Let newtypes in stub files generate correct C type names.
It worked in ghc-4.09 but was not merged into 4.11?

Unboxed returns, + a little peephole optimisation.

FILE RENAME: from fptools/ghc/rts/Bytecodes.h
               to fptools/ghc/includes/Bytecodes.h

Wibble

Correctly unpack constructors onto the stack.

LiberateCase fiddling

Bogons in previous commit

Get CAF info right

Some renaming in HscTypes

Add #include of ghc/includes/config.h so tests of TARGET_OS_mingw32 work.

Correct CPR information.  How it ever worked I don't know.

* The CPR info on a newtype constructor should be NoCPR, whereas
  before it was ReturnsCPR!

* Minor: don't change CPR info on implicit Ids

Slight repackaging in HscMain

Rename omitIfaceSigForId to isImplicitId

config.h depends not just on fptools/mk/config.h but on
fptools/mk/config.mk, for architecture information.

add -fno-{cse,cpr,strictness} flags.

Don't need to set the CafInfo on non-top-level Ids, the default is
correct now.

change the default CafInfo to NoCafRefs, to avoid PrimOps,
constructors and other random Ids from declaring that they refer to
CAFs.

Add documentation on -fno-monomorphism-restriction

Add -fno-monomorphism-restriction

Comment

Complete Simons arity/caf changes

use $(GHC) for $(MKDEPENDHS), since $(GHC_INPLACE) isn't built yet.
There's a minor problem with this, in that if you've added any options
to GhcLibHcOpts that are specific to the new compiler, 'make boot'
won't work.  bleugh.

Figure out CafInfo during CoreTidy.  This is the final piece of the
puzzle in getting the final IdInfo from the Core2Core phases, rather
than waiting for the STG code.

This simplifies the SRT phase, in that it no longer has to have a
complicated circular algorithm to figure out the CafInfo at the same
time as the SRT layout.

touch the object file in the HscNoRecomp case (duh)

Fix many obvious bogons and partially emerge from Wrong Stack Offset Hell.

Minor enhancements to printing machinery to aid debugging the BC generator.

Dont drop inline-mes quite so easily.  This was leading to substWorker warnings.

Dont warn about uses of deprecated things in its own module

Do a better job of eta expansion.

This showed up in one of Manuel's programs, where he got code like:

    $wsimpleGen
	     ww
	     (\ i :: Int ->
		  case i of wild1 { I# i# ->
		  case w of wild2 { I# e# ->
		  __coerce (ST RealWorld ())
		  (\ s# :: (State# RealWorld) ->
		       case writeIntArray# @ RealWorld mba# i# e# s#
		       of s2#1 { __DEFAULT ->
		       (# s2#1, () #)
		       })
		  }
		  })
	     s2#

The argument wasn't eta expanded, so it got right through to
the code generator as two separte lambdas.

Needless to say, I fiddled around with things in a vain attempt
to tidy them up.  Yell if anything seems to go wrong, or perfomance
drops on any programs.

Improve iface diffs messages

Errors and warnings were back to front

Add foldOL

Be very keen to inline simple primops.
  We give a discount of 1 for each arg so that (op# x y z) costs 1.
  I found occasions where we had
	f x y z = case op# x y z of { s -> (# s, () #) }
  and f wasn't getting inlined

Tidy up functional dependencies slightly

wibble

Workers get local names initially; nuke mkDerivedName

Better handling of HsTupCon (tidy up + fix minor versioning bug)

Tidy up the Persistent Renamer State structure a little

Make it work even if mingw32_TARGET_OS is defined

Fix the bogus -fwarn-unused-exports problem