ghc-hetmet.git
20 years ago[project @ 2003-06-19 10:42:24 by simonmar]
simonmar [Thu, 19 Jun 2003 10:42:26 +0000 (10:42 +0000)]
[project @ 2003-06-19 10:42:24 by simonmar]
Add raiseIO# primop.

This is part of ensuring that exceptions are *precise* in the IO monad
(as opposed to imprecise exceptions in the pure world).  If we allow
the strictness analyser to see the definition of throwIO:

  throwIO e = IO $ \s -> throw e

then it might re-order evaluation in the IO monad, with the result
that we get _|_ instead of an exception, or one kind of exception when
we were expecting another.  We therefore must prevent the strictness
analyser from doing these reorderings in the IO monad.  Hiding the
definition of throwIO by making it a primop solves part of the problem
(there's more to come).

See SourceForge bug #752149.

20 years ago[project @ 2003-06-19 10:35:37 by simonmar]
simonmar [Thu, 19 Jun 2003 10:35:37 +0000 (10:35 +0000)]
[project @ 2003-06-19 10:35:37 by simonmar]
Style nits only

20 years ago[project @ 2003-06-17 23:26:30 by sof]
sof [Tue, 17 Jun 2003 23:26:30 +0000 (23:26 +0000)]
[project @ 2003-06-17 23:26:30 by sof]
apply SysTools.normalisePath to all fileish_args, not just source files

20 years ago[project @ 2003-06-16 15:32:16 by simonpj]
simonpj [Mon, 16 Jun 2003 15:32:18 +0000 (15:32 +0000)]
[project @ 2003-06-16 15:32:16 by simonpj]
--------------------------
Remove some wired-in types
--------------------------

ptrTyCon, funPtrTyCon, addrTyCon, stablePtrTyCon have no business
being wired in. This commit makes them into knownKey Names, which
is much better.

20 years ago[project @ 2003-06-16 15:31:13 by simonpj]
simonpj [Mon, 16 Jun 2003 15:31:13 +0000 (15:31 +0000)]
[project @ 2003-06-16 15:31:13 by simonpj]
----------------------------------
Fix the "no interface file" bug
----------------------------------

The problem was this.

* Module A imports module B

* ghc --make A

* Remove B.hs, B.hi, and edit A so that it doesn't import B

* ghc --make A

Then GHC would fail, saying it can't find an interface file for B.
Reason: when loading A.hi (loadOldIface) it makes a Name for the
usages it finds.  To make a Name for B things, it needs B's package;
but it can't find B.hi, so it bombs instead.

Solution: don't complain if loadOldIface fails; instead behave just like
if you can't find A.hi.

Fixes a long-standing infelicity.

20 years ago[project @ 2003-06-14 16:54:24 by panne]
panne [Sat, 14 Jun 2003 16:54:24 +0000 (16:54 +0000)]
[project @ 2003-06-14 16:54:24 by panne]
Attempt to fix the nightly builds (untested):
Use our forefathers' syntax to import rawSystem

20 years ago[project @ 2003-06-13 09:27:13 by simonmar]
simonmar [Fri, 13 Jun 2003 09:27:13 +0000 (09:27 +0000)]
[project @ 2003-06-13 09:27:13 by simonmar]
Force -fvia-C when profiling or ticky-ticky is on.

20 years ago[project @ 2003-06-13 08:45:32 by simonmar]
simonmar [Fri, 13 Jun 2003 08:45:32 +0000 (08:45 +0000)]
[project @ 2003-06-13 08:45:32 by simonmar]
- add the OPTIONS pragma to the section on pragmas, with a cross-ref
  to the place it is described.

- tidy up the section on pragams: re-indent and sort the sections
  alphabetically.

20 years ago[project @ 2003-06-13 07:44:15 by simonpj]
simonpj [Fri, 13 Jun 2003 07:44:15 +0000 (07:44 +0000)]
[project @ 2003-06-13 07:44:15 by simonpj]
Clarifying remarks about libdir/bindir

20 years ago[project @ 2003-06-13 06:22:23 by simonpj]
simonpj [Fri, 13 Jun 2003 06:22:23 +0000 (06:22 +0000)]
[project @ 2003-06-13 06:22:23 by simonpj]
Add the commit message as a comment

20 years ago[project @ 2003-06-12 16:50:19 by simonpj]
simonpj [Thu, 12 Jun 2003 16:50:19 +0000 (16:50 +0000)]
[project @ 2003-06-12 16:50:19 by simonpj]
----------------------------------
Fix the no-GHC.Err problem in GHCi
----------------------------------

Merge this to the stable branch.

Pattern-match failure error-ids are wired-in, which means the
GHC.Err interface isn't loaded, which in turn confused the linker
when tried to find what package GHC.Err is from.

