ghc-hetmet.git
22 years ago[project @ 2001-07-16 10:29:42 by rrt]
rrt [Mon, 16 Jul 2001 10:29:42 +0000 (10:29 +0000)]
[project @ 2001-07-16 10:29:42 by rrt]
Binaries now live in extra-bin

22 years ago[project @ 2001-07-16 09:49:31 by rrt]
rrt [Mon, 16 Jul 2001 09:49:31 +0000 (09:49 +0000)]
[project @ 2001-07-16 09:49:31 by rrt]
Fix binary finding for Unix

22 years ago[project @ 2001-07-16 09:41:26 by simonpj]
simonpj [Mon, 16 Jul 2001 09:41:26 +0000 (09:41 +0000)]
[project @ 2001-07-16 09:41:26 by simonpj]
Tidy up Type/TcType stuff in DsCCall/DsForeign

22 years ago[project @ 2001-07-16 09:30:30 by rrt]
rrt [Mon, 16 Jul 2001 09:30:30 +0000 (09:30 +0000)]
[project @ 2001-07-16 09:30:30 by rrt]
Set -DDLLized when DLLized is true for use in C source

22 years ago[project @ 2001-07-16 09:23:24 by rrt]
rrt [Mon, 16 Jul 2001 09:23:24 +0000 (09:23 +0000)]
[project @ 2001-07-16 09:23:24 by rrt]
Add stubs for ILXized cbits DLL.

Don't call resetNonBlockingFd when DLLized; this stops the cbits DLL
depending on the RTS, which ILX needs; it doesn't matter for Windows, as
that doesn't have non-blocking IO implemented anyway.

22 years ago[project @ 2001-07-16 09:13:39 by rrt]
rrt [Mon, 16 Jul 2001 09:13:39 +0000 (09:13 +0000)]
[project @ 2001-07-16 09:13:39 by rrt]
Fix DLL building, and add support for linking .hsc.o files into cbits libraries

22 years ago[project @ 2001-07-16 09:11:44 by qrczak]
qrczak [Mon, 16 Jul 2001 09:11:44 +0000 (09:11 +0000)]
[project @ 2001-07-16 09:11:44 by qrczak]
s/extra-bin/bin/ because this is where all programs are installed to.
Perhaps it should be changed the other way around - I don't know what
this extra-bin was supposed to be.

22 years ago[project @ 2001-07-16 06:33:22 by qrczak]
qrczak [Mon, 16 Jul 2001 06:33:22 +0000 (06:33 +0000)]
[project @ 2001-07-16 06:33:22 by qrczak]
Fix paths in installed hsc2hs driver.

22 years ago[project @ 2001-07-16 00:43:56 by sof]
sof [Mon, 16 Jul 2001 00:43:56 +0000 (00:43 +0000)]
[project @ 2001-07-16 00:43:56 by sof]
hscStmt: track changes made to the TC entry points typecheckExpr, typecheckStmt

22 years ago[project @ 2001-07-16 00:39:04 by sof]
sof [Mon, 16 Jul 2001 00:39:04 +0000 (00:39 +0000)]
[project @ 2001-07-16 00:39:04 by sof]
eqForeignPtr: implement equality using new primitive

22 years ago[project @ 2001-07-14 00:06:13 by sof]
sof [Sat, 14 Jul 2001 00:06:15 +0000 (00:06 +0000)]
[project @ 2001-07-14 00:06:13 by sof]
Heal HEID

- eqForeignObjZh in include/PrimOps.h didn't have quite the right
  shape (the result is a macro arg). hslibs/lang/ForeignObj
  wasn't up on the change to eqForeignObj now being a primop.

- recent ghc/compiler/deSugar/ changes broke the handling of
  CCall & FFI decls quite a bit. Backed out most the rewrites
  of Type.splitFoo to TcType.tcSplitFoo (i.e., now back to using
  TcType.tcSplitFoo).

  The backed-out newtype-related changes were by no means accidental.
  But, I don't profess to understand their intention to make the
  proper fix, so my change is just a stop-gap measure to get HEAD
  back to the land of the living.

