Pass -lpthread to linker when using a 'threaded' RTS

Add ObjectIO to standard packages

Deadlock is now an exception instead of a return status from
rts_evalIO().

The current behaviour is as follows, and can be changed if necessary:
in the event of a deadlock, the top main thread is taken from the main
thread queue, and if it is blocked on an MVar or an Exception (for
throwTo), then it receives a Deadlock exception.  If it is blocked on
a BLACKHOLE, we instead send it the NonTermination exception.  Note
that only the main thread gets the exception: it is the responsibility
of the main thread to unblock other threads if necessary.

There's a slight difference in the SMP build: *all* the main threads
get an exception, because clearly none of them may make progress
(compared to the non-SMP situation, where all but the top main thread
are usually blocked).

'std' now also needs user32 (under win32)

package rts (mingw32): in installed mode, add gcc-lib/ to library_dirs,
forcing it to be used by default. This allows other backend tools to be
used (e.g., `dllwrap') without haphazardly mixing & matching (import)
libs.

MERGE TO STABLE.

Sets the X11 C compiler and linker flags for package xlib correctly now

std & mingw32: include 'kernel32' in extra_libraries

[btw, this config file needs a bit of work to have it work on a
i386-unknown-cygwin plat.]

Remove HDirect/COM package.

THIS CHANGE AFFECTS ALL OBJECT FILES COMPILED FROM HASKELL.
Please say "make -C ghc/lib/std clean; make -C hslibs clean".

This commit eliminates spurious warning messages when compiling on
the Alpha.  There are two kinds of spurious warning messages:

(1) gcc: -noprefix_recognition: linker input file unused since linking not done

    This warning is because we pass the flag "-Xlinker -noprefix_recognition"
    to gcc.  We remove this warning by no longer passing the flag to gcc,
    and by removing the reason we were passing the flag in the first place:
    __init_* is now renamed to __stginit_*.

(2) .../includes/Regs.h: warning: call-clobbered register used for global
    register variable

    This warning and all other warnings except (1), we eliminate by
    passing the -w flag to gcc.

MERGE TO STABLE BRANCH

Build system hacks to split HSwin32.o into two parts, so that it
can be loaded into GHCi.  Uses the same gruesome hacks as HSstd.o.

Remove dependency of win32 on greencard (because the latter is merged
into the former). This is a temporary hack until the greencard library
is added to the new libraries.

Change the generation of package.conf and package.conf.inplace, so that
-lbfd and -liberty are included when necessary.

libm is a dependency of the RTS now, so don't also depend on it for
package "std".

Added -lm to the extra_libraries list for the rts package.  This is
because ldexp() is used by StgPrimFloat.c and is in -lm on alpha-osf3.

Make -u's and symbols into separate args

Add C includes path for Windows here rather than hardwiring into
SysTools.lhs. Change to include rather than includes.

package util: don't include readline on Win32

Add comment referring to horrible hack to do with HSstd splitting/merging.

Inching closer to making the HGL part of the standard build.

This commit is mostly to remove bits of xlib and greencard whch had
crept into the source tree.

ToDo:

- make greencard part of standard build (waiting for response from
  simonM)
- make greencard install libHSgreencard when it builds
- find a autoconf-respecting way of installing the package
- only build xlib and HGL if you have X11 installed
- merge all these changes into HEAD
- should the demos/ tests be built as part of the standard build?
  what's the convention here?
  (Note that the tests can't be used in batch more - you gotta click
  on the window.)
- a little debugging - the tests aren't quite working.

Tidy up formatting

On mingw32, package std for ghci is split into HSstd1.o and HSstd2.o.

[After fumbling around in the makefile forest for a while, Sigbjorn
enlightened me as to where one registers new hslib packages.  Golly,
never thought of looking over here.  I'd kinda expected that it would
be in the makeforest since makefiles seem to contain all (or almost
all) the information needed to install a package.  (Any chance this
stuff could be moved into the makeforest in the new library
infrastructure?)  It seems that'd be more effective since it'd be able
to benefit from all that good information that autoconf spent so long
figuring out about your X installation, your platform, your
internationalisation library, etc.]

Oh, yeah, the change... Added xlib to the package list.  Result looks
plausible.

Add wsock32 to list of libraries needed for rts under Windows.

comment fixes

Allow paths in package.conf to begin with the string "$libdir" which
is magically replaced by the compiler with either

  (a) the argument to the -B<dir> option, if one is present, or
  (b) the wired-in libdir otherwise.

This means that the standard package.conf file is independent of the
install location and we don't need the post-install-script hack any
more (Julian - could you update the install machinery when the time
comes? thanks.)

Don't include the win32 and com packages unless we're building for Windows.

Plumb LibsReadline into extra_libraries for -package util, rather than
extra_ld_opts, so that GHCi knows to load them too.

HSnet_cbits R.I.P.
(BTW, that's what I meant with "monolithic approach" :-)

- make the pkgconf program independent of the location of the build
  tree - instead of Config.hs, we pass the configuration parameters in
  as options.  This is necessary for .hc bootstrapping.

- remove the ghci driver stuff, it moved to the ghci subdirectory.

- remove ghc5.hs

PutFullMVar went away

- Don't duplicate the definition of PackageConfig, instead hackily #include
  it from ../utils/ghc-pkg/Package.hs.

- Add missing source_dir fields to PackageSrc.hs.

Apply stat fix for mingw. It's ugly, it's implausible, but it seems to work.
Unlike Cygwin...(missing no. 3).

Remove -l from library name in libraries list (oops!)

Add winmm to libraries for RTS under Windows, where it's now used for
threadDelay.

Move library names from extra_ld_opts to extra_libraries.

Fix up inter-package dependencies so the run-time linker doesn't barf.

Make lang depend on data (since it does, really.)

lang package needs an undefined symbol (for deprecated rts_mkAddr)

Changes needed to get the head to compile on sparc-solaris.

add missing -u symbols.