GitLab

The old dph-par and dph-seq CPP libraries are gone. The DPH front end libraries are now dph-lifted-*, and are only built in one way.

This big patch implements a kind-polymorphic core for GHC. The current
implementation focuses on making sure that all kind-monomorphic programs still
work in the new core; it is not yet guaranteed that kind-polymorphic programs
(using the new -XPolyKinds flag) will work.

For more information, see http://haskell.org/haskellwiki/GHC/Kinds

This is a pretty egregious mistake in the recompilation checker: in a
declaration with multiple binders (e.g. data T = A | B) we were giving
all the binders the same fingerprint when referenced, so e.g. an
unfolding that mentioned A would get the same fingerprint as if it
mentioned B instead.  The fix is of course to give them all different
fingerprints.

Fixes some core-lint errors when compiling with profiling

Now included:
    - all language flags
    - all filename-related flags (-i, -osuf, -hidir etc.)
    - all CPP-related flags (-I, -D, -U)

* if -XParallelArrays is given, the symbols for the desugarer are
  taken from 'Data.Array.Parallel' (from whichever package is
  exposed and has the module — the home package is fine, too)
* if -fvectorise is given, the symbols for the vectoriser are
  taken from 'Data.Array.Parallel.Prim' (as above)

(There is one wired in symbol left, namely the data constructor
'base:GHC.PArr.[::]. It'll die another day.)

As usual it was to do with the handling of bottoms,
but this time it wasn't terribly subtle; I was using
andArityType (which is designed for case branches) as
a cheap short cut for the arity trimming needed with
a cast.  That did the Wrong Thing for bottoming
expressions.  Sigh.

(fix build failure with -split-objs)

Also, use the Win32 API (CreateThread) instead of the CRT API
(_beginthreadex) for thread creation.

This means ghci now correctly picks up the package flags
for tab completion.

iWe were adding c:/... to $PATH, but : is the separator
in $PATH.