- recent changes to the behaviour of 'hiding' & qualified names
  broke hslibs/lang/CString hslibs/data/edison/Seq/ListSeq,
  hslibs/data/edison/Coll/TestOrdBag, hslibs/data/edison/Coll/UnbalancedSet,
  hslibs/data/edison/Coll/TestOrdSet, hslibs/data/edison/Seq/TestSeq

- rename 64-bit 'primop' funs that now live in lib/std/cbits/longlong.c
  back to what they used to be called (i.e., prefixed with "stg_").
  Why?
     - less likely they'll clash with other (user supplied) entry points
       at link-time.
     - matches protos in ghc/includes/PrimOp.h

22 years ago[project @ 2001-07-13 15:03:00 by simonmar]
simonmar [Fri, 13 Jul 2001 15:03:00 +0000 (15:03 +0000)]
[project @ 2001-07-13 15:03:00 by simonmar]
add warning about not changing WhatGccIsCalled from build.mk.

22 years ago[project @ 2001-07-13 15:01:28 by simonmar]
simonmar [Fri, 13 Jul 2001 15:01:28 +0000 (15:01 +0000)]
[project @ 2001-07-13 15:01:28 by simonmar]
remove superfluous "return ()"

22 years ago[project @ 2001-07-13 15:00:11 by simonmar]
simonmar [Fri, 13 Jul 2001 15:00:11 +0000 (15:00 +0000)]
[project @ 2001-07-13 15:00:11 by simonmar]
error msg wibble

22 years ago[project @ 2001-07-13 14:59:32 by simonmar]
simonmar [Fri, 13 Jul 2001 14:59:32 +0000 (14:59 +0000)]
[project @ 2001-07-13 14:59:32 by simonmar]
check TMPDIR (got lost somewhere in the conversion to SysTools)

22 years ago[project @ 2001-07-13 13:55:19 by rrt]
rrt [Fri, 13 Jul 2001 13:55:19 +0000 (13:55 +0000)]
[project @ 2001-07-13 13:55:19 by rrt]
Don't build RTS for ILX

22 years ago[project @ 2001-07-13 13:41:42 by rrt]
rrt [Fri, 13 Jul 2001 13:41:42 +0000 (13:41 +0000)]
[project @ 2001-07-13 13:41:42 by rrt]
Improve comment

22 years ago[project @ 2001-07-13 13:40:48 by rrt]
rrt [Fri, 13 Jul 2001 13:40:48 +0000 (13:40 +0000)]
[project @ 2001-07-13 13:40:48 by rrt]
Function stubs for ILX

22 years ago[project @ 2001-07-13 13:40:26 by rrt]
rrt [Fri, 13 Jul 2001 13:40:26 +0000 (13:40 +0000)]
[project @ 2001-07-13 13:40:26 by rrt]
Add StgPrimFloat.c for ILX

22 years ago[project @ 2001-07-13 13:29:56 by simonpj]
simonpj [Fri, 13 Jul 2001 13:29:58 +0000 (13:29 +0000)]
[project @ 2001-07-13 13:29:56 by simonpj]
------------------------------------
Tidy up the "syntax rebinding" story
------------------------------------

I found a bug in the code that dealt with re-binding implicit
numerical syntax:
literals  (fromInteger/fromRational)
negation (negate)
n+k patterns (minus)

This is triggered by the -fno-implicit-prelude flag, and it
used to be handled via the PrelNames.SyntaxMap.

But I found a nicer way to do it that involves much less code,
and doesn't have the bug.  The explanation is with
RnEnv.lookupSyntaxName

22 years ago[project @ 2001-07-13 11:50:46 by rrt]
rrt [Fri, 13 Jul 2001 11:50:46 +0000 (11:50 +0000)]
[project @ 2001-07-13 11:50:46 by rrt]
Need to add -Icbits even for ILX

22 years ago[project @ 2001-07-13 11:48:52 by rrt]
rrt [Fri, 13 Jul 2001 11:48:52 +0000 (11:48 +0000)]
[project @ 2001-07-13 11:48:52 by rrt]
Don't use relative #include paths, as these files will be compiled from
inside cbits (they're now linked into the cbits library). All paths on which
include files may be found must now be given with -I directives.

22 years ago[project @ 2001-07-13 11:39:48 by rrt]
rrt [Fri, 13 Jul 2001 11:39:48 +0000 (11:39 +0000)]
[project @ 2001-07-13 11:39:48 by rrt]
Make eqForeignObj a primop; N.B. Not implemented for the NCG

22 years ago[project @ 2001-07-13 11:11:34 by rrt]
rrt [Fri, 13 Jul 2001 11:11:34 +0000 (11:11 +0000)]
[project @ 2001-07-13 11:11:34 by rrt]
Explain why errno is imported via a function

22 years ago[project @ 2001-07-13 11:03:47 by rrt]
rrt [Fri, 13 Jul 2001 11:03:47 +0000 (11:03 +0000)]
[project @ 2001-07-13 11:03:47 by rrt]
Move 64-bit arithmetic support into std_cbits; there's no reason for it to
be in the RTS.

