Remove part of previous patch which was for debugging only.

Fix a nasty bug: when saving errno in the TSO after running a thread,
beware that the TSO might have moved, e.g. if the thread made a safe
foreign call in the threaded RTS and a GC happened.

We have to grab the new location of the TSO *before* saving errno.

MERGE TO STABLE

This was the most interesting bug hunt I've had in a while.  The crash
only showed up in about 1 in 4 runs of a program with 1000 Haskell
threads, running on the threaded RTS.  To make things worse, gdb
doesn't support watchpoints in programs with multiple threads... :-/

Make getAllocations() return an Int64 to avoid Int overflow.

-----------------------------------------------
	Decent error on re-use of built-in syntax
	-----------------------------------------------

Make the renamer complain nicely on stupid stuff like

	data T = []

(except in the Prelude, of course!)

-----------------------------------------------
	Improve location info on unused-import warnings
	-----------------------------------------------

Improving the location involves plumbing the location of the import a bit
more assiduously -- hence change to imp_mods in TcRnTypes

Comment update

Mention that ordering is important on the GHCi command line

spelling mistake in comment only

Reverse the order in which DLLs are loaded. This makes more sense in general and
should fix the trouble with the broken OpenGL libs on Gentoo. This fix has been
tested on x86 Linux, but reports from other platforms (WinDoze, MacOS X, ...)
are highly welcome, dynamic linking is quite platform-specific and surprising
things can happen from time to time...

Merge to STABLE when tested on other platforms

Add x86_64-*-openbsd* as synonym for amd64-*-openbsd.

Which one you get depends on which config.sub we have, and newer
(2004-*) config.subs normalise all amd64s to x86_64s.

Fix yet another bug in the THUNK_SELECTOR code.  Interestingly, I
spotted this one earlier but left a ToDo in the code rather than
fixing it (I think I wasn't sure whether it could happen or not).

The bug is to close another another way that eval_thunk_selector()
could return a pointer into to-space.  See comments for details.

mingw/msys fix: keep .lcomm directives

Install MinGW32 gcc and perl.

Improve error reporting for static flags

Windows make install sees an eagle.  It's heart aches - one day it will soar.

Enable GHCi on OpenBSD
     1. fix dlopen(NULL, ..) not working
     2. set USE_MMAP to overcome malloc'd memory not being executable

Tidy-up only; behaviour unchanged

---------------------------------
	Another delicate fix to the way that the
	Dreaded Monomorpism Restriction is handled
	---------------------------------

     I think this should merge nicely to the STABLE branch


In TcSimplify 1.130 I changed tcSimplifyRestricted (used for
declarations that fall under the monomorphism restriction) to use Plan
C.  Unfortunately, it later transpired that George Russel and Serge
Mechveliani both made somewhat-dubious use of overlapping instances
that conflicted with this change. Here's the example

  instance (HasBinary ty IO) => HasCodedValue ty

  foo :: HasCodedValue a => String -> IO a

  doDecodeIO :: HasCodedValue a => () -> () -> IO a
  doDecodeIO codedValue view = let { act = foo "" } in  act

You might think this should work becuase the call to 'foo' in the last
line gives rise to a constraint (HasCodedValue t), which can be
satisfied by the type sig for doDecodeIO.  But the restricted binding
act = ... calls tcSimplifyRestricted, and Plan C simplifies the
constraint using the (rather bogus) instance declaration, and now we
are stuffed.


This commit implements Plan D, which is like plan B except that it does no
improvement, and hence avoids plan B's flaw.  See the comments with Plan D.

The only criticism one might make of Plan D is that it may sometimes quantify
a restricted binding over "too few" type variables; but one can solve that
by adding a type sig.  So this seems better than the very subtle problems
with Plan C.

All of this is very much at the margin: George and Sergey are pushing
their luck.

Make the printing of samples really locale-independent

Nuke old version of the FFI spec

Fix mishandling of the BF_COMPACTED flag, which could lead to problems
when using the compacting collector (+RTS -c, or +RTS -M<size>).  In
fact, I'm not sure how it worked at all.

MERGE TO STABLE

Hack to generate correct dependencies for the RTS.  We now #define
PROFILING, THREADED_RTS and DEBUG when generating dependencies so that
we get (hopefully) a superset of all the deps.

In multiple declaration errors, give the location of one of the
declarations as the location of the error message, instead of the
top-level of the file.

Documentation typos

3 more openbsd platforms: m88k, sparc64 & vax
Just in case someone tries to run the testsuite.

Add openbsd platforms: alpha, arm, hppa, powerpc

GCC's __attribute__ handling seems to be a little bit stricter with GCC 3.3.3:

   * When a function declaration uses it, the corresponding definition has to
     use it, too.

   * Syntactically it is allowed only at the beginning of the function
     definition.

Let's hope that the current syntax is backwards compatible...

---------------------------------
	Make sure that GHCi loads all
	necessary orphan-instance modules
	---------------------------------

When you do :m A, we need to load any orphan-instance modules
below A.  That's easily done, but had been forgotten.

	Do not merge to stable: the fix is different there

White space

make a SrcSpan for the whole file that is not "unhelpful"

Report error in GHCi for unlifted bindings

Make addSrcSpan ignore unhelpful spans

Improve error handling for deriving(Typeable)

Better location info

Add ASSERT

Use IfaceDeprecs synonym

Fix (head []) value for src span in mkSimpleMatch

White spacing only

Put parens around binders in debug prints

Remove dead comments