GitLab

As Trac #11222, and #10712 note, the strictness analyser
needs to be rather careful about exceptions.  Previously
it treated them as identical to divergence, but that
won't quite do.

See Note [Exceptions and strictness] in Demand, which
explains the deal.

Getting more strictness in 'catch' and friends is a
very good thing.  Here is the nofib summary, keeping
only the big ones.

--------------------------------------------------------------------------------
        Program           Size    Allocs   Runtime   Elapsed  TotalMem
--------------------------------------------------------------------------------
          fasta          -0.1%     -6.9%     -3.0%     -3.0%     +0.0%
            hpg          -0.1%     -2.0%     -6.2%     -6.2%     +0.0%
       maillist          -0.1%     -0.3%      0.08      0.09     +1.2%
reverse-complem          -0.1%    -10.9%     -6.0%     -5.9%     +0.0%
         sphere          -0.1%     -4.3%      0.08      0.08     +0.0%
           x2n1          -0.1%     -0.0%      0.00      0.00     +0.0%
--------------------------------------------------------------------------------
            Min          -0.2%    -10.9%    -17.4%    -17.3%     +0.0%
            Max          -0.0%     +0.0%     +4.3%     +4.4%     +1.2%
 Geometric Mean          -0.1%     -0.3%     -2.9%     -3.0%     +0.0%

On the way I did quite a bit of refactoring in Demand.hs

This change tidies up and simplifies (a bit) the knot-tying
when kind-checking groups of type and class declarations.

The trouble (shown by Trac #11356) was that we wanted an error message
(a kind-mismatch) that involved a type mentioned a (AThing k), which
blew up.

Since we now seem to have TcTyCons, I decided to use them here.
It's still not great, but it's easier to understand and more robust.

Trac #11437 showed that erroneous constraints from a 'deriving'
clause need to be wrapped in an Implication to properly scope
their skolems.

The main change is in TcDeriv.simplifyDeriv; the call to
buildImplicationFor is new.

simpl_top was being polluted with Safe Haskell stuff which was only
used in one of its four calls.  This moves the Safe Haskell stuff
to the place it is actually used

It was only called in one place; easier to inline it

Previously tidySkolemInfo used tidyOpenType, and returned a new
TidyEnv.  But that's not needed any more, because all the skolems
should be in scope in the constraint tree.

I also removed a (now-unnecessary) field of UnifyForAllSkol

Incredibly enough this hasn't been touched since 6.10.1

Reviewers: austin, hvr, bgamari, thomie

Reviewed By: thomie

Differential Revision: https://phabricator.haskell.org/D1721

Reviewers: hvr, thomie, austin

Reviewed By: austin

Subscribers: duncan

Differential Revision: https://phabricator.haskell.org/D1741

GHC Trac Issues: #8176, #4437

Bump Cabal and Haddock submodules such that they both support GCC-style
response files on Windows.

Previous representation of GADTs in TH was not expressive enough
to express possible GADT return types. See #11341

Test Plan: ./validate

Reviewers: goldfire, austin, bgamari

Subscribers: thomie, RyanGlScott

Differential Revision: https://phabricator.haskell.org/D1738

GHC Trac Issues: #11341

Phab:D493 accidentally changed the way GHC generics looks up `Fixity`
information when deriving `Generic` or `Generic1`. Before, a `Fixity` of
`Infix` would be given only if a data constructor was declared infix,
but now, `Infix` is given to any data constructor that has a fixity
declaration (not to be confused with being declared infix!). This commit
reverts back to the original behavior for consistency's sake.

Fixes #11358.

Test Plan: ./validate

Reviewers: kosmikus, dreixel, austin, bgamari

Reviewed By: bgamari

Subscribers: thomie

Differential Revision: https://phabricator.haskell.org/D1740

GHC Trac Issues: #11358

A set of changes to enable local ghc env files to be useful for tools
like cabal. Ultimately it will allow cabal to maintain a ghc env file so
that users can simple run ghc or ghci in a project directory and get the
expected environment of the project.

Change the name of .ghc.environment files to include the platform and
ghc version, e.g. .ghc.environment.x86_64-linux-7.6.3, since their
content is version specific. Strictly speaking this is not backwards
compatible, but we think this feature is not widely used yet.

"Look up" for a local env file, like the behaviour of git/darcs etc. So
you can be anywhere within a project and get the expected environment.

Don't look for local env files when -hide-all-packages is given.

Extend the syntax of env files to allow specifying package dbs too.

Test Plan:
Currently completely untested. Compiles, that is all.
Sorry, have to disappear for the hols.

Reviewers: hvr, ezyang, austin, bgamari

Reviewed By: ezyang, bgamari

Subscribers: thomie

Differential Revision: https://phabricator.haskell.org/D1668

GHC Trac Issues: #11268

The interfaces to {save,load}ThreadState were quite messy due to the
need to pass in local registers (produced with draws from a unique
supply) since they were used from both FCode and UniqSM.

This, however, is entirely unnecessary as we already have an
abstraction to capture this effect: MonadUnique. Use it.

This is part of an effort to properly represent stack unwinding
information
for foreign calls.

Test Plan: validate

Reviewers: austin, simonmar

Reviewed By: simonmar

Subscribers: thomie

Differential Revision: https://phabricator.haskell.org/D1733

Various people (myself included) have complained about the lack of
useful descriptions for the various packages included in GHC's source
tree. Fix this.

Test Plan: Validate

Reviewers: austin, thomie

Reviewed By: thomie

Subscribers: angerman, ezyang

Differential Revision: https://phabricator.haskell.org/D1736

Since #11316 indicates that having flag `-Wtoo-many-guards`
enabled by default causes issues, the simplest thing is to
remove it. This patch removes it from the default list, it
updates the docs and removes the suppression flags for
`T783` and `types/OptCoercion.hs`

Test Plan: validate

Reviewers: bgamari, austin, goldfire

Subscribers: thomie

Differential Revision: https://phabricator.haskell.org/D1737

GHC Trac Issues: #11316

Unfortunately, I could not add the expected error message, so if someone
accidentally fixes this bug, this test will still be failing (no harm).
But maybe someone stumbles over it then and can update the expected
output.

When translating a `CoPat` to `PmPat` check whether the wrapper
is just a hole or a cast with refl. In these cases we can safely
drop the wrapper and generate less guard patterns. Fixes T11276.

Test Plan: validate

Reviewers: bgamari, austin

Subscribers: thomie

Differential Revision: https://phabricator.haskell.org/D1729

GHC Trac Issues: #11276

On Windows, with msys2, `pwd` works (as can be seen by the use of `pwd`
that slipped into the validate script), so there is really no need for
`ghc-pwd` anymore.

Test Plan: try it

Reviewers: austin, bgamari, Phyx

Reviewed By: Phyx

Subscribers: Phyx, erikd

Differential Revision: https://phabricator.haskell.org/D1731

Note only does this make the markup cleaner and more semantic, but it
clears up some warnings and enables proper hyperlinks. Yay Sphinx!

While I was at it I generally cleaned up the markup

Test Plan: Read it

Reviewers: hvr, thomie, austin

Differential Revision: https://phabricator.haskell.org/D1730

Test Plan: Read it

Reviewers: austin, hvr, #core_libraries_committee

Subscribers: thomie

Differential Revision: https://phabricator.haskell.org/D1720

GHC Trac Issues: #11314

Otherwise Clang seems to complain about unused static definitions.

Since we produce ARM veneer.

ARM has a 24-bit relative jump offset just like PowerPC. Ensure that
symbols don't get too far from their symbol extras.

See #11340.

Differential Revision: https://phabricator.haskell.org/D1728