GitLab

In i386 position-independent symbol stubs, the NCG used to print
	jmp %edx
instead of
	jmp *%edx

Apple's assembler used to silently accept this, but has recently started to print warnings about it.

MERGE TO STABLE.

... so that GHCi doesn't complain about duplicate symbols when two C modules define the same static variable.

MERGE TO STABLE.

The __i686.get_pc_thunk symbols generated by gcc's PIC code generator are weak definitions that appear in every object file, so we need to deal with them.

MERGE TO STABLE.

and hopefully be a bit cleverer about finding things automatically.

 publish-binary-dist

This patch cleans up the GHC API, and adds some functionality: we can
now compile to object code inside GHCi.

Previously we had:

  data GhcMode
    = BatchCompile
    | Interactive
    | OneShot
    | JustTypecheck
    | MkDepend
  
  data HscTarget
    = HscC
    | HscAsm
    | HscJava
    | HscInterpreted
    | HscNothing

There was redundancy here; if GhcMode is Interactive, then only
HscInterpreted makes sense, and JustTypecheck required HscNothing.
Now we have:

  data GhcMode
    = CompManager       -- ^ --make, GHCi, etc.
    | OneShot           -- ^ ghc -c Foo.hs
    | MkDepend          -- ^ ghc -M, see Finder for why we need this

and HscTarget remains as before.

Previously GhcLink looked like this:

  data GhcLink = NoLink | StaticLink

Now we have:

  data GhcLink = NoLink | LinkBinary | LinkInMemory

The idea being that you can have an HscTarget of HscAsm (for example)
and still link in memory.

There are two new flags:

  -fobject-code selects object code as the target (selects
                either -fasm or -fvia-C, whichever is the default)
                This can be usd with ':set' in GHCi, or on the command line.

  -fbyte-code   sets byte-code as the target.  Only works in GHCi.
                One day maybe this could save the byte code in a file
                when used outside GHCi.

  (names chosen for consistency with -fno-code).

Changes to the GHC API: newSession no longer takes the GhcMode
argument.  The GhcMode defaults to CompManager, which is usually what
you want.  To do JustTypecheck now, just set hscTarget to HscNothing.

This use was the only thing keeping parsec in core-packages, and
we already have a dependency on alex+happy anyway.

3 small regexps were responsible for pulling 3 packages into core-packages.

The new code should also do a better job of hiding
"call-clobbered register used" warnings.

It used to be the case that
    autoreconf
in the root would also autoreconf in any libraries that need it, and
    ./configure
in the root would also configure all the libraries.

However, cabal now configures the libraries, so they were getting
configured twice.

Thus now a small shellscript, boot, autoreconfs the root and all libraries
that need it, and ./configure in the root doesn't configure the libraries.

* configure can be told where ld is
* make an hsc2hs-inplace.bat
* tell Setup configure about foo.bar rather than foo on Windows
* tell Setup configure to pass a --with-cc=$(CC) argument to ./configure

This patch updates the External Core creator, pretty-printer, and parser to
agree on a concrete syntax for External Core, including the constructs
required by the change to System FC. Code to create valid ASTs from External
Core files will come later, as will bits for renaming, typechecking, and
desugaring.

(we don't actually do anything useful either, because Cabal doesn't have
an install-docs option).

Some fixes to adjustor functions.  The 8-byte address returned by the 
allocator is adjusted to be aligned to 16-byte boundaries.  Fixed a typo 
in inserting an immediate address into an instruction.

This fixes the calls to 5-argument and 6-argument functions in ffi009.  
Some functions still break.  I suspect it's related to passing arguments on 
the stack.

Save/restore two more registers in StgCRun().  The extra 
registers are used by ffi009.hs, when compiling with gcc 4.1.2.

A thread that was blocked on a blackhole but can now be woken up could
possibly be treated as unreachable by the GC, and sent the
NonTermination exception.

This can give rise to spurious <<loop>>s in concurrent programs, so
it's a good one to fix.

HANDLE is defined to be (void *) anyway, so this shouldn't hurt