Some fixes to this (still non-working) pass.

The decision to not make a static closure should only be taken for
top-level bindings.

Minor cleanup

Nuke -fcase-elim; sometimes things fail if you dont have it (Sergeys message)

print minInt as (mintInt+1)-1, to avoid gcc's conversion to unsigned.

Bump version to 4.04 (to avoid confusion with already released 4.03
for Win32).

Don't inline strLength, it causes gcc to run out of registers.

Test for pattern type signatures.

Make scoped type variables work.

rename -ddump-rdr to -ddump-parsed, and make it work.

Add HsBinds.isPragSig

Do CSE right after a simplification pass; needs no-shadowing invariant

Pragmas for default decls

Add common sub-expression

Remove exponential simplifier run-time bug

Many small tuning changes

Up heap size for Parser.hs a bit more.

Up the heap/stack for Parser.hs a bit.

Literate tidyups from Wolfram.

Typo from Wolfram.

Comment cleanup and literisation(?) by Wolfram Kahl <kahl@DI.Unipi.IT>.

Urk, more fixes.  This thing *really* needs to be taken out and shot.

Fix for default unregisterised builds.

Make use of the $(GhcWithRegisterised) build flag from config.mk.  You
can now get an unregisterised build by *either*

	- setting GhcWithRegisterised to NO in build.mk, or
	- adding 'u' to GhcLibWays.

The former will make all compilations unregisterised, whereas the
latter will require a '-unreg' flag to be passed to the compiler to
get the unregisterised build.

Test for Richard Reid's bug:  syntax C{} for constructing C with all fields set
to bottom.

Allow the syntax C{} for building an unlabelled constructor C and
setting all of C's fields to bottom.  The Haskell report is a bit
vague on whether this is legal, but it turns out to be quite easy to
support.

build Posix before Misc.

Don't rely on bad implementation of 'sum', now that we don't have one
:)

Allow ``...'' as argument of __casm.

Don't chop off the last character of the buffer.

tcrun004 uses existentials, needs -fglasgow-exts.

Forgot 'forall' and 'foreign' as possible variable names.

Oops, forgot that the memory returned from realloc could have moved.

Add tcfail082.stderr

Accept some test changes; add tcfail082, read021

prescript wibble

K&R -> ANSI token pasting (ie, x##y for x/**/y)

- add a test for non-parsing of strict newtype constructor fields
- other small fixes

Missing case in collectTypedPatBinders; Andreas Marth bug report

Add a few SPECIALISE/INLINE pragmas.