GitLab

------------------------------------------
	Improved handling of scoped type variables
	------------------------------------------

The main effect of this commit is to allow scoped type variables
in pattern bindings, thus

	(x::a, y::b) = e

This was illegal, but now it's ok.  a and b have the same scope
as x and y.


On the way I beefed up the info inside a type variable
(TcType.TyVarDetails; c.f. IdInfo.GlobalIdDetails) which
helps to improve error messages. Hence the wide ranging changes.
Pity about the extra loop from Var to TcType, but can't be helped.

1. Correct the handling of selectors for
   newtypes with polymorphic fields

2. Make certain that trivial constructor wrappers get inlined

Add missing MUT_CONS case (reported by Andy Cheadle)

Load the files given on the command line a bit later, and catch any
exceptions so that errors while loading these files don't cause GHCi
to exit completely.

Handle completely empty files without crashing in slurpFileExpandTabs.

The new story where InstallShield just crawls over a make install tree,
rather than having to specify the directory layout by hand. Some tweaks to
the way that external programs are installed, so that they too can be just
crawled over.

Fix a problem when a Haskell process is suspended/resumed using shell
job control in Unix.  The shell tends to put stdin back into blocking
mode before resuming the process, so we have to catch SIGCONT and put
it back into O_NONBLOCK.

Also:

  - fix a bug in the scheduler: reverse the order of the check
    for pending signals and the call to awaitEvent to block on I/O.

  - do a style sweep in Signals.c

Add setNonBlockingFd(), for use in the SIGCONT handler in Signals.c.

Fix typo.

Made db2pdf use jadetex|dvips|ps2pfd, not pdfjadetex. Two benefits:

  1. Smaller PDF files.
  2. jadetex is a little less exotic, and more likely to be on a given
system.

The actual motivation is only knowing of a working version of jadetex
for Windows and not of pdfjadetex (at least, not one that works on the
gargantuan tome that is the combined User's Guide and Libraries
Guide).

Much more in-depth description. Switch to the new more automated
process, and concentrate on dukin' it out with InstallShield.

Update sample output.  All the GHCi tests work as of today.

Add deadlock-detection test.

Add ':set args' and ':set prog' test.

- Fix the free variable calculation in schemeE following some changes
  to the global-vs-local name story in earlier parts of the compiler.
  (fixes GHCi breakage on the HEAD).

- Eliminate some duplicate free variable calculations.

The .NET BigInteger support class (yup, the all-embracing .NET
Framework classes don't include big integers; clearly not important
for Real Programmers).

This is basically Julian's sainteger.c module from STG Hugs
resurrected and with lots of # signs added.

Specialise-constructor rules active only in phase 0 (for Manuel & Gaby)

Give the template tyvars nice print names, as per a suggestion from
Koen Claessen.

Prelude> :i []
-- [] is a data constructor
[] :: forall a. [a]

-- [] is a type constructor
data [] a = [] | (:) a [a]

- document --print-libdir
- replace some <literal>s with <option>s where appropriate

Add --print-libdir flag

Add C++ extern "C" wrappers to generated stubs.

Set libdir to prefix on Windows only. Windows install paths are
currently different from those used under Unix, and it's probably best
to change the layout of the InstallShield to match Unix in the long
term. For now, just Make It Work.

Wrap the include file entry-points in extern "C" { ... } if this is a
C++ compiler.

Understand the various C++ suffixes.

Added Show instance for Concurrent.ThreadId; useful when
debugging/diagnosing.

Move extern declaration for 'environ' from Stg.h into HsPosix.h

merge from stable, revs:

  1.41.4.1  +4 -1      fptools/ghc/compiler/nativeGen/AbsCStixGen.lhs
  1.61.4.3  +10 -0     fptools/ghc/driver/mangler/ghc-asm.lprl

  Place a zero word after each reversed vector table, so that the vtbl
  label is really in the section we would like to claim it is in.  This
  is needed for the GC to work correctly.  Fixes a GHCi segfault
  reported by Ryszard Kubiak.

