Fix bug #1 in the unregisterised RTS:

Now that catch# is a fully-fledged IO operation, it has to respect the
return convention for IO, in particular if we don't have an R1
(eg. when unregisterised) then the return value goes on the stack.

The return convention could be better, after all we use R1 when
returning normal values, but I'll just fix it for now.

Fix x86 NCG so the compiler can compile itself `-O':

-- Implement fake x86 insn GITOD/GITOF.
-- Implement primops ReadMutVarOp and WriteMutVarOp.
-- Pro tem, disable use of %eax as a spill temp.
-- Clarify wording of Rules of the Game comment in MachCode.

Remember all the classes loaded from an object file group, and
call visitClass on them at the end of processInterfaces(), so that
the .level numbers on the class get calculated.

Add hugsprimPmFail.

-- Make default defaults work in combined mode
-- rename some fns in lib/Prelude.hs to match names in HugsPrel.lhs

trivialCode on x86 is the heart of instruction selection for
expressions.  It is definitely a non-trivial in complexity.
To generate correct code it needs to observe preserve several
delicate invariants, but didn't.

-- Recorded in MachCode.lhs the "Rules of the Game"; ie what I think
   the required invariants are.

-- Completely rewrote trivialCode (also shift_code).  I think it should
   handle all cases correctly, and has special treatment for literal
   operands.

-- Updated NOTES file to record issues which need to be resolved
   before x86 nativeGen can be considered ready for public use.

Export findReservedRegs for all architectures.

primOpStrictness(CatchOp): catch# isn't strict in its first arg.

Double the number of RESERVED_C_STACK_BYTES so as to give the
native code generator up to 508 spill slots.

-- Cosmetic changes in register allocator.

-- Implement macro HP_GEN_SEQ_NP.

-- MachCode(trivialCode, x86): because one of the operands is also
   the destination (on this 2-address arch), it's invalid to sequence
   the code to compute the operands using asmParThen [code1, code2].
   since the order of assignments matters.  Fixed.

Spilling and x86 shift-code cleanups.

Add configure target to detect when autoconf needs to be run.

Include the configure script in a src dist.

Change primOpUsg(CatchOp) to reflect changed arity of catch#.  Also
update primOpStrictness(CatchOp).

We don't want -keep-{hc,s}-files-too on by default, and why do we need
-H32M -K32M for PrelBase.lhs?  Hans: I want a word with you :)

Change the type of catch# to

	catch# :: (W# -> (# W#, a #))
	       -> (b -> W# -> (# W#, a #))
	       -> W# -> (# W# , a #)

where W# == State# RealWorld.  In other words, make it explicit that
catch# is an IO operation and takes IO operations as arguments.  The
previous type was too general, and resulted in catch# having the wrong
arity which could cause mis-optimisations.

The down side is that we now have to pass the state token around
inside the primop instead of doing it in the Haskell wrapper, and
raiseAsync() also has to build a PAP(handler,exception,realworld)
instead of just a PAP(handler,exception) when it invokes a handler as
a result of an async exception.

I also added some optimisations to (un)?blockAsyncException to not
grow the stack if it can be avoided, such as when we're about to block
async exceptions and there's a blockAsyncExceptions_ret stack frame on
the top of the stack.

The bd->free field of a block descriptor is supposed to be set to -1
for free blocks, if we're #ifdef DEBUGging.  It wasn't sometimes.

Deal with ThreadRelocated in checkTSO().

Add notes about where the IO representation is wired in to various
parts of the source tree, since I'm about to add some more.

comment fixup

First pass at implicit parameters.  Honest, I didn't really go in *intending*
to modify every file in the typechecker... ;-)  The breadth of the change
is partly due to generalizing contexts so that they are not hardwired to
be (Class, [Type]) pairs.  See types/Type.lhs for details (look for PredType).

Modifications to make x86 register spilling to work reasonably.  It
should work ok most of the time, although there is still a remote
possibility that the allocator simply will be unable to complete
spilling, and will just give up.

-- Incrementally try with 0, 1, 2 and 3 spill regs, so as not to
   unduly restrict the supply of regs in code which doesn't need spilling.

-- Remove the use of %ecx for shift values, so it is always available
   as the first-choice spill temporary.  For code which doesn't do
   int division, make %edx and %eax available for spilling too.
   Shifts by a non-constant amount (very rare) are now done by
   a short test-and-jump sequence, so that %ecx is not tied up.

-- x86 FP: do sin, cos, tan in-line so we get the same answers as gcc.

-- Moved a little code around to remove recursive dependencies.

-- Fix a subtle bug in x86 regUsage, which could cause underestimation
   of live ranges.

bump version to 4.07 on the main trunk.

don't use ^ in patterns, Solaris sh barfs on it.

Commit all changes prior to addressing the x86 spilling situation in
the register allocator.

-- Fix nonsensical x86 addressing mode hacks in mangleIndexTree
   and getAmode.

-- Make char-sized loads work properly, using MOVZBL.

-- In assignIntCode, use primRep on the assign node to determine
   the size of data transfer, not the size of the source.

-- Redo Integer primitives to be in line with current representation
   of Integers.

Added note about the necessity of using the Cygnus DocBook tools (rather
than, e.g., SuSE).

Removed redundant rules (%.dvi : %.tex &c.) that were breaking DocBook
(re)building.

Corrected release notes URLs

Observe the C conventions for use of the FP register stack.  In particular,
free up any live fp registers prior to non-local control transfers.  Sigh.

This is not good.  The FP situation needs to be reviewed once the rest
of x86 nativeGen is stable.

genCCall (x86): pass float args in double format.

spillReg, loadReg (x86): always spill fp registers using double-sized
loads/stores, even if they nominally contain only a float value --
otherwise the spill loses the extra guard bits.

Finish pass 1 over the build system docs.

Fix some of the gross inaccuracies in this documentation.

Removed version number and reelucidated some text swallowed by index entries.

Changed default text size to "normal".

spillReg, loadReg (x86): spill above %esp, not below it.  Duh.  If you
spill below %esp, ccalls, which use stack below %esp, can trash the
spill area.

Updated version to 4.06

Sven's first autoheader commandment: Don't change config.h.in manually...!
(Why is mk/config.h.in in the repository? configure is not, either.)

Teach magicIdRegMaybe about R9 and R10.

Print a useful panic msg if getRegister(x86) can't reduce a tree.