22 years ago[project @ 2001-07-12 16:25:32 by sof]
sof [Thu, 12 Jul 2001 16:25:32 +0000 (16:25 +0000)]
[project @ 2001-07-12 16:25:32 by sof]
ppFilesFromSummaries: to cope with recent driver changes,
only report a file as being preprocessed if it differs from
the source file(name).

Fixes awfully bad breakage when using --make (source files
were being deleted).

22 years ago[project @ 2001-07-12 16:24:49 by simonpj]
simonpj [Thu, 12 Jul 2001 16:24:49 +0000 (16:24 +0000)]
[project @ 2001-07-12 16:24:49 by simonpj]
Remove DsHsSyn; really part of my previous commit

22 years ago[project @ 2001-07-12 16:21:22 by simonpj]
simonpj [Thu, 12 Jul 2001 16:21:24 +0000 (16:21 +0000)]
[project @ 2001-07-12 16:21:22 by simonpj]
--------------------------------------------
Fix another bug in the squash-newtypes story.
--------------------------------------------

[This one was spotted by Marcin, and is now enshrined in test tc130.]

The desugarer straddles the boundary between the type checker and
Core, so it sometimes needs to look through newtypes/implicit parameters
and sometimes not.  This is really a bit painful, but I can't think of
a better way to do it.

The only simple way to fix things was to pass a bit more type
information in the HsExpr type, from the type checker to the desugarer.
That led to the non-local changes you can see.

On the way I fixed one other thing.  In various HsSyn constructors
there is a Type that is bogus (bottom) before the type checker, and
filled in with a real type by the type checker.  In one place it was
a (Maybe Type) which was Nothing before, and (Just ty) afterwards.
I've defined a type synonym HsTypes.PostTcType for this, and a named
bottom value HsTypes.placeHolderType to use when you want the bottom
value.

22 years ago[project @ 2001-07-12 14:51:28 by simonpj]
simonpj [Thu, 12 Jul 2001 14:51:28 +0000 (14:51 +0000)]
[project @ 2001-07-12 14:51:28 by simonpj]
Fix the module import story to match what the Revised
Haskell Report says

1.  Don't import qualified names of things that aren't imported

2. Fix a bug that meant
import A hiding( D )
where D is a data constructor, didn't work.
[The fix is to use IEVar not IEThingAbs in the
want_hiding case of get_item in RnNames.filterImports

22 years ago[project @ 2001-07-12 13:32:25 by rrt]
rrt [Thu, 12 Jul 2001 13:32:25 +0000 (13:32 +0000)]
[project @ 2001-07-12 13:32:25 by rrt]
Rewire so that hsc.c files get built as part of cbits library.

22 years ago[project @ 2001-07-12 12:59:48 by rrt]
rrt [Thu, 12 Jul 2001 12:59:48 +0000 (12:59 +0000)]
[project @ 2001-07-12 12:59:48 by rrt]
Fix spacing

22 years ago[project @ 2001-07-12 12:58:39 by rrt]
rrt [Thu, 12 Jul 2001 12:58:39 +0000 (12:58 +0000)]
[project @ 2001-07-12 12:58:39 by rrt]
Two fixes:

1. Don't emit "native" any more.

2. Add import of mscorlib to the start of every output file.

22 years ago[project @ 2001-07-12 10:37:55 by rrt]
rrt [Thu, 12 Jul 2001 10:37:55 +0000 (10:37 +0000)]
[project @ 2001-07-12 10:37:55 by rrt]
Change access to errorHdrHook back into a foreign label (this is now
supported by the NCG, or should be, and in any case, we don't build
HSstd with the NCG).

22 years ago[project @ 2001-07-12 10:34:54 by rrt]
rrt [Thu, 12 Jul 2001 10:34:54 +0000 (10:34 +0000)]
[project @ 2001-07-12 10:34:54 by rrt]
Remove addrOf_ErrorHdrHook, as the NCG now can (or should) cope with
foreign labels, and in any case, we don't use the NCG to build the
standard library.

The motivation for this fix is preventing HSstd_cbits depending on the
RTS, so that it can be built as a DLL (for .NET) without needing to
get the RTS built as a DLL (that way lies madness).

