GitLab

Wibble in hi-boot file

Behave better in decode_escape failure

track 5.03's change to PrelHandle.fillReadBuffer

Complete previous tcAddImportedIdInfo commit

Improve error reporting

Comments only

--------------------------------------
	Finally get rid of tcAddImportedIdInfo
	--------------------------------------

TcEnv.tcAddImportedIdInfo is a notorious source of space leaks.
Simon M got rid of the need for it on default methods.
This commit gets rid of the need for it for dictionary function Ids,
and finally nukes the beast altogether. Hurrah!

The change really involves putting tcInterfaceSigs *before*
tcInstDecls1, so that any imported DFunIds are in the typechecker's
environment before we get to tcInstDecls.

Add missing files for Rank-N commit

----------------------
	Implement Rank-N types
	----------------------

This commit implements the full glory of Rank-N types, using
the Odersky/Laufer approach described in their paper
	"Putting type annotations to work"

In fact, I've had to adapt their approach to deal with the
full glory of Haskell (including pattern matching, and the
scoped-type-variable extension).  However, the result is:

* There is no restriction to rank-2 types.  You can nest forall's
  as deep as you like in a type.  For example, you can write a type
  like
	p :: ((forall a. Eq a => a->a) -> Int) -> Int
  This is a rank-3 type, illegal in GHC 5.02

* When matching types, GHC uses the cunning Odersky/Laufer coercion
  rules.  For example, suppose we have
	q :: (forall c. Ord c => c->c) -> Int
  Then, is this well typed?
	x :: Int
	x = p q
  Yes, it is, but GHC has to generate the right coercion.  Here's
  what it looks like with all the big lambdas and dictionaries put in:

	x = p (\ f :: (forall a. Eq a => a->a) ->
		 q (/\c \d::Ord c -> f c (eqFromOrd d)))

  where eqFromOrd selects the Eq superclass dictionary from the Ord
  dicationary:		eqFromOrd :: Ord a -> Eq a


* You can use polymorphic types in pattern type signatures.  For
  example:

	f (g :: forall a. a->a) = (g 'c', g True)

  (Previously, pattern type signatures had to be monotypes.)

* The basic rule for using rank-N types is that you must specify
  a type signature for every binder that you want to have a type
  scheme (as opposed to a plain monotype) as its type.

  However, you don't need to give the type signature on the
  binder (as I did above in the defn for f).  You can give it
  in a separate type signature, thus:

	f :: (forall a. a->a) -> (Char,Bool)
	f g = (g 'c', g True)

  GHC will push the external type signature inwards, and use
  that information to decorate the binders as it comes across them.
  I don't have a *precise* specification of this process, but I
  think it is obvious enough in practice.

* In a type synonym you can use rank-N types too.  For example,
  you can write

	type IdFun = forall a. a->a

	f :: IdFun -> (Char,Bool)
	f g = (g 'c', g True)

  As always, type synonyms must always occur saturated; GHC
  expands them before it does anything else.  (Still, GHC goes
  to some trouble to keep them unexpanded in error message.)


The main plan is as before.  The main typechecker for expressions,
tcExpr, takes an "expected type" as its argument.  This greatly
improves error messages.  The new feature is that when this
"expected type" (going down) meets an "actual type" (coming up)
we use the new subsumption function
	TcUnify.tcSub
which checks that the actual type can be coerced into the
expected type (and produces a coercion function to demonstrate).

The main new chunk of code is TcUnify.tcSub.  The unifier itself
is unchanged, but it has moved from TcMType into TcUnify.  Also
checkSigTyVars has moved from TcMonoType into TcUnify.
Result: the new module, TcUnify, contains all stuff relevant
to subsumption and unification.

Unfortunately, there is now an inevitable loop between TcUnify
and TcSimplify, but that's just too bad (a simple TcUnify.hi-boot
file).


All of this doesn't come entirely for free.  Here's the typechecker
line count (INCLUDING comments)
	Before	16,551
	After	17,116

unbreak tcModule.

Use (DefMeth Name) rather than (DefMeth Id) in ClassOpItem.  This not
only eliminates a space leak, because Names generally hold on to much
less stuff than Ids, but also turns out to be a minor cleanup.

Replace a lazy pattern match in tcGetInstLoc with a strict one (fixes
a space leak).

Adorn various constructor fields with strictness annotations - this
fixes at least one space leak, in the mi_usages field of a ModIface.

Collect up _scc_ expressions on the right hand side of a closure
definition and attach them directly to the closure using PushCC-type
cost centres, so that the allocation of the closure gets attributed to
the right place.

Fix a long-standing bug in the cost attribution of cost-center stacks.
The problem case is this:

	let z = _scc_ "z" f x
	in ... z ...

previously we were attributing the cost of allocating the closure 'z'
to the enclosing cost center stack (CCCS), when it should really be
attributed to "z":CCCS.  The effects are particularly visible with
retainer profiling, because the closure retaining 'f' and 'x' would
show up with the wrong CCS attached.

To fix this, we need a new form of CCS representation internally:
'PushCC CostCentre CostCentreStack' which subsumes (and therefore
replaces) SingletonCCS.  SingletonCCS is now represented by 'PushCC cc
NoCCS'.

