Display (overall) averages for per-module statistics as well as for
per-program statistics.

Also, with `-d' (default; `-n' to turn off) compute geometric standard
deviations and display +/- 1 s.d. points as well as average.  Note that
since these are *geometric*, the lower point will appear closer to the
average than the upper.  This is correct behaviour.

Filled in some missing support for importing defs w/ implicit params.
Fixed bug in improvement w/ implicit params.

forceIO_ret_entry, when #ifndef REG_R1: remember to load Node.

Because of the previous HaXml-"improvement" :-}, syslib text now
depends on syslib data.

Temporarily disable Stix inlining whilst we figure out how to fix
obscure bugs in the register allocator.  With inlining off, everything
seems to work, and the compiler can boot itself, but when enabled,
inlining exposes a register allocation bug which causes
ghc/compiler/utils/FastString.lhs to be miscompiled on Linux (the big
arithmetic expressions for computing hash values).

oops, remove assigment to non-existent "emitted" field in a CCS.

Added removeIndirections() and make use of it in lookupStableName() -- increases
the possibilities of StableName sharing.

Change the behaviour of the cost-centre stack profiler for recursive
sccs.  Before, we used to remove the old copy of the CC from the stack
and push the new one on.  Now, we record back-edges in the graph when
a recursive scc is detected.

This should give saner-looking cost-centre stacks for heavily
recursive code.  Indeed, it reduces the number of stacks in the
profile for some examples considerably.

Ctrl-C now interrupts the RTS safely.  Previously it called
shutdownHaskellAndExit() from the signal handler directly, which isn't
safe because we may have been interrupted during GC or whatever.  Now
we set the interrupted flag and wait for the RTS to shut down by
itself.

Make foreign import work in combined mode:
-- Allow interpreter to do ccall primops even in combined mode
-- Implement hugsprimMkIO in combined mode, so as to wrap up a
   an IO value created by Hugs in a form compatible with GHC's IO
   representation.

The GHC Prelude doesn't seem to export Addr.  Make it do so in combined
mode for compatibility with standalone mode.

Fix boundary cases for quotInteger, remInteger & a few others.  These
changes somehow escaped 4.06 and have been sitting in my tree for a
while.

Update sparc-specific parts of NCG to use new infrastructure, so they
will at least compile under Solaris.  Won't work (yet) tho.

Fix signatures w/ implicit parameter types in them (in particular,
correctly handle the case where there are no type variables).  Also
made a few more things Outputable.  Nuke outdated comment in Parser.y.

Add haxml/lib to imports for -syslib text.

Only fill in the ARG_TAG when DEBUG is on, for unregisterised.

Many changes to improve the quality and correctness of generated code,
both for x86 and all-platforms.  The intent is that the x86 NCG will
now be good enough for general use.

-- Add an almost-trivial Stix (generic) peephole optimiser, whose sole
   purpose is elide assignments to temporaries used only once, in the
   very next tree.  This generates substantially better code for
   conditionals on all platforms.  Enhance Stix constant folding to
   take advantage of the inlining.

   The inlining presents subsequent insn selection phases with more
   complex trees than would have previously been used to.  This has
   shown up several bugs in the x86 insn selectors, now fixed.
   (assumptions that data size is Word, when could be Byte,
    assumptions that an operand will always be in a temp reg, etc)

-- x86: Use the FLDZ and FLD1 insns.

-- x86: spill FP registers with 80-bit loads/stores so that
   Intel's extra 16 bits of accuracy are not lost.  If this isn't
   done, FP spills are not suitably transparent.  Increase the
   number of spill words available to 2048.

-- x86: give the register allocator more flexibility in choosing
   spill temporaries.

-- x86, RegAllocInfo.regUsage: fix error for GST, and rewrite to
   make it clearer.

-- Correctly track movements in the C stack pointer, and generate
   correct spill code for archs which spill against the stack pointer
   even when the stack pointer moves.  Redo the x86 ccall mechanism
   to push args on the C stack in the normal way.  Rather than have
   the spiller have to analyse code sequences to determine the current
   stack offset, the insn selectors communicate the current offset
   whenever it changes by inserting a DELTA pseudo-insn.  Then the
   spiller only has to spot DELTAs.

   This means having a new native-code-generator monad (Stix.NatM)
   which carries both a UniqSupply and the current stack offset.

-- Remove the asmPar/asmSeq ways of grouping insns together.
   In the presence of fixed registers, it is hard to demonstrate
   that insn selectors using asmPar always give correct code, and
   the extra complication doesn't help any.

   Also, directly construct code sequences using tree-based ordered
   lists (utils/OrdList.lhs) for linear-time appends, rather than
   the bizarrely complex method using fns and fn composition.

-- Inline some hcats in printing of x86 address modes.

-- Document more of the hidden assumptions which insn selection relies
   on, particular wrt addressing modes.

oops, back out local patch that accidentally got committed with rev. 1.139.

fix = vs. == in conditional

document the -recomp option explicitly.

'IPVARID :: type' moved from btype to type, to reduce conflicts.

Wibble.

(#ifdef incorrectly nested inside begin-end code)

Use createIOThread and not createGenThread, even for the interpreter.

Add simple mkWeakPtr/addFinalizer test.

The finalizer passed to mkWeak# might be NULL, so we'd better not
claim it's always pointer if we need to GC.  Another bug squashed...

*ahem*  Re-enable -fwarn-unused-binds and -fwarn-unused-imports...  :-}

Deprecations of variables now works, although the source location is
not yet reported correctly and the code needs some cleanup. Added a
new flag -fwarn-deprecations to the set of standard warnings. The
syntax of deprecations has been extended to deprecate types, classes,
or even constructors, although this does not work yet.

add -ltermcap when looking for -lreadline.

From: Michael Weber.

Nuke PTR_ON_HEAP and all associated ifdeffery; we always need it.

Make the desugarer a little more liberal about the class of case
expressions which get the special unboxed tuple treatment.

silence linker warnings on HP/UX.

From: Eric Schweitz <schweitz@nortelnetworks.com>

update for format changes in cacheprof.out.summary.

primUnpackString --> hugsprimUnpackString

Correct [un]blockAsyncExceptionszh_ret in the #ifndef REG_R1 case.

Don't always use GMP from ghc/rts; instead consult HaveLibGmp first.

Remove debugging junk accidentally left in :-(

Wibbles for Win32 standalone compilation of Hugs:
-- Turn off debugging miniinterpreter
-- Change SIZEOF_INTP (which no longer exists) into SIZEOF_VOID_P

Use the GMP library built in ghc/rts, not the system-supplied one,
since that doesn't exist on Cygwin.

#ifdef wurbles to aid Win32 compilation.

Handle `with' more cleverly.  I was generating partially applied methods
for the case where the `with' expression was also overloaded, but this
was buggy, and completely unnecessary.  Instead, simply force the method
binding at the point of the `with' expression (we reap no benefits from
pushing the sharing further out anyway), and release the remainder of
the method's context into the LIE.