GitLab

This is the result of Bernie Pope's internship work at MSR Cambridge,
with some subsequent improvements by me.  The main plan was to

 (a) Reduce the overhead for breakpoints, so we could enable 
     the feature by default without incurrent a significant penalty
 (b) Scatter more breakpoint sites throughout the code

Currently we can set a breakpoint on almost any subexpression, and the
overhead is around 1.5x slower than normal GHCi.  I hope to be able to
get this down further and/or allow breakpoints to be turned off.

This patch also fixes up :print following the recent changes to
constructor info tables.  (most of the :print tests now pass)

We now support single-stepping, which just enables all breakpoints.

  :step <expr>     executes <expr> with single-stepping turned on
  :step            single-steps from the current breakpoint

The mechanism is quite different to the previous implementation.  We
share code with the HPC (haskell program coverage) implementation now.
The coverage pass annotates source code with "tick" locations which
are tracked by the coverage tool.  In GHCi, each "tick" becomes a
potential breakpoint location.

Previously breakpoints were compiled into code that magically invoked
a nested instance of GHCi.  Now, a breakpoint causes the current
thread to block and control is returned to GHCi.

See the wiki page for more details and the current ToDo list:

  http://hackage.haskell.org/trac/ghc/wiki/NewGhciDebugger

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.

  ** MERGE into 6.6.1 **

	MERGE to 6.6.1

For class-ops, record selectors, data constructors, we want the ru_local
field of the Rule to be False.  We do not attach the rule to the binding
for the Id, because there simply isn't a binding until the code gen stage.

(NB: the ru_local field is different to the orphan-hood of the rule.)

This fixes a bug that meant that RULES on class ops were never exported.

Merge to 6.6 branch.  Test case is dsrun014.

Note [Desugaring seq]  cf Trac #1031
~~~~~~~~~~~~~~~~~~~~~
   f x y = x `seq` (y `seq` (# x,y #))

The [CoreSyn let/app invariant] means that, other things being equal, because 
the argument to the outer 'seq' has an unlifted type, we'll use call-by-value thus:

   f x y = case (y `seq` (# x,y #)) of v -> x `seq` v

But that is bad for two reasons: 
  (a) we now evaluate y before x, and 
  (b) we can't bind v to an unboxed pair

Seq is very, very special!  So we recognise it right here, and desugar to
	case x of _ -> case y of _ -> (# x,y #)

The special case would be valid for all calls to 'seq', but it's only *necessary*
for ones whose second argument has an unlifted type. So we only catch the latter
case here, to avoid unnecessary tests.

My previous patch was creating problems with hs-boot files. 

This patch performs several optimizations with the goal of minimizing the cost of building the arguments to breakpointJump:
  - Group them all in a single tuple, to minimize closure creation in heap
  - Wrap this with the GHC.Base.lazy combinator, to induce max laziness
  - Remove as many literal strings as possible 
    * injecting a module-local CAF to store the module name and use that 
    * eliminating the package string (not needed).

Conal Eliott (Trac #1145) exposed a nasty flaw in the way in which
orphan instances are computed, when there are functional dependencies
in the class.  It took me some time to figure out what was going on,
and led to more refactoring.

Briefly:

* Elaborate comments about orphan-hood and versioning added to IfaceSyn
* The is_orph field vanishes from InstEnv.Instance
* Similarly ru_orph vanishes from CoreSyn.CoreRule
* Orphan-hood is computed in MkIface.instanceToIfaceInst, and
	MkIface.coreRuleToIfaceRule

Elsewhere just tidying up.

Benchmarks have shown that making the formation of the list of locals more lazy can improve performance of -fdebugging up to 50% in some cases

This patch clears up a long-standing wart.   For some time it's been the
case that 
	the RHS of a non-recursive let can be unlifed iff 
	the RHS is ok-for-speculation

This patch extends the invariant to the argument of an App, and 
establishes it by the smart constructors mkDsApp, mkDsApps in the desugarer.

Once established, it should be maintained by the optimiser.

This tides up some awkward cases, notably in exprIsHNF, and I think it
fixes a outright strictness bug in Simplify.prepareRhs.

By popular request, in a breakpoint it is possible now to inspect the result of the expression wrapped by the breakpoint.

The user interface for this is right now preliminar; there is a new binding called '_result' at every breakpoint. Suggestions are welcome!

  I have added a check, and while there removed a few kludges in my code.
  Kudos to -dcore-lint for uncovering this.
 
  I think that this restriction could be lifted, if GHC.Base.breakpoint could have kind ?? -> ??. But is this a legal type? Does not look so to me.

The class is named IsString with the single method fromString.
Overloaded strings work the same way as overloaded numeric literals.
In expressions a string literals gets a fromString applied to it.
In a pattern there will be an equality comparison with the fromString:ed literal.

Use -foverloaded-strings to enable this extension.
 

Consider this FC program:
	data family AT a :: *
	data instance AT Int = T1 Int Int

	f :: AT Int -> Int
	f t = case t of DEFAULT -> <body>

We'd like to replace the DEFAULT by a use of T1, so that if 
we scrutinise t inside <body> we share the evaluation:

	f t = case (t `cast` co) of T1 x y -> <body>

I decided to do this as part of the liberate-case transformation,
which is already trying to avoid redundant evals.  

The new transformation requires knowledge of the family instance
environment, so I had to extend ModGuts to carry the fam_inst_env,
and put that envt into the liberate-case environment. 

Otherwise it's all pretty straightforward.

Adding a {-# GENERATED "SourceFile" SourceSpan #-} <expr> pragma.
This will be used to generate coverage for tool generated (or quoted) code.
The pragma states the the expression was generated/quoted from the stated
source file and source span.

- With -findexed-types, equational constraints can appear in contexts 
  wherever class predicates are allowed.
- The two argument types need to be boxed and rank 0.

Hopefully this will help to restore ability to build HEAD on GHC 5.0x systems

Specifically, this disables the special support in the RTS for looking up the datacon name corresponding to an address. 
Correspondingly, the debugging commads in GHCi will not be available, and neither will the '-fdebugging' flag

Used in the desugaring of the breakpoint primitive

Instrumentation gets activated by the '-fdebugging' dynflag.

All the instrumentation occurrs in the desugarer; it consists of inserting 'breakpoint' combinators at a number of places in the AST, namely: 
 - Binding sites
 - Do-notation statements 
These 'breakpoint' combinators will later be further desugared (at DsExpr) into ___Jump functions.
For more info about this and all the ghci.debugger see the page at the GHC wiki:

http://hackage.haskell.org/trac/ghc/wiki/GhciDebugger

This changes the internal representation of TickBoxes,
from
        Note (TickBox "module" n)  <expr>
into

        case tick<module,n> of
          _ -> <expr>

tick has type :: #State #World, when the module and tick numbe
are stored inside IdInfo.

Binary tick boxes change from

         Note (BinaryTickBox "module" t f) <expr>

into

          btick<module,t,f> <expr>

btick has type :: Bool -> Bool, with the module and tick number
stored inside IdInfo.