22 years ago[project @ 2001-07-11 22:47:32 by sof]
sof [Wed, 11 Jul 2001 22:47:32 +0000 (22:47 +0000)]
[project @ 2001-07-11 22:47:32 by sof]
Append final newline to _stub.{c,h} output

22 years ago[project @ 2001-07-11 19:48:07 by sof]
sof [Wed, 11 Jul 2001 19:48:07 +0000 (19:48 +0000)]
[project @ 2001-07-11 19:48:07 by sof]
Prev commit which tried to get rid of the 'ineffective CPP'
stage broke an invariant/assumption made by DriverPipeline.genPipeline,
I'm afraid.

The invariant being that from the input filename to a compilation
pipeline, the start stage/phase can be uniquely determined
(via DriverPhases.startPhase). run_phase no longer guarantees
this, its result filename may now be equal to the input filename.
[This resulted in -M not working properly when -cpp wasn't also used].

Patch this up by having the input filename to a compilation pipeline
stage be tagged with the assumed suffix of the output. This is
really just papering over the cracks, the proper way to tidy this
up is to avoid having to do the two-stage compilation pipeline in
Main.main.compileFile, i.e., reconsider how OPTIONS pragmas are
handled. I decided not to try to do this now for fear of destabilising
HEAD even further.

22 years ago[project @ 2001-07-11 14:51:49 by sof]
sof [Wed, 11 Jul 2001 14:51:49 +0000 (14:51 +0000)]
[project @ 2001-07-11 14:51:49 by sof]
runSomething: don't export it

22 years ago[project @ 2001-07-11 14:50:49 by sof]
sof [Wed, 11 Jul 2001 14:50:49 +0000 (14:50 +0000)]
[project @ 2001-07-11 14:50:49 by sof]
run_phase: added tysig

22 years ago[project @ 2001-07-11 13:24:24 by rrt]
rrt [Wed, 11 Jul 2001 13:24:25 +0000 (13:24 +0000)]
[project @ 2001-07-11 13:24:24 by rrt]
Replace GhcLibToolsHcOpts with FptoolsHcOpts, and make it apply to all
programs in fptools, not just the hslibs tools. Now there's a way of
controlling compiler flags to all programs in the tree, so

a) by default all utilities should be optimised (before this was only done
for one or two like hsc2hs)

b) if you're doing development work, you can get everything built quicker by
setting FptoolsHcOpts appropriately

22 years ago[project @ 2001-07-11 13:10:31 by rrt]
rrt [Wed, 11 Jul 2001 13:10:31 +0000 (13:10 +0000)]
[project @ 2001-07-11 13:10:31 by rrt]
Don't add -O here; should be up to the user

22 years ago[project @ 2001-07-11 11:01:59 by rrt]
rrt [Wed, 11 Jul 2001 11:01:59 +0000 (11:01 +0000)]
[project @ 2001-07-11 11:01:59 by rrt]
Make ghc-pkg find its path on Windows

22 years ago[project @ 2001-07-11 10:20:43 by rrt]
rrt [Wed, 11 Jul 2001 10:20:43 +0000 (10:20 +0000)]
[project @ 2001-07-11 10:20:43 by rrt]
Fix spacing bogon.

22 years ago[project @ 2001-07-11 10:17:51 by rrt]
rrt [Wed, 11 Jul 2001 10:17:51 +0000 (10:17 +0000)]
[project @ 2001-07-11 10:17:51 by rrt]
Rewrite getExecDir to use GetModuleFileName on Windows (thanks, sof),
and switch unDosifyPath's definitions between Windows and Unix so that
it actually does something on Windows.

22 years ago[project @ 2001-07-11 01:27:04 by sof]
sof [Wed, 11 Jul 2001 01:27:04 +0000 (01:27 +0000)]
[project @ 2001-07-11 01:27:04 by sof]
Do away with that 'ineffective CPP' malarkey. run_phase now returns
the FilePath of its output file; in the event CPP isn't required,
we then just return the input file. (The extra LINE pragma that
the 'ineffective CPP' inserted was a NOP).

Appears to speed things up a bit; mingw/Win32 builds no longer need to
do a slow copyFile operation - other plats are saved from doing a
System.system call.

BTW, there's an opportunity to shorten the critical path for .hs files
some in the driver: at the moment 'unlit' is run over all input.

22 years ago[project @ 2001-07-11 00:13:22 by sof]
sof [Wed, 11 Jul 2001 00:13:22 +0000 (00:13 +0000)]
[project @ 2001-07-11 00:13:22 by sof]
Avoid needless redefn of HAVE_ALLOCA

22 years ago[project @ 2001-07-10 14:25:24 by rrt]
rrt [Tue, 10 Jul 2001 14:25:24 +0000 (14:25 +0000)]
[project @ 2001-07-10 14:25:24 by rrt]
Add win32 package for Windows ghc.

