Make this more Haskell 1.3-insensitive

Allow the following (not strictly H98) construct:

    do
	blah
	e $ do
	blah

This doesn't break any existing code, just allows more programs
through.  Only the semantics for layout after a 'do' have changed.

Add a test for deriving Show (f (f a)); makes Hugs hang

Make the r.u.n message happen only with -DDEBUG

Back out a change that creeped in by mistake.

Clean out objects generated from assembly files.

Made the compilation of the RTS almost warning-free and improved the
output of some "barf"s in the garbage collector . This is part of my
epic crusade against "scavenge_mut_list: strange object?"... :-(

Suppress warnings by making some static symbols extern.  Sigh.

Someone with a big stick told me to remove -ddump-cpranalyse
Use -ddump-cpranal instead.

CPR Analysis Mark 2. Slightly more elegant,  and (I believe) now
copes correctly with references to CAFS and non-top level function
bindings.

(keving)

Add dump-cpranal to agree with the users_guide.  dump-cpranalyse will
still work but is deprecated and could be removed by someone with
a bigger stick than me.

Back out changes for "specialid"s.  It didn't work this way: there was
a conflict in the grammar.

Add explanatory comments

Fix SetLevels so that it does not clone top-level bindings, but it
*does* clone bindings that are destined for the top level.

The global invariant is that the top level bindings are always
unique, and never cloned.

Some renamer fixes

* Correct the defn of Rename.isOrphanRule (caused a Sergey bug)

* Tidy up the Rename.implicitFVs stuff

Propagate changes for tcSimplifyToDicts

Reduce renamer warnings

* Add Type.repType

* Re-express splitRepTyConApp_maybe using repType

* Use the new repType in Core2Stg

	The bug was that we ended up with a binding like
		let x = /\a -> 3# +# y
		in ...
	and this should turn into an STG case, but the big lambda
	fooled the core-to-STG pass

Improve common sub-expression stuff
	- better hash function
	- add Const.isBoxedDataCon, and use it in CSE
	- don't CSE for nullary constructors

Fix lost specialisations.  There were two problems

	{-# SPECIALISE f :: Int -> Rational #-}
	fromIntegral	=  fromInteger . toInteger

This generates

	fromIntegral_spec = fromIntegral d

for some suitable dictionary d.  But since fromIntegral is small,
it got inlined into fromIntegral_spec, thus losing the specialised
call (fromIntegral d) that was the whole raison d'etre of fromIntegral_spec.
Haskish solution: add an inlne pragma for the _spec things:

	fromIntegral_spec = _inline_me (fromIntegral d)

Now we won't inline inside.  But this showed up a related problem.  The
typechecker tries to common up overloaded things, so it actually generates

	m = fromIntegral d
	fromIntegral_spec = _inline_me m

which is pretty stupid.  Using tcSimplifyToDicts (instead of tcSimplify)
in TcBinds.tcSpecSigs fixes this.

a couple more weird cases for lexing of qualified names.

Fix some pretty subtle bugs in the lexing of qualified names.

writeHandle has been removed.

Move -fwarn-unused-{binds,matches} back into -W (two votes for, one
against).

nit-picking lexical tests for M.<reservedid> and M.<specialid>

Jump to the join point when returning a new constructor to a bind
default.  Fixes:  recent panic in mkStaticAlgReturnCode.

Don't treat GHC extension keywords as such when recognising qualified
varids.  This is a quick fix for the bug that M.label isn't recognised
as a qvarid with -fglasgow-exts on.

Don't attempt to close stdin or stdout.  From Kevin Atkinson
<kevinatk@home.com>.

Document the SocketPrim.PortNumber type

'do' expressions must have a non-empty list of statements.

Fix some race holes in the handle locking code, and clean it up a little.

Fix `defined but not used' warning to omit *all* identifiers beginning
with underscore, not just top-level ones, following Haskell report.

Rescue UsageSP analysis from bit-rot.

Resurrect -funfold-casms-in-hi-file and probably make Sigbjorn go wild
:-)

Readline needs Posix stuff

Make -optc-DDEBUG work again

Just a small test for my CVS setup: Changed two references to ghc-4.03
to ghc-4.04

+RTS -s gives summary statistics only.

Some fixes to the TICK_ macros in raiseAsync.

Set context_switch to 1 after scheduling a signal handler.  This will
ensure the running thread is interrupted as soon as possible.