This caused the CAF analysis to occasionally miss a CAF sometimes,
resulting in a very hard to diagnose crash.

This change means that we don't recover from erroneous type
declarations, thereby reporting fewer errors.  But trying to
recover confused GHC (leading to a pattern match failure in
the compiler Trac #7175), and it turned out to be tricky
to fix that.

So this patch takes the more conservative path of failing
earlier, perhaps reporting fewer real errors.

Pls merge to 7.6

I changed the behaviour slightly, e.g. i386/FreeBSD will no longer
fall through and use the Linux "i386-pc-linux-gnu", but will get the
final empty case instead. I assume that that's the right thing to do.

HaskellMachRegs.h is no longer included in anything under compiler/

Also, includes/CodeGen.Platform.hs now includes "stg/MachRegs.h"
rather than <stg/MachRegs.h> which means that we always get the file
from the tree, rather than from the bootstrapping compiler.

All the mechanism was there, but it wasn't being use for the
name-shadowing test.

Now the StableName#s don't need to have the same type parameters:

eqStableName# :: StableName# a -> StableName# b -> Int#

The CmmBlocks inside CmmExprs were not getting the PIC treatment

Return type was correct when TABLES_NEXT_TO_CODE was defined.

Also fix a bug where a dodgy import warning was emitted for data
families with a single constructor.

All the wibble seem to have cancelled out, and (non-debug) object sizes
are back to where they started.

I'm not 100% sure that the types are optimal, but at least now the
functions have types and we can fix them if necessary.

This caused a couple of .o files to change size. I had a look at one,
and it seems to be caused by the difference in size of these two
instructions:
    49 8b 5d 08            mov    0x8(%r13),%rbx
    49 8b 5c 24 08         mov    0x8(%r12),%rbx
(with a few nops being added or removed later in the file, presumably
for alignment reasons).

This has several advantages:
* It can be called from gdb
* There is more type information for the user, and type checking
  for the compiler
* Less opportunity for things to go wrong, e.g. due to missing
  parentheses or repeated execution

The sizes of the non-debug .o files hasn't changed (other than
Inlines.o), so I'm pretty sure the compiled code is identical.