The CCS argument to SET_HDR may now be an arbitrary expression, such
as PushCostCentre(CCCS,foo_cc), as may be the argument to CCS_ALLOC().
So we combine SET_HDR and CCS_ALLOC into a single macro, SET_HDR_, to
avoid repeated calls to PushCostCentre().

Call LDV_ENTER() on entry to a constructor in profiling mode.

Call LDV_ENTER() on entry to a thunk or function in profiling mode.

Add
	ldvEnter :: Code

for LDV profiling: it records that the closure pointed to by R1 has
just been entered.

We don't need to consider Hp as a volatile register across C calls; it
is already saved by the CALLER_SAVES_SYSTEM macro.

Inlining pragma buglet

Nuke #include of MachDeps.h - nothing in there that cannot be
(better) gotten from the in-tree config.h

This is the only (last?) use of MachDeps.h, so if anyone won't argue
for its continued existence, I'll nuke it sometime soon.

- change the interpretation of FPTOOLS_TOP_ABS, it is now the UNIXy path.
  For Win32, it is of the form <drive>:<path> where '/' is the directory
  separator. Prior to this commit, the directory separator was '\'.
- for the (only) place we need to have a platform-native representation
  of FPTOOLS_TOP_ABS, use FPTOOLS_TOP_ABS_PLATFORM.

Tidy up imports

Improve error msg

--------------------------------------
	Yet another cut at the DmdAnal domains
	--------------------------------------

This version of the domain for demand analysis was developed
in discussion with Peter Sestoft, so I think it might at last
be more or less right!

Our idea is mentally to separate
	strictness analysis
from
	absence and boxity analysis

Then we combine them back into a single domain.  The latter
is all you see in the compiler (the Demand type, as before)
but we understand it better now.

---------------------------------------
	Add continuation splitting to Simplify
	---------------------------------------

When the simplifier finds a 'case', it calls mkDupableAlt
to make the "continuation" (that is, the context of the
case expression) duplicatable, so that it can push it into
the case branches.  This is crucial for the case-of-case
transformation.

But it turns out that it's a bad idea to do that when
the context is "I'm the argument of a strict function".  Consider

	f (case x of { True -> False; False -> True }) arg2

where f is a strict function.  Then we *could* (and were)
transforming to

	let $j a = f a arg2
	in
	case x of { True -> $j False; False -> $j True }

But this is in general a terribly bad thing to do.
See the example in comments with Simplify.mkDupableCont.

Remove pprTrace

GhcCanonVersion: leave out GhcPatchLevel, causes problems when
the GhcPatchLevel isn't a patchlevel but a date.

lookupInstEnv: "#ifdef DEBUG"-protect trace code
that got included in prev. commit.

(I'm guessing it either should be removed or moved
into the ASSERT, but I'll let the original committer
decide which).

* Moved CANON_HC_VERSION out of ghc/compiler/Makefile and into mk/config.mk.in
  (and renamed it as GhcCanonVersion).
* Have ghc/driver/Makefile use it; cheaper and more robust than the version
  testing it was already doing.

Optimised defn of CANON_HC_VERSION, synthesise it from config.mk info,
not via an expensive $(shell ...) call.

Use -fvia-C when compiling rename/ParseIface.hs and parser/Parser.hs -
Int16 primops aren't supported by the NCG on all plats.

---------------------------------------
	Fix an obscure overlapping-instance bug
	---------------------------------------

	MERGE TO STABLE BRANCH

When searching for instances, we used bale out if the type
we seek could be instantiated to match the instance (because
it might be so instantiated later, in which case we don't
want to miss the opportunity).

The bug was that we used *matching* whereas we should use
*unification*.

Comments in the file InstEnv.

gen_Eq_binds: when comparing constructor tags, emit just

   a == b = case con2tag_Foo# a of
              a# -> case con2tag_Foo# b of b# -> a# PrelGHC.==# b#

and not

   a == b = case con2tag_Foo# a of
              a# -> case con2tag_Foo# b of
                      b# -> if a# PrelGHC.==# b# then PrelBase.True else PrelBase.False

(Not that this wouldn't get simplified, but still).

rnHsForeignDecl: 'foreign import's (incl 'f.e.d's) _define_ local toplevel
names, so better use RnEnv.lookupTopBndrRn and not RnEnv.lookupOccRn to
resolve the name.

As was, declaring ForeignImports with the same name as an imported entity
wasn't permitted.

gencode: for completeness sake only, handle CCallTypedefs

ghc-inplace: single-quote -B option to avoid de-escaping those b-slashes

Updates to the native code generator following the changes to fix the
large block allocation bug, and changes to use the new
function-address cache in the register table to reduce code size.

Also: I changed the pretty-printing machinery for assembly code to use
Pretty rather than Outputable, since we don't make use of the styles
and it should improve performance.  Perhaps the same should be done
for abstract C.

Remove the heap-check-size panic, following the RTS fixes for this problem.

When generating dependencies, look for both source files _and_
interface files.

If STABLE is still a branch with a future, I'd encourage merging.