STABLE BRANCH ***

21 years ago[project @ 2002-07-08 13:39:10 by simonmar]
simonmar [Mon, 8 Jul 2002 13:39:10 +0000 (13:39 +0000)]
[project @ 2002-07-08 13:39:10 by simonmar]
HEAD is now 5.05

21 years ago[project @ 2002-07-06 10:14:31 by chak]
chak [Sat, 6 Jul 2002 10:14:31 +0000 (10:14 +0000)]
[project @ 2002-07-06 10:14:31 by chak]
Reverted to check for `null srcs' (instead of `null real_srcs') to determine
whether to throw "no input files".  Otherwise, something like

  ghc -o foo ObjectGeneratedByGHC.o

doesn't work anymore.

21 years ago[project @ 2002-07-05 20:30:38 by sof]
sof [Fri, 5 Jul 2002 20:30:38 +0000 (20:30 +0000)]
[project @ 2002-07-05 20:30:38 by sof]
isSourceFile: widened to also return True for .hc files

21 years ago[project @ 2002-07-05 19:05:41 by sof]
sof [Fri, 5 Jul 2002 19:05:41 +0000 (19:05 +0000)]
[project @ 2002-07-05 19:05:41 by sof]
5.04

21 years ago[project @ 2002-07-05 16:15:14 by sof]
sof [Fri, 5 Jul 2002 16:15:14 +0000 (16:15 +0000)]
[project @ 2002-07-05 16:15:14 by sof]
When validating the number of source files on the command-line wrt other
options, only consider source files. i.e., don't consider .a's and .o's
that inadvertently, and idempotently, might be present.

21 years ago[project @ 2002-07-05 01:23:45 by mthomas]
mthomas [Fri, 5 Jul 2002 01:23:45 +0000 (01:23 +0000)]
[project @ 2002-07-05 01:23:45 by mthomas]
Allow module names up to 50 chars wide without misalignment of the
profiling table.  For example: "Graphics.UI.ObjectIO.Process.Toolbar"
no longer messes up the output.

The output is now well over 80 chars wide now.

21 years ago[project @ 2002-07-04 13:06:28 by simonpj]
simonpj [Thu, 4 Jul 2002 13:06:28 +0000 (13:06 +0000)]
[project @ 2002-07-04 13:06:28 by simonpj]
Fix error message

21 years ago[project @ 2002-07-04 10:38:49 by simonmar]
simonmar [Thu, 4 Jul 2002 10:38:49 +0000 (10:38 +0000)]
[project @ 2002-07-04 10:38:49 by simonmar]
In hPutLitString, catch the empty string case before calling hPutBuf.
Some older versions of hPutBufFull choke on a zero-length buffer.

Fixes occasional problems with the Sparc native code generator, which
uses SLIT("") in a couple of places.

21 years ago[project @ 2002-07-04 07:09:47 by mthomas]
mthomas [Thu, 4 Jul 2002 07:09:47 +0000 (07:09 +0000)]
[project @ 2002-07-04 07:09:47 by mthomas]
Mingw32 has problems with "long long" display so convert to float output.

21 years ago[project @ 2002-07-03 15:17:15 by sof]
sof [Wed, 3 Jul 2002 15:17:15 +0000 (15:17 +0000)]
[project @ 2002-07-03 15:17:15 by sof]
comment out para mentioning old ghc-win32 page; doesn't contain all that much useful info any longer

21 years ago[project @ 2002-07-03 15:16:09 by sof]
sof [Wed, 3 Jul 2002 15:16:09 +0000 (15:16 +0000)]
[project @ 2002-07-03 15:16:09 by sof]
updated external core frontend info

21 years ago[project @ 2002-07-03 15:15:24 by sof]
sof [Wed, 3 Jul 2002 15:15:25 +0000 (15:15 +0000)]
[project @ 2002-07-03 15:15:24 by sof]
Remove -fcore option, a NOP. (The Core frontend is enabled by feeding
GHC .hcr files.)