22 years ago[project @ 2001-07-10 14:23:36 by rrt]
rrt [Tue, 10 Jul 2001 14:23:36 +0000 (14:23 +0000)]
[project @ 2001-07-10 14:23:36 by rrt]
Rewrite getExecDir with Win32Registry calls. Duh.

22 years ago[project @ 2001-07-10 11:32:28 by simonpj]
simonpj [Tue, 10 Jul 2001 11:32:28 +0000 (11:32 +0000)]
[project @ 2001-07-10 11:32:28 by simonpj]
Two bug-fixes to the new newtype story

1.  Be consistent about using TcType (not Type) in the
typechecker.  There was an odd function in TcMType that
used splitTyConApp instead of tcSplitTyConApp, which
resulted in bogus error messages

2. TcType.isTauTy should not look through SourceTy

22 years ago[project @ 2001-07-10 11:23:43 by rrt]
rrt [Tue, 10 Jul 2001 11:23:43 +0000 (11:23 +0000)]
[project @ 2001-07-10 11:23:43 by rrt]
Make getExecDir work on Windows by reading the registry

22 years ago[project @ 2001-07-10 11:22:15 by rrt]
rrt [Tue, 10 Jul 2001 11:22:15 +0000 (11:22 +0000)]
[project @ 2001-07-10 11:22:15 by rrt]
ie -> i.e.

22 years ago[project @ 2001-07-10 11:21:05 by rrt]
rrt [Tue, 10 Jul 2001 11:21:05 +0000 (11:21 +0000)]
[project @ 2001-07-10 11:21:05 by rrt]
Fix jade detection on Windows for jades that don't understand Unix paths

22 years ago[project @ 2001-07-09 19:45:16 by sof]
sof [Mon, 9 Jul 2001 19:45:16 +0000 (19:45 +0000)]
[project @ 2001-07-09 19:45:16 by sof]
Hmm..are the Cambridge offices running low on oxygen? Desloppified to make stage2 work again

22 years ago[project @ 2001-07-09 19:09:39 by sof]
sof [Mon, 9 Jul 2001 19:09:39 +0000 (19:09 +0000)]
[project @ 2001-07-09 19:09:39 by sof]
Undo prev. commit, it was just plain wrong

22 years ago[project @ 2001-07-09 17:44:08 by sof]
sof [Mon, 9 Jul 2001 17:44:08 +0000 (17:44 +0000)]
[project @ 2001-07-09 17:44:08 by sof]
-- 'hybrid' support for DOS-style paths in directory lists.
--
-- That is, if "foo:bar:baz" is used, this interpreted as
-- consisting of three entries, 'foo', 'bar', 'baz'.
-- However, with "c:/foo:c:\\foo;x:/bar", this is interpreted
-- as four elts, "c:/foo", "c:\\foo", "x", and "/bar" --
-- *provided* c:/foo exists and x:/bar doesn't.
--
-- Notice that no attempt is made to fully replace the 'standard'
-- split marker ':' with the Windows / DOS one, ';'. The reason being
-- that this will cause too much breakage for users & ':' will
-- work fine even with DOS paths, if you're not insisting on being silly.
-- So, use either.

22 years ago[project @ 2001-07-09 16:30:50 by sof]
sof [Mon, 9 Jul 2001 16:30:50 +0000 (16:30 +0000)]
[project @ 2001-07-09 16:30:50 by sof]
get rid off gcc warning

22 years ago[project @ 2001-07-09 13:21:09 by sewardj]
sewardj [Mon, 9 Jul 2001 13:21:09 +0000 (13:21 +0000)]
[project @ 2001-07-09 13:21:09 by sewardj]
Zap another ghc_411_at_least.

22 years ago[project @ 2001-07-09 10:46:46 by simonmar]
simonmar [Mon, 9 Jul 2001 10:46:46 +0000 (10:46 +0000)]
[project @ 2001-07-09 10:46:46 by simonmar]
Change the test for whether to include GHCi support in the compiler to
be closer to "are we bootstrapping with the same compiler".

22 years ago[project @ 2001-07-09 08:24:08 by simonmar]
simonmar [Mon, 9 Jul 2001 08:24:08 +0000 (08:24 +0000)]
[project @ 2001-07-09 08:24:08 by simonmar]
oops: I changed the names of some of the GC stubs, and didn't realise they
were mentioned here too.

