GitLab

This references #8121Signed-off-by: Austin Seipp <aseipp@pobox.com>

This is a test commit for addressing #8121

A simple heuristic (i.e. replacing the `libraries/` suffix by
`packages/` in the inferred remotepath, if the remotepath in the
`packages` file is "-") used to infer the remotepath for Git
submodules. A proper solution would require to parse the `.gitmodules`
file to obtain the actual localpath<->remotepath mapping.

This addresses #8121 and is a follow-up to a22e908a which changed the
`.gitmodules` urls to use the `/packages/` folder instead of
`/libraries/`.

This changes the `/libraries/<name>.git/` suffix in the urls
to `/packages/<name>.git`. This is a minor preparation step
helping the Gitolite transition (see #8121).

See also http://permalink.gmane.org/gmane.comp.lang.haskell.ghc.devel/1713

My original change to the calling convention mistakenly used all 6 XMM
registers---which live in the global register table---on x86 (32 bit). This
royally screwed up the floating point code generated for that platform because
floating point arguments were passed in global registers instead of on the
stack!

We were using SSE is some places and XMM in others. Better to keep a consistent
naming scheme.

As documented in the users' guide, you can now write

type family Foo a where ..

in a hs-boot file to declare an abstract closed type family.

Dynamic libraries on powerpc linux are currently broken.
See: http://hackage.haskell.org/trac/ghc/ticket/8024

This check is somewhat subtle. See Note [Checking GADT return types]
in TcTyClsDecls. The new plan is to check *before* desugaring the type
from HsType to Type. This avoids problems with the pattern-match
in rejigConRes.

As a nice side benefit to this, I discovered that Template Haskell
splices were a little conservative in their treatment of valid data
constructors. (For example, a kind signature in the return type caused
failure.) Now, the TH code uses exactly the same function as the
"real" code, which is nice. See hsTyGetAppHead_maybe in HsTypes.

Roles are a solution to the GeneralizedNewtypeDeriving type-safety
problem.

Roles were first described in the "Generative type abstraction" paper,
by Stephanie Weirich, Dimitrios Vytiniotis, Simon PJ, and Steve Zdancewic.
The implementation is a little different than that paper. For a quick
primer, check out Note [Roles] in Coercion. Also see
http://ghc.haskell.org/trac/ghc/wiki/Roles
and
http://ghc.haskell.org/trac/ghc/wiki/RolesImplementation
For a more formal treatment, check out docs/core-spec/core-spec.pdf.

This fixes Trac #1496, #4846, #7148.

The solution is to use a different notion of apartness. See
http://research.microsoft.com/en-us/um/people/simonpj/papers/ext-f/axioms-extended.pdf
for the gory details. Some comments are also in Notes [Compatibility]
and [Apartness] in FamInstEnv.

We now give a warning about enumerations like [5 .. 3] :: Int8.

I'd been meaning to do this for some time, but finally got around to it
due to the overflowing literals warning. With that enabled, we were
getting a warning for
    -128 :: Int8
as that is parsed as
    negate (fromInteger 128)
which just happens to do the right thing, as
    negate (fromInteger 128) = negate (-128) = -128

This patch adds rules for converting floating point multiplication
of the form 2.0 * x and x * 2.0 into addition x + x.

It thought that .cmm files were Haskellish, but then didn't know
how to compile them. Now they aren't treated as Haskellish, and
it can compile and link them.

Makes it look less likely that people will confuse what it is for
(e.g. #8104).

This means that ghc-tarballs is only needed on Windows

We need to desugar the code, or we don't get the warnings from the
desugarer.

No functional changes

Signed-off-by: Austin Seipp <aseipp@pobox.com>