21 years ago[project @ 2002-07-03 14:50:27 by wolfgang]
wolfgang [Wed, 3 Jul 2002 14:50:27 +0000 (14:50 +0000)]
[project @ 2002-07-03 14:50:27 by wolfgang]
added a note about MacOS X support

21 years ago[project @ 2002-07-03 11:28:12 by simonmar]
simonmar [Wed, 3 Jul 2002 11:28:12 +0000 (11:28 +0000)]
[project @ 2002-07-03 11:28:12 by simonmar]
version 5.04

21 years ago[project @ 2002-07-03 11:27:43 by simonmar]
simonmar [Wed, 3 Jul 2002 11:27:43 +0000 (11:27 +0000)]
[project @ 2002-07-03 11:27:43 by simonmar]
Announce for 5.04

21 years ago[project @ 2002-07-03 11:16:56 by simonmar]
simonmar [Wed, 3 Jul 2002 11:16:56 +0000 (11:16 +0000)]
[project @ 2002-07-03 11:16:56 by simonmar]
Small fix to the documentation installation

21 years ago[project @ 2002-07-03 11:07:58 by simonmar]
simonmar [Wed, 3 Jul 2002 11:07:58 +0000 (11:07 +0000)]
[project @ 2002-07-03 11:07:58 by simonmar]
Remove old BIN_DIST specific stuff

21 years ago[project @ 2002-07-03 10:04:38 by simonmar]
simonmar [Wed, 3 Jul 2002 10:04:38 +0000 (10:04 +0000)]
[project @ 2002-07-03 10:04:38 by simonmar]
Remove the stuff on building the documentation to the building guide.

21 years ago[project @ 2002-07-03 10:02:19 by simonmar]
simonmar [Wed, 3 Jul 2002 10:02:19 +0000 (10:02 +0000)]
[project @ 2002-07-03 10:02:19 by simonmar]
Add documentation on building the documentation.  (I've moved the
section on building documentation from the User's Guide, it doesn't
really seem appropriate there).

21 years ago[project @ 2002-07-03 08:53:50 by simonpj]
simonpj [Wed, 3 Jul 2002 08:53:50 +0000 (08:53 +0000)]
[project @ 2002-07-03 08:53:50 by simonpj]
Reorder sections of type-system extensions

21 years ago[project @ 2002-07-03 08:30:27 by simonmar]
simonmar [Wed, 3 Jul 2002 08:30:27 +0000 (08:30 +0000)]
[project @ 2002-07-03 08:30:27 by simonmar]
Direct the punter to GHC.Exts rather than GHC.Prim.

21 years ago[project @ 2002-07-02 16:27:38 by simonmar]
simonmar [Tue, 2 Jul 2002 16:27:38 +0000 (16:27 +0000)]
[project @ 2002-07-02 16:27:38 by simonmar]
After loading a package, set the context to something sensible

21 years ago[project @ 2002-07-02 16:21:23 by simonmar]
simonmar [Tue, 2 Jul 2002 16:21:23 +0000 (16:21 +0000)]
[project @ 2002-07-02 16:21:23 by simonmar]
Note that literate Haskell is not supported by hsc2hs at the moment

21 years ago[project @ 2002-07-02 14:55:29 by simonmar]
simonmar [Tue, 2 Jul 2002 14:55:29 +0000 (14:55 +0000)]
[project @ 2002-07-02 14:55:29 by simonmar]
Updates, mainly to get the documentation installed

21 years ago[project @ 2002-07-02 14:53:30 by simonmar]
simonmar [Tue, 2 Jul 2002 14:53:30 +0000 (14:53 +0000)]
[project @ 2002-07-02 14:53:30 by simonmar]
Give the book an id so it gets a nice HTML filename

21 years ago[project @ 2002-07-02 14:48:51 by simonmar]
simonmar [Tue, 2 Jul 2002 14:48:51 +0000 (14:48 +0000)]
[project @ 2002-07-02 14:48:51 by simonmar]
make $(datadir)/html before attempting to install into it

21 years ago[project @ 2002-07-02 14:24:54 by simonmar]
simonmar [Tue, 2 Jul 2002 14:24:54 +0000 (14:24 +0000)]
[project @ 2002-07-02 14:24:54 by simonmar]
For a binary distribution, install docs using the standard
install-docs target rather than hacky copying (needed now that we use
Haddock for building some of the docs too).

21 years ago[project @ 2002-07-02 14:15:22 by simonmar]
simonmar [Tue, 2 Jul 2002 14:15:22 +0000 (14:15 +0000)]
[project @ 2002-07-02 14:15:22 by simonmar]
Don't omit docs when doing a binary dist

21 years ago[project @ 2002-07-02 13:56:22 by simonmar]
simonmar [Tue, 2 Jul 2002 13:56:22 +0000 (13:56 +0000)]
[project @ 2002-07-02 13:56:22 by simonmar]
Remove $(GhcBinDistDocs), this will be done differently

21 years ago[project @ 2002-07-02 13:50:56 by simonmar]
simonmar [Tue, 2 Jul 2002 13:50:56 +0000 (13:50 +0000)]
[project @ 2002-07-02 13:50:56 by simonmar]
Install SGML docs in the right places:

$(datadir)/html for HTML docs
$(datedir)      for everything else

(our main HTML documentation tree will live in $(datadir)/html now).