22 years ago[project @ 2001-07-08 17:04:04 by sof]
sof [Sun, 8 Jul 2001 17:04:04 +0000 (17:04 +0000)]
[project @ 2001-07-08 17:04:04 by sof]
getTimes: fix Win95 fallback

22 years ago[project @ 2001-07-06 15:39:47 by rrt]
rrt [Fri, 6 Jul 2001 15:39:47 +0000 (15:39 +0000)]
[project @ 2001-07-06 15:39:47 by rrt]
Fix path for Windows jade

22 years ago[project @ 2001-07-06 14:18:53 by simonmar]
simonmar [Fri, 6 Jul 2001 14:18:53 +0000 (14:18 +0000)]
[project @ 2001-07-06 14:18:53 by simonmar]
Add test for broken takeMVar/putMVar in 5.00.2.

Bug reported by Greg Mildenhall <gregm@pc-121.cs.uwa.edu.au> on
comp.lang.functional..

22 years ago[project @ 2001-07-06 14:15:50 by rrt]
rrt [Fri, 6 Jul 2001 14:15:50 +0000 (14:15 +0000)]
[project @ 2001-07-06 14:15:50 by rrt]
lib/gcc-lib -> gcc-lib

22 years ago[project @ 2001-07-06 14:11:38 by simonmar]
simonmar [Fri, 6 Jul 2001 14:11:38 +0000 (14:11 +0000)]
[project @ 2001-07-06 14:11:38 by simonmar]
Fix a couple of nasty bugs in the take/putMVar implementation.

Now we keep the invariant that a full MVar only has blocked putMVars
on its queue, and an empty MVar only has blocked takeMVars on its
queue.  It was the absence of this invariant that led to accidental
deadlock before.

The second bug is that there was a window between a blocked thread
being restarted and it actually retrying the takeMVar/putMVar
operation when it could receive an exception, which would also lead to
deadlock.

The solution to both these problems (as suggested by Simon P.J.) is to
atomically wake up and perform the next blocked putMVar when we do a
take, and vice versa.  As a side effect, takeMVar & putMVar should be
much faster when blocking & restarting, because we now shortcut the
retrying of the blocked operation and we use a more specialised stack
layout for the blocked thread.  Unfortunately, things got more
complicated too, but there are comments explaining what's going on.

22 years ago[project @ 2001-07-06 14:05:14 by apt]
apt [Fri, 6 Jul 2001 14:05:14 +0000 (14:05 +0000)]
[project @ 2001-07-06 14:05:14 by apt]
fix eqType bug

22 years ago[project @ 2001-07-06 10:53:04 by rrt]
rrt [Fri, 6 Jul 2001 10:53:04 +0000 (10:53 +0000)]
[project @ 2001-07-06 10:53:04 by rrt]
New version for GHC 5

22 years ago[project @ 2001-07-05 13:52:40 by rrt]
rrt [Thu, 5 Jul 2001 13:52:40 +0000 (13:52 +0000)]
[project @ 2001-07-05 13:52:40 by rrt]
Final stage of the PACKAGE oopsathon, or else...

22 years ago[project @ 2001-07-05 13:31:09 by sewardj]
sewardj [Thu, 5 Jul 2001 13:31:10 +0000 (13:31 +0000)]
[project @ 2001-07-05 13:31:09 by sewardj]
Changes to nuke RAWCPP.

22 years ago[project @ 2001-07-05 13:01:14 by rrt]
rrt [Thu, 5 Jul 2001 13:01:16 +0000 (13:01 +0000)]
[project @ 2001-07-05 13:01:14 by rrt]
Change HSLIB to PACKAGE and HSLIB_DEPS to PACKAGE_DEPS, which makes more
sense. Also unbreaks the build system, in an arse-about-face sort of way
(or, if you prefer, cart-before-horse).

22 years ago[project @ 2001-07-05 08:55:29 by simonpj]
simonpj [Thu, 5 Jul 2001 08:55:29 +0000 (08:55 +0000)]
[project @ 2001-07-05 08:55:29 by simonpj]
Correct spelling in error message

22 years ago[project @ 2001-07-04 16:55:48 by rrt]
rrt [Wed, 4 Jul 2001 16:55:48 +0000 (16:55 +0000)]
[project @ 2001-07-04 16:55:48 by rrt]
Only use -package-name when in a package

22 years ago[project @ 2001-07-04 16:29:35 by rrt]
rrt [Wed, 4 Jul 2001 16:29:36 +0000 (16:29 +0000)]
[project @ 2001-07-04 16:29:35 by rrt]
Move -package-name and -package flag setting up a level