  1.41.4.2  +7 -0      fptools/ghc/compiler/nativeGen/AbsCStixGen.lhs
  1.61.4.4  +18 -0     fptools/ghc/driver/mangler/ghc-asm.lprl

  Add comments, much longer than the fix itself, giving explaination
  for yesterday's dummy-word-after-vtbl fix.

merge from stable, revs:

  1.74.4.1  +12 -11    fptools/ghc/compiler/nativeGen/MachCode.lhs
  1.30.4.1  +23 -0     fptools/ghc/compiler/nativeGen/Stix.lhs
  1.70.4.1  +2 -5      fptools/ghc/compiler/nativeGen/StixPrim.lhs

  Route all NCG panics to do with missing primop implementations and any
  other panic which could be caused by compiling legitimate sources
  through the function Stix.ncgPrimopMoan.  This emits a helpful message
  explaining what has happened, advises the use of -fvia-C as a
  workaround, and says please mail us.

merge from stable, rev 1.23.4.1:

  Win32 only: implement sane failure semantics + message when out of
  memory, rather than continuing and allowing the rest of the system
  to seg out.  Also, increase the 128M limit to 256M.

  This allegedly fixes the Antony Courtney Win32 segfault.

merge from stable, rev 1.60.2.10:

  Implement a debugging aid: look in GHCi's object symbol tables for
  symbols within DELTA bytes of the specified address, and show their
  names.  Only compiled with -DDEBUG.  The new fn is called ghci_enquire
  and you can call it from within a gdb session.

pipeLoop: Handle pipelines that end with ineffective
final stages (e.g., CPP and/or PP) correctly. i.e.,
need to copy the outcome of the pipeline into the
expected output file, since the final stage didn't.

Custom pre-processor documentation

Added support for a custom pre-processor pass:

  ghc -F -pgmF/path/to/a/pre/processor ...

will now run /path/to/a/pre/processor over Haskell
input sources. It is positioned in the compilation
pipeline just before the compiler proper, but after
unlit'ing and CPP'ing. The pre-processor is passed
the following command-line when invoked:

   /path/to/a/pre/processor orig_input_source_file_path
   			    input_source_file
			    output_source_file
			    <other options>

Additionally options can be fed directly to the
pre-processor via -optF<option> options.

The -F option causes the pre-processor to run _iff_ one
has been specified via -pgmF (there's some redundancy
here, but I went for this cmd-line interface as it's
consistent with the general -pgm<Foo> story).

Motivation:

 * hooking in a pre-processor is occasionally useful;
   e.g., cheap&cheerful way to integrate language
   extensions with GHC, compile-time syntax/style
   checking etc.

 * Artfully re-using the CPP phase (by specifying your
   own via -pgmP) doesn't really work as the driver
   really assumes that GNU cpp is what's being invoked
   (and path mangling is also performed on Win32 platforms).

   Additionally, there are cases when you want to be
   able to run CPP _and_ a pre-processor.

 * The alternative of running the pre-processor as a
   separate program in a Makefile (say) doesn't work
   in interpreted mode, and this approach also forces
   you to give up on recompilation checking when in
   batch mode.

Tidy up after my assumption-busting.

Why should pdfjadetex be in the same directory as Jade? Make the much
less silly assumption that it's on the PATH.

-------------------------------------------------------
  Correct an error in the handling of implicit parameters
  -------------------------------------------------------

	MERGE WITH STABLE BRANCH UNLESS HARD TO DO

Mark Shields discovered a bug in the way that implicit parameters
are dealt with by the type checker.  It's all a bit subtle, and
is extensively documented in TcSimplify.lhs.

This commit makes the code both simpler and more correct.  It subtly
changes the way in which type signatures are treated, but not in a way
anyone would notice: see notes with "Question 2: type signatures"
in TcSimplify.lhs.

Renumber tuples (decrement by one) to conform to new primop naming scheme.

ILX fixes to get the standard library to build and verify.

-Wall cleanup:

  - move some imports inside #ifdef GHCI
  - remove some unused bindings

- only generate split markers if we're splitting
- remove a couple of unused imports