GitLab

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>

Unfortunately, I couldn't remove all incidences of runCmmContFlowOpt
from HscMain; in particular, there is a Cmm conversion testing
facility which may run with only control flow optimizations, which
I preserved the semantics of.  Given the state of the current
codegen, this code might be moribund anyway.
Signed-off-by: Edward Z. Yang <ezyang@mit.edu>

  Module imports form a cycle:
    module `Foo4' imports `Foo'
            which imports `Foo2'
            which imports `Foo3'
            which imports `Foo4'

as requested by Bryan Richter

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>

It doesn't make sense to specialise class methods -- their
code is simply a record selector.  Moreover, such methods
don't (currently) have an inl_sat field in their InlinePragma
which made DsBinds.makeCoreArity bleat.

This patch just ignores such pragmas, with a warning.

The following patches were ported:

    d0faaa6f Fix segfault in array copy primops on 32-bit
    18691d44 Make assignTemp_ less pessimistic
    9c23f06f Make array copy primops inline
Signed-off-by: Edward Z. Yang <ezyang@mit.edu>

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>

Specifically
 - Turn the error
       "You cannot SPECIALISE f because it is not INLINEABLE"
   into a warning

 - More importantly, suppress it altogether when
     HscTarget = HscNothing or HscInterpreted
   because then we aren't going to be generating unfoldings

The change to the packages file format broke it

We were getting
    InstallExtraPackages="YES    "

We no longer pull directly from upstream repos, so it wasn't useful
to have it programatically available. The info, and more, is now on
    http://hackage.haskell.org/trac/ghc/wiki/Commentary/Libraries#Repositorylocations
instead.

Template Haskell doesn't support GADTs directly but
we can use equality constraints to do the job. Here's
an example of the dump from splicing such a declaration:

    [d| data T a b
            where
              T1 :: Int -> T Int Char
              T2 :: a -> T a a
              T3 :: a -> T [a] a
              T4 :: a -> b -> T b [a] |]
  ======>
    T5217.hs:(6,3)-(9,53)
    data T a[aQW] b[aQX]
        = (b[aQX] ~ Char, a[aQW] ~ Int) => T1 Int |
          b[aQX] ~ a[aQW] => T2 a[aQW] |
          a[aQW] ~ [b[aQX]] => T3 b[aQX] |
          forall a[aQY]. b[aQX] ~ [a[aQY]] => T4 a[aQY] a[aQW]

The point here is that if a data type chooses a representation that
unpacks an argument field, the representation of the argument field
must be visible to clients.  And it may not be if OmitInterfacePragmas
is on.

Otherwise Haddock (which compiles stuff without -O) falls over

The issue here was: what import declaration brings into
scope the 'op here

   import qualified Foo( op )
   import Bar( C(op) )
   instance C Int where
     op = ...

Well, the import of Bar, obviously.  But what if the
import Bar had been
   import Bar( C )
Then the instance is still supposed to work, getting
op from the Foo.op imported from Foo.  (I'm assuming its
the same op, of course.)

We have to do this until the earliest bootstrap compiler
has fixed #5252.  It's no big deal; it just means that
SrcLoc.hi has lots of stuff even when compiling GHC without -O.
Lacking this, you get

  (GHC version 7.0.3 for i386-unknown-mingw32):
	reboxProduct: not a product ghc-7.1:SrcLoc.RealSrcLoc{tc r8J}

The general rule is now that we don't complain about
a duplicate export from a "...".  See RnNames.dupExport_ok

The new monad comprehension stuff introduce LastStmt,
and DsMeta must convert that to TH syntax

...all to do with the new codgen path

It was only working when followed by something, e.g. "$topdir/base".

Take vectorisation declarations into account during the initial occurrence analysis (right after desugaring).

The "Unhelpful" cases are now in a separate type. This allows us to
improve various things, e.g.:
* Most of the panic's in SrcLoc are now gone
* The Lexer now works with RealSrcSpans rather than SrcSpans, i.e. it
  knows that it has real locations and thus can assume that the line
  number etc really exists
* Some of the more suspicious cases are no longer necessary, e.g.
  we no longer need this case in advanceSrcLoc:
      advanceSrcLoc loc _ = loc -- Better than nothing

More improvements can probably be made, e.g. tick locations can
probably use RealSrcSpans too.