* HTML documentation for "foo.xml" goes into directory "foo" again,
  not "foo-html". This is nicer and consistent with the behaviour for
  building the docs from SGML.

* Disabled building PostScript documentation in the spec files for
  now, there are some strange issues with the FO->PS conversion for
  some files which have to be clarified first.

Re-nuked SGML support, I crept back in with a previous commit. >:-(

The HTML documentation subdirectories have a "-html" suffix nowadays

Check for exit code *and* stderr when testing for "-input" in
FP_PROG_AR_SUPPORTS_INPUTS macro.

Cleanup: all (well, most) messages from the RTS now go through the
functions in RtsUtils: barf(), debugBelch() and errorBelch().  The
latter two were previously called belch() and prog_belch()
respectively.  See the comments for the right usage of these message
functions.

One reason for doing this is so that we can avoid spurious uses of
stdout/stderr by Haskell apps on platforms where we shouldn't be using
them (eg. non-console apps on Windows).

Fix bug in isEmptyMVar#

Preserve ExprWithTySig after type checking

Make error messages consistent

devolve some library-specific configuration

Rearrange so that Regs.h include comes before inline function
definitions.  I'm guessing that newer versions of gcc are less picky
about this, which is why we hadn't noticed it.

Make the UPD_IND macros compatible with older gccs that don't accept
declarations in the middle of a block.

$T_PRE_APP char now appears to be a '#' on OpenBSD/x86.

devolve the recently added dirent checks to a new libraries/base/configure.ac

Guerrilla tactics: Added --enable-hopengl

Nuked SGMLDocWays

Minore package GHC fixes, and a couple of changes for Visual Studio.
Messages from the compiler should now go through a new API in
ErrUtils, so that they can be redirected by the GHC client if
necessary.  (currently not all messages go through this interface, but
some of them do).

Better fix for previous from Ian Lynagh.

Apply the fix from

   [ 1015205 ] ghc-split fails to quote path when used as regex

namely escape backslashes in the pathname $Tmp_prefix before using it
as a regex.  The fix doesn't look like it solves the general case, but
it'll do for now.

Catch exit(127) result from raw system, and report that we couldn't
execute the program.  This is a semi-hack; all exec() errors get
reported as exit(127) by rawSystem, but if we treat it just like a
program that does exit(127) then the user sees no output.

Fix profiling (ASSIGN_CCSID macros had gone away).

RTS_SYMBOLS: added closure_flags (innards-peeking user code may use this); merge to STABLE

Fixed version/release magic. Actually I think we're doing things the wrong way
here: We should use AC_INIT with a version parameter as the central place for
this info and substitute what's needed into the rest of the files.

markup fixes (and one typo) in arrow notation section

add the Debian location for XSL stylesheets

Get the version number into ghc-pkg by creating Version.hs from the
Makefile to avoid GHC and possibly shell versionitis.

Try to fix up the previous commit.  I'm not sure if this is quite
right, but it should be closer.

Warning fix on Win32.

Rationalise the install settings a bit.  Before, if you had $OWNER or
$GROUP set in your environment, then these would lead to options being
given to install, possibly with unexpected results.  This caused
install to fail for at least one person.

Now, $(INSTALL_OWNER) and $(INSTALL_GROUP) are set via
config.mk/build.mk as with other settings.

turns out that 5.04.x differs in how args are quoted across plats; try to cope.

Cleaned up FPTOOLS_UNDERSCORE (changequote is evil!) and renamed it to
FP_LEADING_UNDERSCORE

The crusade against configuration bit rot continues: Cleaned up header checks

Removed duplicate tests for headers

Nuked tests already done via standard AC_INCLUDES_DEFAULT

Removed unused function checks

doh

escape GHC_PKG_VERSION double quotes on all platforms when compiling with <=6.0.x

Trimmed a few function checks

Beautified test for ghc-pkg

Aesthetic fix: Redirect error messages for non-GNU ar to /dev/null

all target: if building GMP, only descend into gmp/ when way=""