Now we exclude wired-in names from the free variables looked at
by the linker (in Linker.linkExpr), and make sure the base package
is loaded unconditionally (DriverState.initPackageList).

20 years ago[project @ 2003-06-12 16:25:31 by simonmar]
simonmar [Thu, 12 Jun 2003 16:25:31 +0000 (16:25 +0000)]
[project @ 2003-06-12 16:25:31 by simonmar]
Urk, don't quote/escape the command name on Windows, because the
compiler is exceedingly naughty and sometimes uses 'perl "..."' as the
command name.

20 years ago[project @ 2003-06-12 16:06:05 by simonmar]
simonmar [Thu, 12 Jun 2003 16:06:07 +0000 (16:06 +0000)]
[project @ 2003-06-12 16:06:05 by simonmar]
Change the type of System.Cmd.rawSystem:

  rawSystem :: FilePath -> [String] -> IO ExitCode

and implement it properly on both Windows & Unix.  The intended
meaning is that the program is executed with *exactly* these
arguments.

We now re-use this rawSystem in the compiler itself (using it directly
from the library if __GLASGOW_HASKELL__ >= 601).

The previous implementation of SysTools.runSomething was broken on
4.08, because Posix.executeFile was broken.  However, implementing the
new rawSystem on 4.08 is tricky, because it uses the FFI marshalling
libraries which weren't present on 4.08.  Hence, bootstrapping from
4.08 is now not possible (it was already not possible on Windows).  It
could be made possible by importing enough FFI marshalling support,
but I won't bother doing that unless/until it is needed.

20 years ago[project @ 2003-06-12 14:37:21 by simonpj]
simonpj [Thu, 12 Jun 2003 14:37:21 +0000 (14:37 +0000)]
[project @ 2003-06-12 14:37:21 by simonpj]
Comments

20 years ago[project @ 2003-06-12 14:36:59 by simonpj]
simonpj [Thu, 12 Jun 2003 14:36:59 +0000 (14:36 +0000)]
[project @ 2003-06-12 14:36:59 by simonpj]
Fix lack of deriving(Typeable) in existentials; merge to stable

20 years ago[project @ 2003-06-12 11:13:07 by simonmar]
simonmar [Thu, 12 Jun 2003 11:13:07 +0000 (11:13 +0000)]
[project @ 2003-06-12 11:13:07 by simonmar]
More intelligent test for ghc-pkg:  if there was a --with-ghc option
to configure, then we now use that as a hint for which ghc-pkg to use.

20 years ago[project @ 2003-06-12 09:48:17 by simonpj]
simonpj [Thu, 12 Jun 2003 09:48:17 +0000 (09:48 +0000)]
[project @ 2003-06-12 09:48:17 by simonpj]
Add more comments from Sigbjorn

20 years ago[project @ 2003-06-11 10:28:43 by simonmar]
simonmar [Wed, 11 Jun 2003 10:28:43 +0000 (10:28 +0000)]
[project @ 2003-06-11 10:28:43 by simonmar]
Ugh. gcc 3.3 has started turning 'static int foo = 0' into a .comm
declaration, presumably to save a few bytes in the binary.  So we must
Deal With It.

20 years ago[project @ 2003-06-11 07:23:06 by simonpj]
simonpj [Wed, 11 Jun 2003 07:23:06 +0000 (07:23 +0000)]
[project @ 2003-06-11 07:23:06 by simonpj]
Add coments

20 years ago[project @ 2003-06-10 17:58:25 by sof]
sof [Tue, 10 Jun 2003 17:58:25 +0000 (17:58 +0000)]
[project @ 2003-06-10 17:58:25 by sof]
hardtop{mingw32}: for now, stick with 'cygpath -w' instead of 'cygpath -m'

20 years ago[project @ 2003-06-10 17:54:56 by sof]
sof [Tue, 10 Jun 2003 17:54:56 +0000 (17:54 +0000)]
[project @ 2003-06-10 17:54:56 by sof]
Normalise filepaths of all source files before doing any processing
of them. This, unsurprisingly, simplifies filepath comparisons on
platforms that support multiple path syntaxes (cf., bidirectional
slashes under mingw/win32.)

This tentatively fixes Sourceforge bug #751175; assuming now bad
interactions arise, a similar change will be made on STABLE.

20 years ago[project @ 2003-06-10 17:46:06 by sof]
sof [Tue, 10 Jun 2003 17:46:06 +0000 (17:46 +0000)]
[project @ 2003-06-10 17:46:06 by sof]
- renamed unDosifyPath as normalisePath
- renamed dosifyPath as platformPath
- dropped dosifyPaths; unused.
- platformPath (ne dosifyPath): drop slicing off of /cygdrive prefixes

20 years ago[project @ 2003-06-10 13:40:11 by simonpj]
simonpj [Tue, 10 Jun 2003 13:40:12 +0000 (13:40 +0000)]
[project @ 2003-06-10 13:40:11 by simonpj]
----------------------------------
Fix the crossDllArg crash (take 2)
----------------------------------

I got this fix completely wrong, again.

