GitLab

This patch executes the plan described in the discussion in Trac #2497.
Specficially:

    * Inside a RULE, switch on the forall-as-keyword in the lexer,
      unconditionally. (Actually this is done by an earlier patch.)

    * Merge the -XScopedTypeVariables and -XPatternSignatures flags,
      and deprecate the latter. Distinguishing them isn't senseless,
      but it's jolly confusing.

    * Inside a RULE, switch on -XScopedTypeVariables unconditionally. 

    * Change -frewrite-rules to -fenable-rewrite-rules; deprecate the former. 
      Internally the DynFlag is now Opt_EnableRewriteRules.

There's a test in typecheck/should_compile/T2497.hs

It was replaced by -fwarn-warnings-deprecations, but I think we want
to keep it for backwards compatibility.  I'm not sure we want to
deprecate it either...

Now you can say
  
  import "network" Network.Socket

and get Network.Socket from package "network", even if there are
multiple Network.Socket modules in scope from different packages
and/or the current package.

This is not really intended for general use, it's mainly so that we
can build backwards-compatible versions of packages, where we need to
be able to do

module GHC.Base (module New.GHC.Base) where
import "base" GHC.Base as New.GHC.Base

The flag is off by default, but GHCi turns it on (in Main.hs).  For
GHCi it can be overriden on the command-line or using :set.

They are now handled by the main flag parser, rather than having their
own praser that runs later.

As an added bonus, 5 global variables are also gone.

This is all a bit of a mess, but can hopefully be improved when we get
encoding/decoding support in Handles.

-fglasgow-exts also turns it on.

We shouldn't need it, as we don't call cygwin's gcc, and it was causing
problems with the nightly builders passing it to GHC.

From the entry in the User's guide:

-fwarn-dodgy-foreign-imports causes a warning to be emitted for
foreign imports of the following form:

foreign import "f" f :: FunPtr t

on the grounds that it probably should be

foreign import "&f" f :: FunPtr t

The first form declares that `f` is a (pure) C function that takes no
arguments and returns a pointer to a C function with type `t`, whereas
the second form declares that `f` itself is a C function with type
`t`.  The first declaration is usually a mistake, and one that is hard
to debug because it results in a crash, hence this warning.

It now also doesn't automatically link base and rts either.
We need this when we've done a build, so base and rts are in the
package.conf, but we've then cleaned the libraries so they don't
physically exist any more.

It's -fdph-seq and -fdph-par at the moment, I'll think of a nicer setup later.

  This is a fairly substantial rewrite of the Static Argument Transformatoin,
  done by Max Bolingbroke and reviewed and modified by Simon PJ.
  
  * Fix a subtle scoping problem; see Note [Binder type capture]
  * Redo the analysis to use environments
  * Run gentle simlification just before the transformation

* Allow -ffoo flags to be deprecated
* Mark some -ffoo flags as deprecated
* Avoid using deprecated flags in error messages, in the build system, etc
* Add a flag to en/disable the deprecated flag warning

We might want to put the values initSysTools finds in their own type,
rather than having them flattened into DynFlags

From the docs:
   <para>Suppress any unsolicited debugging output.  When GHC
     has been built with the <literal>DEBUG</literal> option it
     occasionally emits debug output of interest to developers.
     The extra output can confuse the testing framework and
     cause bogus test failures, so this flag is provided to
     turn it off.</para>

This is the optimisation level recommended when compiling DPH programs. At the
moment, it is equivalent to -O2 -fno-method-sharing -fdicts-cheap
-fmax-simplifier-iterations20 -fno-spec-constr-threshold.

We want -Odph to be a dynamic flag and that should imply -fno-method-sharing.
This doesn't add a lot of complexity.

This patch revives the Static Argument Transformation, thanks to
Max Bolingbroke.  It is enabled with 
	-fstatic-argument-transformation
or	-O2

Headline nofib results

                  Size    Allocs   Runtime
Min             +0.0%    -13.7%    -21.4%
Max             +0.1%     +0.0%     +5.4%
Geometric Mean  +0.0%     -0.2%     -6.9%

Modules that need it import it themselves instead.

This patch makes a significant improvement to SpecConstr, based on
Roman's experience with using it for stream fusion.  The main change is
this:

  * For local (not-top-level) declarations, seed the specialisation 
    loop from the calls in the body of the 'let'.

See Note [Local recursive groups] for discussion and example.  Top-level
declarations are treated just as before.

Other changes in this patch:

  * New flag -fspec-constr-count=N sets the maximum number of specialisations
    for any single function to N.  -fno-spec-constr-count removes the limit.

  * Refactoring in specLoop and friends; new algebraic data types 
    OneSpec and SpecInfo instead of the tuples that were there before

  * Be less keen to specialise on variables that are simply in scope.
    Example
      f p q = letrec g a y = ...g....  in g q p
    We probably do not want to specialise 'g' for calls with exactly
    the arguments 'q' and 'p', since we know nothing about them.