GitLab

All the work was done by Dan Winograd-Cort.

The main thing is that arrow comamnds now have their own
data type HsCmd (defined in HsExpr).  Previously it was
punned with the HsExpr type, which was jolly confusing,
and made it hard to do anything arrow-specific.

To make this work, we now parameterise
  * MatchGroup
  * Match
  * GRHSs, GRHS
  * StmtLR and friends
over the "body", that is the kind of thing they
enclose.  This "body" parameter can be instantiated to
either LHsExpr or LHsCmd respectively.

Everything else is really a knock-on effect; there should
be no change (yet!) in behaviour.  But it should be a sounder
basis for fixing bugs.

See Trac #7287

Make sure that we check for type errors strictly in a Template Haskell splice, even if -fdefer-type-errors is on

We're going to run this code, so there's no point in deferring type
errors.  Worse, TcSplice uses unsafeCoerce (for vevy good reasons),
so splicing a type into a place where an expression is expected would
yield a seg-fault if we plough on regardless.

Fixes Trac #7276

This is just a bug that's been around since we introduced
polymorphic kinds.  The roots are in Type.cmpTypeX, but
there was a variant in TcRnDriver.checkBootTyCon, which
is where it came up.

Fixes Trac #7272

* Get rid of the lookupInInerts stage

* Re-introduce the flat-cache for flattening type-family equations
  See Note [Type family equations] in TcSMonad. My previous clever attempt
  with organising the work list proved too fragile.

  There's a (static) flag -fno-flat-cache to switch if off,
  so you can try with and without

* Improve the -ddump-cs-trace output

* The usual round of refactoring

Signed-off-by: Gabor Greif <ggreif@gmail.com>

The constraint solver sometimes goes to a lot of effort that turns
out to be Refl in the end, and avoiding zonking those proofs is a
useful optimisation. (Trac #5030)

These variables all have uniformative names like "cobox" or "d".
Make them System Names tells the pretty printer to show their
uniques even without -dppr-debug.  That in turn makes the
less-voluminous debug traces *much* more informative.

This fixes Trac #7278

* Treat kind-equality constraints as *derived* equalities,
  with no evidence.  That is really what they are at the moment.

* Get rid of EvKindCast and friends.

* Postpone kind errors properly to the constraint solver
  (lots of small knock-on effects)

I moved SwapFlag to BasicTypes as well

Rather than having a separate
    foo_INSTALL_SHELL_WRAPPER
variable, we just use
    foo_INSTALL && foo_SHELL_WRAPPER

The current fix is relatively dumb as far as where to add HpLim
checks: it will always perform a check unless we know that we're
returning from a closure or we are doing a non let-no-escape case
analysis.  The performance impact on the nofib suite looks like this:

            Min          +5.7%     -0.0%     -6.5%     -6.4%    -50.0%
            Max          +6.3%     +5.8%     +5.0%     +5.5%     +0.8%
 Geometric Mean          +6.2%     +0.1%     +0.5%     +0.5%     -0.8%

Overall, the executable bloat is the biggest problem, so we keep the old
omit-yields optimization on by default. Remember that if you need an
interruptibility guarantee, you need to recompile all of your libraries
with -fno-omit-yields.

A better fix would involve only inserting the yields necessary to break
loops; this is left as future work.
Signed-off-by: Edward Z. Yang <ezyang@mit.edu>

It's no longer needed in order to build the compiler itself.