The original CoreUtils.rhsIsNonUpd is now renamed again, to
CoreUtils.rhsIsStatic. Yet more comments explain its (now
simplified) working.

20 years ago[project @ 2003-06-10 13:21:26 by simonmar]
simonmar [Tue, 10 Jun 2003 13:21:26 +0000 (13:21 +0000)]
[project @ 2003-06-10 13:21:26 by simonmar]
Add some more optional space to a pattern for compatibility with GCC 3.3.

20 years ago[project @ 2003-06-10 13:01:09 by simonmar]
simonmar [Tue, 10 Jun 2003 13:01:09 +0000 (13:01 +0000)]
[project @ 2003-06-10 13:01:09 by simonmar]
Explain that the hierarchy is not cleanly divided into packages - two
people (at least) have so far been confused that System.Posix.Types is
in base but System.Posix is in unix.

20 years ago[project @ 2003-06-10 11:57:02 by simonmar]
simonmar [Tue, 10 Jun 2003 11:57:02 +0000 (11:57 +0000)]
[project @ 2003-06-10 11:57:02 by simonmar]
Tidy up the $hardtop test:

  - use cygpath -m to calculate $hardtop on Windows, rather than
    our own partial solution.  Fixes building under c:/cygwin
    on Windows.

  - remove some old stuff dealing with paths we used to use at
    Glasgow.

  - use proper AC_MSG macros instead of explicit echo.

MERGE TO STABLE

20 years ago[project @ 2003-06-09 15:37:37 by simonpj]
simonpj [Mon, 9 Jun 2003 15:37:38 +0000 (15:37 +0000)]
[project @ 2003-06-09 15:37:37 by simonpj]
-------------------------
Fix the crossDllArg crash
-------------------------

Test codeGen/should_compile/cg004 tests for this one.  The problem was
that the test for static-ness (i.e. no need to CAFify the thing) encountered
a form we didn't previously expect.  (See cg004)

This fix tidies up CoreUtils.rhsIsNonUpd, which is the original entry point,
renaming it CoreUtils.hasNoRedexes

20 years ago[project @ 2003-06-09 14:10:04 by matthewc]
matthewc [Mon, 9 Jun 2003 14:10:04 +0000 (14:10 +0000)]
[project @ 2003-06-09 14:10:04 by matthewc]
Change argument of stg_exit to int in line with definition.

20 years ago[project @ 2003-06-09 13:17:38 by matthewc]
matthewc [Mon, 9 Jun 2003 13:17:41 +0000 (13:17 +0000)]
[project @ 2003-06-09 13:17:38 by matthewc]
Update IA64 port to work with gcc 3.3.

20 years ago[project @ 2003-06-09 11:41:47 by simonmar]
simonmar [Mon, 9 Jun 2003 11:41:47 +0000 (11:41 +0000)]
[project @ 2003-06-09 11:41:47 by simonmar]
Add a TODO.

20 years ago[project @ 2003-06-09 11:40:06 by simonmar]
simonmar [Mon, 9 Jun 2003 11:40:06 +0000 (11:40 +0000)]
[project @ 2003-06-09 11:40:06 by simonmar]
Add comment.

20 years ago[project @ 2003-06-09 11:22:14 by simonpj]
simonpj [Mon, 9 Jun 2003 11:22:14 +0000 (11:22 +0000)]
[project @ 2003-06-09 11:22:14 by simonpj]
Remove duplicate infix type-con section (thanks to Ross P)

20 years ago[project @ 2003-06-08 20:33:31 by ross]
ross [Sun, 8 Jun 2003 20:33:31 +0000 (20:33 +0000)]
[project @ 2003-06-08 20:33:31 by ross]
fix more gcc 3.3 preprocessor warnings

20 years ago[project @ 2003-06-08 18:12:23 by igloo]
igloo [Sun, 8 Jun 2003 18:12:27 +0000 (18:12 +0000)]
[project @ 2003-06-08 18:12:23 by igloo]
Introduce a ListP for consistency with ListE. Splicing in something with
a list pattern now works too.

Added various list tests.

20 years ago[project @ 2003-06-07 11:20:47 by ross]
ross [Sat, 7 Jun 2003 11:20:49 +0000 (11:20 +0000)]
[project @ 2003-06-07 11:20:47 by ross]
Add $($(HcFlavour)_HAPPY_OPTS) to HAPPY_OPTS, where HcFlavour defaults to GHC.
Packages may GHC_HAPPY_OPTS if they don't want -agc with ghc.