22 years ago[project @ 2001-07-04 15:52:24 by simonpj]
simonpj [Wed, 4 Jul 2001 15:52:24 +0000 (15:52 +0000)]
[project @ 2001-07-04 15:52:24 by simonpj]
Add notes

22 years ago[project @ 2001-07-04 15:51:23 by simonpj]
simonpj [Wed, 4 Jul 2001 15:51:23 +0000 (15:51 +0000)]
[project @ 2001-07-04 15:51:23 by simonpj]
Add comments; and copy system.c on demand rather than via make boot

22 years ago[project @ 2001-07-04 15:43:38 by simonmar]
simonmar [Wed, 4 Jul 2001 15:43:38 +0000 (15:43 +0000)]
[project @ 2001-07-04 15:43:38 by simonmar]
Only print out the GHCi banner when verbosity >= 1, so you can omit
the banner with -v0 (useful for tests where you don't want the GHC
version number appearing in the output).

22 years ago[project @ 2001-07-04 11:46:37 by sewardj]
sewardj [Wed, 4 Jul 2001 11:46:37 +0000 (11:46 +0000)]
[project @ 2001-07-04 11:46:37 by sewardj]
Scummy hacks to make GMP configure ok for mingw.
-- Copy files instead of symlinking them.
-- Allow correct detection of the ld used with gcc, so that
   that ar rather than the Windows LIB program is used.

22 years ago[project @ 2001-07-04 11:06:02 by simonmar]
simonmar [Wed, 4 Jul 2001 11:06:02 +0000 (11:06 +0000)]
[project @ 2001-07-04 11:06:02 by simonmar]
Allow .hi-boot files to be demand-loaded even in
interactive/make mode, and fix a wibble.

22 years ago[project @ 2001-07-04 10:58:37 by rrt]
rrt [Wed, 4 Jul 2001 10:58:37 +0000 (10:58 +0000)]
[project @ 2001-07-04 10:58:37 by rrt]
gcc-2.95.3-4 for mingw accepts -mwin32, so allow it

22 years ago[project @ 2001-07-04 09:18:38 by sewardj]
sewardj [Wed, 4 Jul 2001 09:18:38 +0000 (09:18 +0000)]
[project @ 2001-07-04 09:18:38 by sewardj]
Un-break on Unix.

22 years ago[project @ 2001-07-03 17:07:52 by rrt]
rrt [Tue, 3 Jul 2001 17:07:52 +0000 (17:07 +0000)]
[project @ 2001-07-03 17:07:52 by rrt]
Kill --enable-minimal-unix-deps (now == mingw32) and emit an error if a
Cygwin compiler is detected.

22 years ago[project @ 2001-07-03 16:57:03 by sewardj]
sewardj [Tue, 3 Jul 2001 16:57:03 +0000 (16:57 +0000)]
[project @ 2001-07-03 16:57:03 by sewardj]
Plumb top-level --with-gcc=... value around the place.

22 years ago[project @ 2001-07-03 16:52:45 by rrt]
rrt [Tue, 3 Jul 2001 16:52:45 +0000 (16:52 +0000)]
[project @ 2001-07-03 16:52:45 by rrt]
Allow child process to inherit parent's handles.

22 years ago[project @ 2001-07-03 16:46:21 by rrt]
rrt [Tue, 3 Jul 2001 16:46:21 +0000 (16:46 +0000)]
[project @ 2001-07-03 16:46:21 by rrt]
Get rid of MinimalUnixDeps (now == mingwin)

22 years ago[project @ 2001-07-03 16:45:15 by rrt]
rrt [Tue, 3 Jul 2001 16:45:15 +0000 (16:45 +0000)]
[project @ 2001-07-03 16:45:15 by rrt]
Quote -D args on Unix

22 years ago[project @ 2001-07-03 16:26:03 by rrt]
rrt [Tue, 3 Jul 2001 16:26:03 +0000 (16:26 +0000)]
[project @ 2001-07-03 16:26:03 by rrt]
Clarify and refine, for a smoother texture

22 years ago[project @ 2001-07-03 14:53:35 by simonpj]
simonpj [Tue, 3 Jul 2001 14:53:35 +0000 (14:53 +0000)]
[project @ 2001-07-03 14:53:35 by simonpj]
Dump package info only for -v3

22 years ago[project @ 2001-07-03 11:14:33 by simonmar]
simonmar [Tue, 3 Jul 2001 11:14:33 +0000 (11:14 +0000)]
[project @ 2001-07-03 11:14:33 by simonmar]
Better exit with a failure code in ghc --make if something went wrong...