21 years ago[project @ 2002-07-02 13:20:53 by simonmar]
simonmar [Tue, 2 Jul 2002 13:20:53 +0000 (13:20 +0000)]
[project @ 2002-07-02 13:20:53 by simonmar]
Updates

21 years ago[project @ 2002-07-02 12:24:48 by simonmar]
simonmar [Tue, 2 Jul 2002 12:24:48 +0000 (12:24 +0000)]
[project @ 2002-07-02 12:24:48 by simonmar]
Don't install a SIGFPE handler: this causes us to go into an infinite
loop on a divide by zero on some systems (BSD, but not Linux it
seems).  I don't think we really ought to be ignoring SIGFPE: for
floating point exceptions, the system usually has a way to request
that operations generate exceptional values rather than signals (this
is the required IEEE behaviour), and for integral division we should
really check beforehand for division by zero (we don't yet).

21 years ago[project @ 2002-07-02 10:35:12 by wolfgang]
wolfgang [Tue, 2 Jul 2002 10:35:12 +0000 (10:35 +0000)]
[project @ 2002-07-02 10:35:12 by wolfgang]
mkMarshalCode implemented for PowerPC

21 years ago[project @ 2002-07-02 10:31:39 by wolfgang]
wolfgang [Tue, 2 Jul 2002 10:31:39 +0000 (10:31 +0000)]
[project @ 2002-07-02 10:31:39 by wolfgang]
GHCi lib autogeneration for MacOS X

21 years ago[project @ 2002-07-02 10:22:13 by wolfgang]
wolfgang [Tue, 2 Jul 2002 10:22:13 +0000 (10:22 +0000)]
[project @ 2002-07-02 10:22:13 by wolfgang]
added a missing SymX(__eprintf)
improved handling of local symbols

21 years ago[project @ 2002-07-02 10:16:35 by wolfgang]
wolfgang [Tue, 2 Jul 2002 10:16:35 +0000 (10:16 +0000)]
[project @ 2002-07-02 10:16:35 by wolfgang]
MacOS X: Make GHCi load frameworks specified in package.conf.
Frameworks specified directly from the command line are still ignored.

21 years ago[project @ 2002-07-02 08:41:28 by simonmar]
simonmar [Tue, 2 Jul 2002 08:41:28 +0000 (08:41 +0000)]
[project @ 2002-07-02 08:41:28 by simonmar]
Allow Haddock docs to be turned off by setting $(NO_HADDOCK_DOCS) to "YES"

21 years ago[project @ 2002-07-01 15:19:48 by simonmar]
simonmar [Mon, 1 Jul 2002 15:19:48 +0000 (15:19 +0000)]
[project @ 2002-07-01 15:19:48 by simonmar]
set $(INSTALL_SGML_DOC)

21 years ago[project @ 2002-07-01 15:16:33 by simonmar]
simonmar [Mon, 1 Jul 2002 15:16:33 +0000 (15:16 +0000)]
[project @ 2002-07-01 15:16:33 by simonmar]
Be a bit more verbose when installing documentation

21 years ago[project @ 2002-07-01 14:29:05 by simonmar]
simonmar [Mon, 1 Jul 2002 14:29:05 +0000 (14:29 +0000)]
[project @ 2002-07-01 14:29:05 by simonmar]
I'm pretty sure we don't need to set WAYS in here.

21 years ago[project @ 2002-07-01 14:27:26 by simonmar]
simonmar [Mon, 1 Jul 2002 14:27:26 +0000 (14:27 +0000)]
[project @ 2002-07-01 14:27:26 by simonmar]
Add a top-level HTML page pulling together all the GHC documentation.

21 years ago[project @ 2002-07-01 14:13:16 by simonmar]
simonmar [Mon, 1 Jul 2002 14:13:16 +0000 (14:13 +0000)]
[project @ 2002-07-01 14:13:16 by simonmar]
set ==> users_guide

21 years ago[project @ 2002-07-01 12:31:21 by simonmar]
simonmar [Mon, 1 Jul 2002 12:31:21 +0000 (12:31 +0000)]
[project @ 2002-07-01 12:31:21 by simonmar]
"set" is no more - the User's Guide will be separate from the hslibs and
libraries documentation from now on.