20 years ago[project @ 2003-06-06 16:04:23 by igloo]
igloo [Fri, 6 Jun 2003 16:04:26 +0000 (16:04 +0000)]
[project @ 2003-06-06 16:04:23 by igloo]
Template Haskell Renamings as described in
http://www.haskell.org/pipermail/template-haskell/2003-May/000110.html

20 years ago[project @ 2003-06-06 15:25:08 by sof]
sof [Fri, 6 Jun 2003 15:25:08 +0000 (15:25 +0000)]
[project @ 2003-06-06 15:25:08 by sof]
upd binary mode addition

20 years ago[project @ 2003-06-06 12:35:19 by simonpj]
simonpj [Fri, 6 Jun 2003 12:35:19 +0000 (12:35 +0000)]
[project @ 2003-06-06 12:35:19 by simonpj]
Add a note about ^Z on Win32

20 years ago[project @ 2003-06-06 12:30:38 by reid]
reid [Fri, 6 Jun 2003 12:30:38 +0000 (12:30 +0000)]
[project @ 2003-06-06 12:30:38 by reid]
Add definition for comma.  Include GC_C_OBJS in BUILD_LIB

20 years ago[project @ 2003-06-06 11:18:35 by simonmar]
simonmar [Fri, 6 Jun 2003 11:18:35 +0000 (11:18 +0000)]
[project @ 2003-06-06 11:18:35 by simonmar]
unbreak GHCi

20 years ago[project @ 2003-06-06 10:43:22 by panne]
panne [Fri, 6 Jun 2003 10:43:22 +0000 (10:43 +0000)]
[project @ 2003-06-06 10:43:22 by panne]
Quick fix: Escape double quotes when quoting; affects only WinDoze
builds. While I'm there: Move quoting hell into #ifdef, so its scope
is obvious and changed an #ifndef into an #ifdef (negated conditions
are evil, at least for my small brain).

20 years ago[project @ 2003-06-06 09:43:50 by stolz]
stolz [Fri, 6 Jun 2003 09:43:50 +0000 (09:43 +0000)]
[project @ 2003-06-06 09:43:50 by stolz]
Drop unused import of System.system.

20 years ago[project @ 2003-06-05 16:50:30 by reid]
reid [Thu, 5 Jun 2003 16:50:30 +0000 (16:50 +0000)]
[project @ 2003-06-05 16:50:30 by reid]
Swap rules for building .o files by compiling and by assembling so that we compile if possible

20 years ago[project @ 2003-06-05 14:39:44 by reid]
reid [Thu, 5 Jun 2003 14:39:44 +0000 (14:39 +0000)]
[project @ 2003-06-05 14:39:44 by reid]
Put GC_C_OBJS into OBJS not STUBOBJS

20 years ago[project @ 2003-06-05 14:05:36 by reid]
reid [Thu, 5 Jun 2003 14:05:36 +0000 (14:05 +0000)]
[project @ 2003-06-05 14:05:36 by reid]
s/green-card/greencard/

20 years ago[project @ 2003-06-05 14:04:28 by stolz]
stolz [Thu, 5 Jun 2003 14:04:28 +0000 (14:04 +0000)]
[project @ 2003-06-05 14:04:28 by stolz]
Work around broken getProcessStatus in 5.04.x

20 years ago[project @ 2003-06-05 12:29:31 by reid]
reid [Thu, 5 Jun 2003 12:29:31 +0000 (12:29 +0000)]
[project @ 2003-06-05 12:29:31 by reid]
Add GC_C_OBJS to MOSTLY_CLEAN_FILES

20 years ago[project @ 2003-06-05 12:09:52 by panne]
panne [Thu, 5 Jun 2003 12:09:52 +0000 (12:09 +0000)]
[project @ 2003-06-05 12:09:52 by panne]
Attention: Tentative fix only! But at least this works much better
than before.

From the Linux man page for waitpid:

   [...]
   ERRORS
          ECHILD if  the  process specified in pid does not exist or
                 is not a child of the calling process.   (This  can
                 happen  for  one's  own  child  if  the  action for
                 SIGCHLD is set to SIG_IGN. See also the LINUX NOTES
                 section about threads.)
   [...]

Consequently, we don't ignore SIGCHLD anymore. Further changes:

   * SIGCONT is not ignored anymore, either. What was this for?

   * Don't use WUNTRACED in waitpid. Again, what was this for?

20 years ago[project @ 2003-06-05 10:11:22 by simonmar]
simonmar [Thu, 5 Jun 2003 10:11:22 +0000 (10:11 +0000)]
[project @ 2003-06-05 10:11:22 by simonmar]
- Fix a couple of bugs in yesterday's pipeline cleanup.