22 years ago[project @ 2001-07-02 18:05:10 by rrt]
rrt [Mon, 2 Jul 2001 18:05:10 +0000 (18:05 +0000)]
[project @ 2001-07-02 18:05:10 by rrt]
Print out the long long GC_tot_alloc separately to avoid printf bogons
on mingwin.

22 years ago[project @ 2001-07-02 13:52:10 by rrt]
rrt [Mon, 2 Jul 2001 13:52:10 +0000 (13:52 +0000)]
[project @ 2001-07-02 13:52:10 by rrt]
Fix up properly not to use _WIN32.

22 years ago[project @ 2001-07-02 13:08:49 by rrt]
rrt [Mon, 2 Jul 2001 13:08:49 +0000 (13:08 +0000)]
[project @ 2001-07-02 13:08:49 by rrt]
Don't use _WIN32; use mingw32_TARGET_OS and cygwin32_TARGET_OS
instead.

22 years ago[project @ 2001-07-02 13:02:29 by rrt]
rrt [Mon, 2 Jul 2001 13:02:29 +0000 (13:02 +0000)]
[project @ 2001-07-02 13:02:29 by rrt]
Don't use _WIN32, use mingw32_TARGET_OS and cygwin32_TARGET_OS.

22 years ago[project @ 2001-07-02 12:00:44 by simonmar]
simonmar [Mon, 2 Jul 2001 12:00:44 +0000 (12:00 +0000)]
[project @ 2001-07-02 12:00:44 by simonmar]
oops, cut-n-paste removed tabs

22 years ago[project @ 2001-07-02 10:08:52 by simonmar]
simonmar [Mon, 2 Jul 2001 10:08:52 +0000 (10:08 +0000)]
[project @ 2001-07-02 10:08:52 by simonmar]
The hc-file bundle should contain the .hs files generated from .hsc
files because of a chicken-egg problem with building hsc2hs.

22 years ago[project @ 2001-06-30 22:57:20 by qrczak]
qrczak [Sat, 30 Jun 2001 22:57:20 +0000 (22:57 +0000)]
[project @ 2001-06-30 22:57:20 by qrczak]
Don't add leading space to libdir (was harmful for drivers).

Don't add $(ProjectNameShort)-$(ProjectVersion) to libdir if
$(ProjectNameShort) is empty; I hope it's correct.

22 years ago[project @ 2001-06-30 11:21:12 by simonmar]
simonmar [Sat, 30 Jun 2001 11:21:12 +0000 (11:21 +0000)]
[project @ 2001-06-30 11:21:12 by simonmar]
sigh, the split script broke due to changes in gcc on FreeBSD (again).

22 years ago[project @ 2001-06-30 00:12:52 by sof]
sof [Sat, 30 Jun 2001 00:12:52 +0000 (00:12 +0000)]
[project @ 2001-06-30 00:12:52 by sof]
FPTOOLS_CC_FLAG: Save and restore CFLAGS

22 years ago[project @ 2001-06-30 00:00:49 by sof]
sof [Sat, 30 Jun 2001 00:00:49 +0000 (00:00 +0000)]
[project @ 2001-06-30 00:00:49 by sof]
Tidy up the handling of -mno-cygwin.

22 years ago[project @ 2001-06-29 16:58:06 by sewardj]
sewardj [Fri, 29 Jun 2001 16:58:06 +0000 (16:58 +0000)]
[project @ 2001-06-29 16:58:06 by sewardj]
Test for mingw32_TARGET_OS and cygwin32_TARGET_OS instead of WIN32.

22 years ago[project @ 2001-06-29 16:17:28 by sewardj]
sewardj [Fri, 29 Jun 2001 16:17:28 +0000 (16:17 +0000)]
[project @ 2001-06-29 16:17:28 by sewardj]
Let the PEi386 machinery recognise .rodata sections.

22 years ago[project @ 2001-06-29 15:37:52 by simonmar]
simonmar [Fri, 29 Jun 2001 15:37:52 +0000 (15:37 +0000)]
[project @ 2001-06-29 15:37:52 by simonmar]
The "Warning: deleting non-existent foo" message should only be
printed in verbose mode.

22 years ago[project @ 2001-06-29 15:30:49 by simonmar]
simonmar [Fri, 29 Jun 2001 15:30:49 +0000 (15:30 +0000)]
[project @ 2001-06-29 15:30:49 by simonmar]
reformat & add a comment