This reverts commit 5ea49271.

jpm@cs.ox.ac.uk authored 12 years ago and ian@well-typed.com committed 12 years ago

Signed-off-by: Jose Pedro Magalhaes <jpm@cs.ox.ac.uk>

    commit f184d9ca
    Author: Simon Marlow <marlowsd@gmail.com>
    Date:   Mon Dec 10 12:00:54 2012 +0000

    Fix a bug in the handling of nested orElse

    Exposed by the following snippet, courtesy of Bas van Dijk and Patrick
    Palka on libraries@haskell.org:

    import Control.Concurrent.STM
    main = do
      x <- atomically $ do
             t <- newTVar 1
             writeTVar t 2
             ((readTVar t >> retry) `orElse` return ()) `orElse` return ()
             readTVar t
      print x

Without -O, we previously tried to make interface files smaller
by not including the data constructors of data types.  But
there are a lot of exceptions, notably when Template Haskell is
involved or, more recently, DataKinds.

However Trac #7445 shows that even without TemplateHaskell, using
the Data class and invoking Language.Haskell.TH.Quote.dataToExpQ
is enough to require us to expose the data constructors.

So I've given up on this "optimisation" -- it's probably not
important anyway.  Now I'm simply not attempting to trim off
the data constructors.  The gain in simplicity is worth the
modest cost in interface file growth, which is limited to the
bits reqd to describe those data constructors.

Conflicts:
	compiler/main/TidyPgm.lhs

Thanks to Richard Eisenberg for writing this.

This reverts the change in Trac #7347, which prevented promotion
of existential data constructors.  Ones with constraints in
their types, or kind polymorphism, still can't be promoted.

This patch fixes a quite egregious bug. runTcS was intialising
the 'untouchables' to NoUntouchables, which is quite wrong for
the invocation in TcSimplify.simplifyInfer.  Result: missed
skolem-escape check and seg-fault city.

All this is done differently, and better, in HEAD, so this
patch is needed only on the branch.

This is a kludge, and means that ghc/haddock won't be able to use
more than 2G of RAM. But it'll make sure that ghci works in the short
term while we work on a proper fix.

Trac #7410 pointed out a terrible error message, which is
much improved by this patch.

Conflicts:
	compiler/typecheck/TcHsType.lhs

Fixes Trac #7451. See Note [Kind Constraint and kind *] in Kind.lhs.

Conflicts:
	compiler/types/Kind.lhs
	compiler/types/Type.lhs

I don't konw how this was left out before; Trac #7347.

In fixing this I did the usual round of refactoring.  In particular, I
cached the fact that a DataCon can be promoted in the DataCon
itself (the dcPromoted field).

    commit 2c207b6f

    Author: Simon Peyton Jones <simonpj@microsoft.com>
    Date:   Tue Oct 2 18:11:08 2012 +0100

    Compare the kinds of type variables when comparing types

    This is just a bug that's been around since we introduced
    polymorphic kinds.  The roots are in Type.cmpTypeX, but
    there was a variant in TcRnDriver.checkBootTyCon, which
    is where it came up.

    Fixes Trac #7272

    commit 419af69c
    Author: Simon Peyton Jones <simonpj@microsoft.com>
    Date:   Tue Oct 2 17:39:46 2012 +0100

    Give PrimOps a NOINLINE pragma, to suppress silly warnings from dsRule

    See Trac #7287

Make sure that we check for type errors strictly in a Template Haskell splice, even if -fdefer-type-errors is on

We're going to run this code, so there's no point in deferring type
errors.  Worse, TcSplice uses unsafeCoerce (for vevy good reasons),
so splicing a type into a place where an expression is expected would
yield a seg-fault if we plough on regardless.

Fixes Trac #7276

Fixes Trac #7438

This bug caused Trac #7386, because in the (rather tricky) "type
inference" (aka run time type reconstruction) done by the GHCi
debugger, we were failing to instantiate a data type family
correctly.  When that code was written we didn't *have* data
families.

I wrote Note [Constructor arg types] to explain the new scheme.

Fixes Trac #7338

We should be lazy when type-checking the equality-contraint
part of a data constructor's type, to make the knot-tying
work out right.  The fact that it's always worked before
is a fluke: no one else wrote a GADT whose type index mentions
itself
  data T a wher
    MkT :: T (T Int)

The issue here was with a function type written prefix
  (->) a b
where we were not generating a FunTy, which blew the
invariant that function types are always FunTys.  We
can't look at the TyCon directly because it may be
knot-tied, so we look at the name instead.

Erik de Castro Lopo authored 12 years ago and ian@well-typed.com committed 12 years ago

Was getting an ocassional hang or segfault when building GHC in a
Qemu user space emulation of ARM. Turned out that the ITIMER_SIGNAL
was being delivered *after* the call to timer_delete(). Setting the
signal to SIG_IGN before deleting the timer solves the problem.

This is the right fix to Trac #7276 (part 2), which makes the
interaction with -fdefer-type-errors and TH work properly.

commit 9ef4d48e
Author: Ian Lynagh <ian@well-typed.com>
Date:   Fri Nov 23 16:54:15 2012 +0000

    Fix building on Sparc/Solaris with non-GNU linker; fixes #5407

    Patch from maeder.

Conflicts:

	rts/sm/Sanity.c

Simon Marlow authored 12 years ago and pcapriotti committed 12 years ago

MERGED from commit 598ee1ad

Based on:

  commit 5bae803a
  Author: Simon Peyton Jones <simonpj@microsoft.com>
  Date:   Sat Sep 15 23:09:25 2012 +0100

      Fix UNPACK with -fomit-interface-pragmas.

      We were missing a case, so that you could expose a constructor
      with UNPACKed fields, but the field tpye was trimmed, and hence
      can't be expanded.

      Fixes Trac #5252 (revived)