- Do some more tidying: share the code for filename generation between
  the two entry points to HscMain, and also share some of the other
  machinery in the Hsc phase.  This fixes some wibbles (things that
  were done in --make mode but not in one-shot, and vice-versa).

  One thing that works now is that if you say 'ghc -keep-hc-files Foo.hs',
  then it automatically switches to -fvia-C mode to generate the .hc file.

20 years ago[project @ 2003-06-04 16:52:06 by stolz]
stolz [Wed, 4 Jun 2003 16:52:07 +0000 (16:52 +0000)]
[project @ 2003-06-04 16:52:06 by stolz]
System.Cmd.system-less invocation of sub-tasks
(Sven approved reverting the OpenGL-Makefile change)

20 years ago[project @ 2003-06-04 15:47:58 by simonmar]
simonmar [Wed, 4 Jun 2003 15:47:59 +0000 (15:47 +0000)]
[project @ 2003-06-04 15:47:58 by simonmar]
Grrr, started off making a small bugfix and ended up doing a major
cleanup operartion.

Anyway, the problem was that -odir wasn't putting the object files in
the right place when the module in question has a hierarchical name.
This was due to the object filename being generated in two different
places: once by the compilation pipeline machinery, and again in the
Finder.  It now works properly when --make is used; I haven't managed
to fix it for one-shot compilations though (some replumbing is
needed).

While I was here, I cleaned up the compilation pipeline machinery
somewhat.  The previous scheme of generating a data structure
representing the phases that need to be executed before actually
executing them was wrong because the structure of the pipeline can
change while it is being executed (eg. if we see {-# OPTIONS -fasm #-}
during the CPP phase).  There were various hacks to deal with this,
but it turned out to be quite messy.

So the new story is that each compilation phase returns the name of
the next phase to execute, and also figures out which file to put its
output in.  This unfortunately means that the knowledge about what
phases are done in what order is now spread throughout the module, but
there are fewer hacks at the higher levels, and overall it seems to be
an improvement.

20 years ago[project @ 2003-06-04 15:18:29 by panne]
panne [Wed, 4 Jun 2003 15:18:30 +0000 (15:18 +0000)]
[project @ 2003-06-04 15:18:29 by panne]
Tentative fixes for Cywgin/MinGW-style pathname confusion

20 years ago[project @ 2003-06-04 15:11:17 by panne]
panne [Wed, 4 Jun 2003 15:11:17 +0000 (15:11 +0000)]
[project @ 2003-06-04 15:11:17 by panne]
Always check for X Window System, otherwise $no_x is not set to "yes"
when there is no X11. Confusing...

20 years ago[project @ 2003-06-04 12:37:09 by reid]
reid [Wed, 4 Jun 2003 12:37:09 +0000 (12:37 +0000)]
[project @ 2003-06-04 12:37:09 by reid]
All changes in this commit are standalone changes only but could usefully be used inplace as well: Don't store package.conf.installed; use a pipe to avoid problems from using stale package.confs.  Also, use PACKAGE, PACKAGE_DEPS, LIBRARY and LIBDIR settings from Makefile when installing pkg.  This requires some ugly sed hackery to overcome ghc-pkg not allowing trailing commas in a list: [a,b,c,]

20 years ago[project @ 2003-06-03 15:07:11 by panne]
panne [Tue, 3 Jun 2003 15:07:11 +0000 (15:07 +0000)]
[project @ 2003-06-03 15:07:11 by panne]
Nuked debugging output

20 years ago[project @ 2003-06-03 11:44:28 by simonpj]
simonpj [Tue, 3 Jun 2003 11:44:28 +0000 (11:44 +0000)]
[project @ 2003-06-03 11:44:28 by simonpj]
Windows ssh info

20 years ago[project @ 2003-06-03 09:41:48 by ross]
ross [Tue, 3 Jun 2003 09:41:51 +0000 (09:41 +0000)]
[project @ 2003-06-03 09:41:48 by ross]
change trailing comments on #else/#endif lines to C style to avoid
warnings from gcc 3.3's preprocessor.

20 years ago[project @ 2003-06-03 09:37:14 by ross]
ross [Tue, 3 Jun 2003 09:37:15 +0000 (09:37 +0000)]
[project @ 2003-06-03 09:37:14 by ross]
fix for gcc 3.3 preprocessor: no layout, end-of-line comments or complex
Haskell comments in macro arguments.

please merge to STABLE

20 years ago[project @ 2003-06-02 16:24:07 by sof]
sof [Mon, 2 Jun 2003 16:24:07 +0000 (16:24 +0000)]
[project @ 2003-06-02 16:24:07 by sof]
Surround copied argv entries in double quotes to avoid quoting issues
(but don't quote me on that.)

Merge to STABLE.