21 years ago[project @ 2002-06-28 14:06:52 by simonpj]
simonpj [Fri, 28 Jun 2002 14:06:54 +0000 (14:06 +0000)]
[project @ 2002-06-28 14:06:52 by simonpj]
-----------------------------------
Fix the CAF info field of error Ids
-----------------------------------

A bizarre bug.   In MkId, we build the Id for various error-y
Ids (like pAT_ERROR_ID) that we grab out of thin air in various
places (like the desugarer).  They were marked as not referring
to any CAFs, but this was a lie!  In fact, they refer to 'untangle'
(see GHC.Err) and thence to a CAF.

Result: GC crash under very obscure circumstances.  (Rob's optimistic
evaluator tickled it.)

Solution: give them more conservative IdInfo.

Two other better solutions to think about:

* Don't grab them out of thin air; instead get them from
  an interface file.

* Treat them as always-live (requires mod to garbage collector)
  so they don't need to be mentioned in SRTs at all

21 years ago[project @ 2002-06-27 15:38:56 by simonmar]
simonmar [Thu, 27 Jun 2002 15:38:58 +0000 (15:38 +0000)]
[project @ 2002-06-27 15:38:56 by simonmar]
Finally fix foreign export and foreign import "wrapper" so that
exceptions raised during the call are handled properly rather than
causing the RTS to bomb out.

In particular, calling System.exitWith in a foreign export will cause
the program to terminate cleanly with the desired exit code.  All
other exceptions are printed on stderr (and the program is
terminated).

Details:

GHC.TopHandler.runMain is now called runIO, and has type IO a -> IO a
(previously it had type IO a -> IO (), but that's not general enough
for a foreign export).  The stubs for foreign export and forein import
"wrapper" now automatically wrap the computation in runIO or its dual,
runNonIO.  It turned out to be simpler to do it this way than to do
the wrapping in Haskell land (plain foreign exports don't have
wrappers in Haskell).

21 years ago[project @ 2002-06-27 15:15:05 by simonmar]
simonmar [Thu, 27 Jun 2002 15:15:05 +0000 (15:15 +0000)]
[project @ 2002-06-27 15:15:05 by simonmar]
Don't process OPTIONS in the HsPp phase as well as the cpp phase.
Fixes problems with include files appearing twice in stub files.

21 years ago[project @ 2002-06-27 12:17:47 by simonmar]
simonmar [Thu, 27 Jun 2002 12:17:48 +0000 (12:17 +0000)]
[project @ 2002-06-27 12:17:47 by simonmar]
Improve the documentation for hierarchical libraries.

21 years ago[project @ 2002-06-27 09:41:18 by simonmar]
simonmar [Thu, 27 Jun 2002 09:41:18 +0000 (09:41 +0000)]
[project @ 2002-06-27 09:41:18 by simonmar]
Updates to make sure that the CVS instructions mention 'libraries' in
addition to 'hslibs'.

21 years ago[project @ 2002-06-26 13:33:05 by simonpj]
simonpj [Wed, 26 Jun 2002 13:33:05 +0000 (13:33 +0000)]
[project @ 2002-06-26 13:33:05 by simonpj]
Add 5.04 release notes draft

21 years ago[project @ 2002-06-26 08:20:25 by stolz]
stolz [Wed, 26 Jun 2002 08:20:25 +0000 (08:20 +0000)]
[project @ 2002-06-26 08:20:25 by stolz]
The "Ouch"-commit:
 - didn't notice my local change to libraries/Makefile
 - forgot to register new files

21 years ago[project @ 2002-06-26 08:18:38 by stolz]
stolz [Wed, 26 Jun 2002 08:18:45 +0000 (08:18 +0000)]
[project @ 2002-06-26 08:18:38 by stolz]
- Make TSO "stable" again: The thread label was changing the size of the
   TSO if you were building a debugging-RTS, leading to binary
   incompatibility. Now we map TSOs to strings using Hash.c.

- API change for labelThread: Label arbitrary threads.

21 years ago[project @ 2002-06-25 15:49:58 by mthomas]
mthomas [Tue, 25 Jun 2002 15:49:59 +0000 (15:49 +0000)]
[project @ 2002-06-25 15:49:58 by mthomas]
Latest Mingw32 needs libmingwex, but for now not when installing as
bundled gcc is out of date.

