Make typeVarsIn non-static since interface.c needs to see it.

Change CFunDescriptor to match new foreign import implementation.

Reimplement back-end for foreign import (calling out).

Return to a cleaned-up version of Alastair's callfun.S,
wherein an architecture and calling-convention specific
piece of assembly code is used to construct arguments and
then call the specified function, under the direction of
a type descriptor string.

Defined an interface to this function (universal_call_c)
which I hope will work regardless of 32-or-64 bitness,
endianness and calling convention.

Current implementation is for x86-linux only.

* Time.CalendarTime.ctMonth's type should be Month (was Int.)
* fixed Time.addToClockTime - the original implementation was
  completely wrong (thanks to George Russell for indirectly
  reporting the bug.)
* Added the non-std Time.noTimeDiff, handy when you want to do
  calendar calculations, e.g.,

            ct <- getClockTime
            print (toUTCTime (addToClockTime noTimeDiff{tdMonth=1} ct))

* many 'foreign import' decls were not marked as unsafe, even though they
  were passing out MutableByteArray and ByteArray vals. Fixed.

The fix to ignore error() cases when doing CPR analysis exposed a
problem with the Void type.  A function that always constructs a void
result was converted to w/w, but the worker was producing an unboxed
tuple with 0 components.  Not good.

Fixed so that constructing a void gives CPR value Top.  This is OK
because we won't really be constructing a void each time, we will be
returning a pointer to a shared void cell.

Adding diffs between Hugs98 (Jan99) and Hugs98 (Sep99)
manually to STG Hugs.

Brings in large change to typechecking sub-system.

Adding diffs between Hugs98 (Jan99) and Hugs98 (Sep99)
manually to STG Hugs.

These are the changes to input.c, with minor tweeks to
connect.h and parser.y to make this work.

Adding diffs between Hugs98 (Jan99) and Hugs98 (Sep99) manually to STG Hugs.

Updating all copyright messages to the same as Hugs98.

Changing order of check for use of DLL for dynamic linking.
(Under cygwin, we still want to use DLL's).

Added basic support for foreign export dynamic.

Many aspects of it are still broken:
* Only supports x86-linux.
* The range of allowable types is small: Char Int Float Double
  Addr and Word.
* Adjustor thunks are never freed.
* Returning Doubles or Floats doesn't work at all.

I expect to fix some of these shortly.  foreign import also
needs redoing, so it can accept any number of arguments of
any type.

Also:
* Fixed setRtsFlags in Evaluator.c to make it endian-independent.
* Fixed raisePrim in Evaluator.c so things like division by zero,
  array index errors, etc, throw an exception instead of
  terminating StgHugs.  raisePrim is renamed makeErrorCall.

Add macros for two-register call/return convention, for experimentation.

set $(HC) to the inplace ghc driver script.

Swap ticky column order

Allow unboxed tuples with zero components in unfoldings, CPR seems to
generate them occasionally (perhaps it shouldn't - there is very
little difference between returning () and (# #)).

oops: unbreak non-ticky compiles

Add a test for scoped type variables

Crude allocation-counting extension to ticky-ticky profiling.

Allocations are counted against the closest lexically enclosing
function closure, so you need to map the output back to the STG code.

typo

#ifdefs for bootstrapping

add a bunch of #ifdefs so we can bootstrap again

Don't use variables beginning with underscore, 3.02 doesn't grok them.

typo

install (INSTALL_PROGS): a ; too many

Disable object loading until Hugs/GHC class/instance sync is done.

Minor library handling mods to make nofib work again.

Try and make finding of Prelude libraries for hugs a bit saner.
Guess (badly) the absolute pathname to the hugs executable, and
use that to derive the location of the default library.

use $(GHC), not $(HC)

use $(GHC), not $(HC)

use $(GHC), not $(HC)

Added the {index,read,write}{PrimTy}OffForeignObj ops to Foreign's
export list + upd. the docs on same a little.

Incorporate some fixes for Hugs

While fiddling around with concurrency & exceptions this afternoon, I
realised an important property of raiseInThread: it is synchronous.
The exception is raised immediately, and doesn't sit in some queue
somewhere waiting to be raised at a later date.

This is important if you have eg. two threads that can kill each
other.  We can guarantee that only one of the threads gets to kill the
other, under no circumstances can there be two killThreads fired in
opposite directions at the same time.

Documented this.

Change that slipped through the net in the recent build system changes.

Improved on the install rules for libexecs and bins a little
(at least from a Win32 pov - you can now install "happy.bin"
without trouble.)

Support for generating a VERSIONINFO resource that can be pinned
onto generated DLLs and EXEs (Win32 specific.)

Delay shutting down WinSock until exit() time

Always prepend `#! $(INTERP)' to a script, even if we're BIN_DISTing.
It doesn't do any harm, and means that we can execute the BIN_DIST
version of a script from the build tree, provided it doesn't depend on
any global paths.

GHC versions are now of the form AA.BB.C, where AA is the major
version, BB is the minor version, and C is the patchlevel.

Part of this commit accidentally ended up in an earlier commit to
fptools/ghc/mk/version.mk, which contains some comments on the version
rationale.  Excerpt:

# ProjectVersionInt does *not* contain the patchlevel (rationale: this
# figure is used for conditional compilations, and library interfaces
# etc. are not supposed to change between patchlevels).

i.e. a patchlevel is for bugfixes only, no new features or
interface-changing changes.  Unfortunately, we made this mistake in
4.04pl1, and won't be making it again.

Build System Cleanup
--------------------

This commit is intended to clear the way for automatic RPM building.
In particular, the idea is that 'make install' doesn't build anything,
so you can do

	make install prefix=/some/tmp/prefix

to install everything in /some/tmp/prefix, but leave the hardwired-in
install directories the same.

For scripts that depend on the install dir (currently just the GHC
driver and mkdependHS), we now build two versions: <script>-inplace,
which is used for running the script from the build tree, and
<script>, which is the to-be-installed version.

NOTE: binary distributions are now a little bit trickier to build.
You *must* include the line "BIN_DIST=1" in your build.mk if you
intend to make a binary distribution from the current build tree.
This is because certain scripts have to be built differently, and we
don't rebuild them when doing 'make binary-dist' anymore (since 'make
binary-dist' just does a 'make install' with a re-targetted prefix,
just like the RPM builder).

Other changes

	- the binary-dist machinery is now all in fptools/Makefile
	- removed a gratuitous $(package)-$(version) level of
	  directories from the binary distribution.
	- binary distributions are now placed under the package
	  name, rather than fptools/fptools.
	- various other minor cleanups.