20 years ago[project @ 2003-06-02 16:10:19 by simonpj]
simonpj [Mon, 2 Jun 2003 16:10:19 +0000 (16:10 +0000)]
[project @ 2003-06-02 16:10:19 by simonpj]
Wibbles to tuples

20 years ago[project @ 2003-06-02 14:26:54 by simonpj]
simonpj [Mon, 2 Jun 2003 14:26:55 +0000 (14:26 +0000)]
[project @ 2003-06-02 14:26:54 by simonpj]
Wibbles to nested tuples

20 years ago[project @ 2003-06-02 13:28:08 by simonpj]
simonpj [Mon, 2 Jun 2003 13:28:09 +0000 (13:28 +0000)]
[project @ 2003-06-02 13:28:08 by simonpj]
-------------------------------------
      Fix the big-tuple-from-desugaring problem
-------------------------------------

The desugarer generates a tuple from
- mutually recursive bindings
- pattern bindings

If either bind a lot of variables, GHC can generate a big
tuple that isn't in the library, with subsequent disaster.

This commit fixes the problem, by using nested tuples.  It
does *not* fix the problem with big tuples written by the
user. And there's still a potential desugarer problem with
parallel list comprehensions that bind a lot of variables
(and parallel array comprehensions) -- but I expect they are
much much rarer.

The fix isn't fully tested yet -- I'll try to do that today.

20 years ago[project @ 2003-06-02 13:27:53 by simonpj]
simonpj [Mon, 2 Jun 2003 13:27:53 +0000 (13:27 +0000)]
[project @ 2003-06-02 13:27:53 by simonpj]
Add comments

20 years ago[project @ 2003-06-02 13:27:33 by simonpj]
simonpj [Mon, 2 Jun 2003 13:27:41 +0000 (13:27 +0000)]
[project @ 2003-06-02 13:27:33 by simonpj]
Prune imports

20 years ago[project @ 2003-06-02 13:24:49 by reid]
reid [Mon, 2 Jun 2003 13:24:49 +0000 (13:24 +0000)]
[project @ 2003-06-02 13:24:49 by reid]
GC-generated stubs go in STUBOBJS not OBJS

20 years ago[project @ 2003-06-02 08:52:28 by simonmar]
simonmar [Mon, 2 Jun 2003 08:52:28 +0000 (08:52 +0000)]
[project @ 2003-06-02 08:52:28 by simonmar]
Back out part of previous commit: the text has already changed to not
recommend using 'cvs update -Pd' any more.

20 years ago[project @ 2003-06-02 07:19:15 by stolz]
stolz [Mon, 2 Jun 2003 07:19:15 +0000 (07:19 +0000)]
[project @ 2003-06-02 07:19:15 by stolz]
- warn about running 'cvs update -Pd'
    (run into by: Peter Simons)
- suggest running 'make html' in fptools/libraries before committing
    (*cough*)
- fix speling blunders

(Could somebody validate the SGML, please? Can't do it here :/)

20 years ago[project @ 2003-06-02 06:44:12 by mthomas]
mthomas [Mon, 2 Jun 2003 06:44:12 +0000 (06:44 +0000)]
[project @ 2003-06-02 06:44:12 by mthomas]
Correct a spelling mistake while doing some reading.

20 years ago[project @ 2003-05-31 16:26:33 by panne]
panne [Sat, 31 May 2003 16:26:33 +0000 (16:26 +0000)]
[project @ 2003-05-31 16:26:33 by panne]
*sigh* Reverted to DocBook 3.1, because I'm completely unable to
provide a catalog for DocBook 4.x on SuSE 8.1. If there is somebody
out there with more SGML knowledge, the autoconf magic should be fixed
so that a newer DTD can be used...

Was there a deep reason for DocBook 4.1 or the usual "newer is better"?
:-]

20 years ago[project @ 2003-05-30 22:29:41 by wolfgang]
wolfgang [Fri, 30 May 2003 22:29:41 +0000 (22:29 +0000)]
[project @ 2003-05-30 22:29:41 by wolfgang]
MacOS X:
pass the -framework option and its parameter as two separate arguments,
i.e. ["-framework", "Foo"] instead of ["-framework Foo"].
This is necessary because all options passed to gcc are now quoted
(since rev. 1.88 of SysTools.lhs), and "-framework Foo" is not a valid
option (while "-framework" "Foo" is).

MERGE TO STABLE

20 years ago[project @ 2003-05-30 13:32:20 by simonmar]
simonmar [Fri, 30 May 2003 13:32:21 +0000 (13:32 +0000)]
[project @ 2003-05-30 13:32:20 by simonmar]
Rename --list-packages-local to --list-local-packages, because it
seems that having one long option which is a prefix of another doesn't
work too well with GetOpt.

20 years ago[project @ 2003-05-30 10:13:05 by simonmar]
simonmar [Fri, 30 May 2003 10:13:05 +0000 (10:13 +0000)]
[project @ 2003-05-30 10:13:05 by simonmar]
Don't unconditionally define WANT_DOTNET_SUPPORT.