21 years ago[project @ 2002-06-25 12:05:14 by simonmar]
simonmar [Tue, 25 Jun 2002 12:05:15 +0000 (12:05 +0000)]
[project @ 2002-06-25 12:05:14 by simonmar]
Various hacking on the User's Guide to knock it into shape for the release.

Committers: please take a look at the 5.04 release notes and feel free
to tweak or embelish the description for your favourite new feature.

21 years ago[project @ 2002-06-25 11:29:55 by simonmar]
simonmar [Tue, 25 Jun 2002 11:29:55 +0000 (11:29 +0000)]
[project @ 2002-06-25 11:29:55 by simonmar]
Update license date.

21 years ago[project @ 2002-06-25 08:32:19 by simonmar]
simonmar [Tue, 25 Jun 2002 08:32:19 +0000 (08:32 +0000)]
[project @ 2002-06-25 08:32:19 by simonmar]
We need Happy version 1.13 now, not 1.9

21 years ago[project @ 2002-06-24 14:50:59 by simonmar]
simonmar [Mon, 24 Jun 2002 14:50:59 +0000 (14:50 +0000)]
[project @ 2002-06-24 14:50:59 by simonmar]
use $(GHC_INPLACE) rather than $(GHC) when pre-processing source for Haddock.

21 years ago[project @ 2002-06-24 14:38:06 by simonmar]
simonmar [Mon, 24 Jun 2002 14:38:06 +0000 (14:38 +0000)]
[project @ 2002-06-24 14:38:06 by simonmar]
dump and read interfaces for Haddock

21 years ago[project @ 2002-06-24 14:35:44 by simonmar]
simonmar [Mon, 24 Jun 2002 14:35:44 +0000 (14:35 +0000)]
[project @ 2002-06-24 14:35:44 by simonmar]
comment updates only

21 years ago[project @ 2002-06-24 14:35:28 by simonmar]
simonmar [Mon, 24 Jun 2002 14:35:28 +0000 (14:35 +0000)]
[project @ 2002-06-24 14:35:28 by simonmar]
add HADDOCK_OPTS

21 years ago[project @ 2002-06-21 13:34:42 by simonpj]
simonpj [Fri, 21 Jun 2002 13:34:43 +0000 (13:34 +0000)]
[project @ 2002-06-21 13:34:42 by simonpj]
---------------------------------------------
    Calculate the free vars of a type 'right'
---------------------------------------------

type C u a = u

Question: is 'a' free in 'C u a'?

I think the answer should be 'no'; see typecheck/should_compile/tc157.hs
for an example of why it matters.  This commit makes it so, and adds
comments to explain a dark corner in the zonking code.

21 years ago[project @ 2002-06-21 13:31:50 by simonpj]
simonpj [Fri, 21 Jun 2002 13:31:50 +0000 (13:31 +0000)]
[project @ 2002-06-21 13:31:50 by simonpj]
Dont use the isomorphic-newtype deriving trick for recursive newtypes

21 years ago[project @ 2002-06-21 12:53:03 by simonmar]
simonmar [Fri, 21 Jun 2002 12:53:03 +0000 (12:53 +0000)]
[project @ 2002-06-21 12:53:03 by simonmar]
bugfix

21 years ago[project @ 2002-06-20 16:03:56 by simonmar]
simonmar [Thu, 20 Jun 2002 16:03:56 +0000 (16:03 +0000)]
[project @ 2002-06-20 16:03:56 by simonmar]
Add $(INSTALL_IFACES) and $(INSTALL_IFACES_WITH_DIRS) for installing
interfaces (the latter replaces $(INSTALL_DATAS_WITH_DIRS).

21 years ago[project @ 2002-06-20 16:03:19 by simonmar]
simonmar [Thu, 20 Jun 2002 16:03:19 +0000 (16:03 +0000)]
[project @ 2002-06-20 16:03:19 by simonmar]
Add support for building & installing Haddock docs.

Also minor cleanups: use $(ifacedir) rather than overriding $(datadir)
when installing interfaces.

21 years ago[project @ 2002-06-20 16:00:18 by simonmar]
simonmar [Thu, 20 Jun 2002 16:00:18 +0000 (16:00 +0000)]
[project @ 2002-06-20 16:00:18 by simonmar]
Add:
$(ifacedir) where we install interfaces
$(HADDOCK_INPLACE) how to invoke Haddock