ghc-pkg doesn't understand the old syntax any more, so 'ghc-pkg -l' fails

Having 2.0.1 causes some unicode tests to fail

Now that we build haddock in the GHC build, we don't need to look for it
on the path.

* New packages format lets you select source control system in use
* Packages file now includes root repo explicitly
* Scripts darcs-all and push-all updated for the new packages format only
* New sync-all script, intended for use after Git changeover and for buildbots right now
* Had to remove libraries/bootstrapping from tree since Git cannot track empty directories without a hack
* Determine checkout date with Git using Darcs fallback in aclocal.m4

We shouldn't need it, as we don't call cygwin's gcc, and it was causing
problems with the nightly builders passing it to GHC.

On a 2.6.24 Linux kernel, it appears that timer_settime() for
CLOCK_REALTIME is sometimes hanging for a random amount of time when
given a very small interval (we were using 1ns).  Using 1ms seems to
be fine.  Also I installed a 1-second timeout to catch hangs in the
future.

They are now the same as hardtop/FPTOOLS_TOP_ABS, so use those instead.

Also removed some substitutions of / for \, as we now use a Haskell
program to find the top path, and it only makes paths with /s in.

Added a missing "test" and replaced the "-e" option of the shell-builtin
"test" with "-f". The former is not portable, see autoconf documentation.

We now have various flags that need to be passed to gcc, but only if
the local gcc version supports them.  So instead of wiring this
knowledge into ghc when it is built, we now put these "extra gcc
flags" into a text file in $libdir, extra-gcc-flags, which is created
by configure.  It is also created by the configure script of a binary
distribution, so a binary dist is independent of the gcc version used
to build it.

Hopefully I got the binary-dist stuff right, but binary dists aren't
currently working so we'll need to come back and check this.

  GHC only works with Perl versions 5.6 and later; however
  the current test greps in a rather non-future-proof way
  (explicitly against 5.6 and 5.8).
  
  We instead exploit the "v-string" change in 5.6, which
  displays the version string as "v5.x.y", rather than
  the earlier format "version 5.xxx_yyy".  Hence grepping
  against "v5" will make it future proof for a while.
  (At least until Perl 6 is installed as the system perl.)

Bindists should now work again, when doing "make install" at least.
"make in-place" is probably still broken.

The old code didn't work on Mac OS X, as "which" returns successfully
when it can't find the program.

* configure can be told where ld is
* make an hsc2hs-inplace.bat
* tell Setup configure about foo.bar rather than foo on Windows
* tell Setup configure to pass a --with-cc=$(CC) argument to ./configure

This happens when the build tree is a link-tree to the source tree,
because lndir will normally omit the _darcs directory.

Instead of asking for the last 100 patches, we just get the first 500
lines of output from darcs changes (ugh, better ideas welcome).

For non-release builds, we want to append a date to the version number
(e.g. 6.7.20070206).  Previously this was done by the nightly build
script, this new method figures out the snapshot version by querying
the darcs repository and finding the date of the most recent patch
(actually it finds the most recent of the last 100 patches, but that
should be good enough).  This is done by the configure script.

To handle source distributions, we create a file VERSION in the
top-level directory that contains the version number, and ship this in
the source distribution.  The configure script picks up the version
from this file if it doesn't see a _darcs directory.