20 years ago[project @ 2003-05-30 09:09:09 by simonmar]
simonmar [Fri, 30 May 2003 09:09:09 +0000 (09:09 +0000)]
[project @ 2003-05-30 09:09:09 by simonmar]
merge rev. 1.116.2.1:

FPTOOLS_O_BINARY: lift #include to the top.

20 years ago[project @ 2003-05-30 09:06:24 by simonmar]
simonmar [Fri, 30 May 2003 09:06:24 +0000 (09:06 +0000)]
[project @ 2003-05-30 09:06:24 by simonmar]
merge rev. 1.119.2.1:

Couple of extra local syms needed when using mingw2

20 years ago[project @ 2003-05-30 09:04:28 by simonmar]
simonmar [Fri, 30 May 2003 09:04:28 +0000 (09:04 +0000)]
[project @ 2003-05-30 09:04:28 by simonmar]
Fix Network.Socket link.

20 years ago[project @ 2003-05-29 14:47:05 by reid]
reid [Thu, 29 May 2003 14:47:05 +0000 (14:47 +0000)]
[project @ 2003-05-29 14:47:05 by reid]
Use --force which ghc-pkg'ing standalone libs because ghc-pkg is run before the libs have actually been installed.  It would be better to delay running ghc-pkg.

20 years ago[project @ 2003-05-29 14:39:26 by sof]
sof [Thu, 29 May 2003 14:39:31 +0000 (14:39 +0000)]
[project @ 2003-05-29 14:39:26 by sof]
Support for interop'ing with .NET via FFI declarations along the
lines of what Hugs98.NET offers, see

 http://haskell.org/pipermail/cvs-hugs/2003-March/001723.html

for FFI decl details.

To enable, configure with --enable-dotnet + have a look
in ghc/rts/dotnet/Makefile for details of what tools are needed to
build the .NET interop layer (tools from VS.NET / Framework SDK.)

The commit doesn't include some library additions + wider-scale
testing is required before this extension can be regarded as available
for general use. 'foreign import dotnet' is currently only supported
by the C backend.

20 years ago[project @ 2003-05-29 13:54:44 by simonmar]
simonmar [Thu, 29 May 2003 13:54:44 +0000 (13:54 +0000)]
[project @ 2003-05-29 13:54:44 by simonmar]
wibble to previous commit

20 years ago[project @ 2003-05-29 13:37:59 by simonmar]
simonmar [Thu, 29 May 2003 13:37:59 +0000 (13:37 +0000)]
[project @ 2003-05-29 13:37:59 by simonmar]
Prune more autoconf droppings from a source dist.

20 years ago[project @ 2003-05-29 13:36:39 by simonmar]
simonmar [Thu, 29 May 2003 13:36:39 +0000 (13:36 +0000)]
[project @ 2003-05-29 13:36:39 by simonmar]
Also clean the Haddock interface file.

20 years ago[project @ 2003-05-29 13:22:51 by simonpj]
simonpj [Thu, 29 May 2003 13:22:52 +0000 (13:22 +0000)]
[project @ 2003-05-29 13:22:51 by simonpj]
n+k patterns should be in Integral, according to the Report

20 years ago[project @ 2003-05-29 13:03:56 by simonmar]
simonmar [Thu, 29 May 2003 13:03:56 +0000 (13:03 +0000)]
[project @ 2003-05-29 13:03:56 by simonmar]
Prune autoconf droppings from the source dist.

20 years ago[project @ 2003-05-29 12:59:55 by simonmar]
simonmar [Thu, 29 May 2003 12:59:55 +0000 (12:59 +0000)]
[project @ 2003-05-29 12:59:55 by simonmar]
Clean AutoApply.hc too.

20 years ago[project @ 2003-05-29 12:58:00 by simonmar]
simonmar [Thu, 29 May 2003 12:58:00 +0000 (12:58 +0000)]
[project @ 2003-05-29 12:58:00 by simonmar]
Clean .depend*, not just .depend (in fptools/ghc/compiler we have
several .depend files).

20 years ago[project @ 2003-05-29 09:03:02 by simonmar]
simonmar [Thu, 29 May 2003 09:03:02 +0000 (09:03 +0000)]
[project @ 2003-05-29 09:03:02 by simonmar]
Note that autoconf 2.52 is required.

20 years ago[project @ 2003-05-29 09:01:20 by simonmar]
simonmar [Thu, 29 May 2003 09:01:20 +0000 (09:01 +0000)]
[project @ 2003-05-29 09:01:20 by simonmar]
Use AC_PREREQ to ensure that we have at least autoconf 2.52, and
remove some old messages about version 2.13.

20 years ago[project @ 2003-05-28 12:10:43 by simonmar]
simonmar [Wed, 28 May 2003 12:10:43 +0000 (12:10 +0000)]
[project @ 2003-05-28 12:10:43 by simonmar]
Surround import Util with #ifdef DEBUG, to avoid the warning (and so I
don't accidentally delete it again).

20 years ago[project @ 2003-05-28 09:43:05 by simonmar]
simonmar [Wed, 28 May 2003 09:43:05 +0000 (09:43 +0000)]
[project @ 2003-05-28 09:43:05 by simonmar]
- Update on compatibility with recent GCC versions
- Mention that Happy might need to be checked out

(from Volker Stolz).

20 years ago[project @ 2003-05-28 09:05:21 by simonmar]
simonmar [Wed, 28 May 2003 09:05:21 +0000 (09:05 +0000)]
[project @ 2003-05-28 09:05:21 by simonmar]
Don't recommend that people do 'cvs update -d' at the top level of an
fptools tree.

20 years ago[project @ 2003-05-28 08:22:22 by simonpj]
simonpj [Wed, 28 May 2003 08:22:22 +0000 (08:22 +0000)]
[project @ 2003-05-28 08:22:22 by simonpj]
Remove duplicate #includes arising from foreign import decls

20 years ago[project @ 2003-05-28 01:12:07 by mthomas]
mthomas [Wed, 28 May 2003 01:12:07 +0000 (01:12 +0000)]
[project @ 2003-05-28 01:12:07 by mthomas]
Assume that notNull in ASSERT is still needed, so revert import.

20 years ago[project @ 2003-05-27 21:14:21 by wolfgang]
wolfgang [Tue, 27 May 2003 21:14:22 +0000 (21:14 +0000)]
[project @ 2003-05-27 21:14:21 by wolfgang]
Fix two bugs in the PowerPC NCG:
1. it generated a 'subfi' (subtract from with immediate) instruction,
which doesn't exist in the PowerPC architecture.
2. didn't correctly handle switch tables (test case cg048.hs).

MERGE TO STABLE

20 years ago[project @ 2003-05-27 14:52:34 by simonmar]
simonmar [Tue, 27 May 2003 14:52:34 +0000 (14:52 +0000)]
[project @ 2003-05-27 14:52:34 by simonmar]
Small updates for 6.x

20 years ago[project @ 2003-05-27 14:15:40 by simonmar]
simonmar [Tue, 27 May 2003 14:15:40 +0000 (14:15 +0000)]
[project @ 2003-05-27 14:15:40 by simonmar]
Use ErrUtils.showPass for printing the '*** Core Linted...' message
instead of rolling our own (also has the effect of sending this to
stderr for consistency).

Tidy up unused imports while I'm here.

20 years ago[project @ 2003-05-27 14:05:11 by simonmar]
simonmar [Tue, 27 May 2003 14:05:11 +0000 (14:05 +0000)]
[project @ 2003-05-27 14:05:11 by simonmar]
Send "Result size =" messages to stderr.

20 years ago[project @ 2003-05-27 14:02:04 by simonmar]
simonmar [Tue, 27 May 2003 14:02:06 +0000 (14:02 +0000)]
[project @ 2003-05-27 14:02:04 by simonmar]
Enable large-file support using AC_SYS_LARGEFILE.

HEADS UP: autoconf 2.5 is now needed to process configure.in.  I
figured this was preferable to copying the AC_SYS_LARGEFILE macro from
the autoconf sources and trying to make it work with autoconf 2.13.

20 years ago[project @ 2003-05-27 12:47:55 by simonpj]
simonpj [Tue, 27 May 2003 12:47:55 +0000 (12:47 +0000)]
[project @ 2003-05-27 12:47:55 by simonpj]
Track changes in Generics.hs

20 years ago[project @ 2003-05-27 12:40:19 by simonmar]
simonmar [Tue, 27 May 2003 12:40:19 +0000 (12:40 +0000)]
[project @ 2003-05-27 12:40:19 by simonmar]
If we're going to surround all arguments to external functions with
quotes (see SysTools.lhs rev. 1.88) then we shouldn't add quotes to
the arguments to -D and -U passed to cpp.

20 years ago[project @ 2003-05-27 11:06:41 by simonpj]
simonpj [Tue, 27 May 2003 11:06:41 +0000 (11:06 +0000)]
[project @ 2003-05-27 11:06:41 by simonpj]
Advise against SplitObjs for Win32

20 years ago[project @ 2003-05-25 20:54:18 by stolz]
stolz [Sun, 25 May 2003 20:54:18 +0000 (20:54 +0000)]
[project @ 2003-05-25 20:54:18 by stolz]
Fix quoted source filenames on *nix

Noticed by: Calle Lejdfors
(please merge)