GitLab

As #19668 showed, there was an /asymptotic/ slow-down in zonking in
GHC 9.0, exposed in test T9198.  The bug was actually present in earlier
compilers, but by a fluke didn't actually show up in any of our tests;
but adding Quick Look exposed it.

The bug was that in zonkTyVarOcc we
1. read the meta-tyvar-env variable
2. looked up the variable in the env
3. found a 'miss'
4. looked in the variable, found `Indirect ty`
5. zonked `ty`
6. update the env *gotten from step 1* to map the variable
   to its zonked type.

The bug is that we thereby threw away all teh work done in step 4.
In T9198 that made an enormous, indeed asymptotic difference.

The fix is easy: use updTcRef.

I commented in `Note [Sharing when zonking to Type]`

-------------------------
Metric Decrease:
    T9198
-------------------------

We now use DsM as the base monad for writing hie files and properly
initialise it from the TcGblEnv.

Before, we would end up reading the interface file from disk for the
module we were currently compiling. The modules iface then ended up in
the EPS causing all sorts of subtle
carnage, including difference in the generated core and haddock emitting
a lot of warnings. With the fix, the
module in the TcGblEnv is set correctly so the lookups happen in the
local name env rather than thinking the identifier comes from an
external package.

Fixes #19693 and #19334

There were two different issues:

1. integralFractionalLit needed to be passed an already negated value. (T19680)
2. negateFractionalLit did not actually negate the argument, only
flipped the negation flag. (T19680A)

Fixes #19680

Fixes #19683

follow-up from !4675

This vastly reduces memory usage when compiling with `--make` mode, from
about 900M when compiling Cabal to about 300M.

As a matter of uniformity, it also ensures that reading from an
interface performs the same as using the in-memory cache. We can also
delete all the horrible knot-tying in updateIdInfos.

Goes some way to fixing #13586

Accept new output of tests fixing some bugs along the way

-------------------------
Metric Decrease:
    T12545
-------------------------

Bumps the `haddock` submodule.

Fixes #19688.

Ticket #13873 unexpectedly showed that a SPECIALISE pragma made a
program run (a lot) slower, because less specialisation took place
overall. It turned out that the specialiser was missing opportunities
because of quantified type variables.

It was quite easy to fix. The story is given in
    Note [Specialising polymorphic dictionaries]

Two other minor fixes in the specialiser

* There is no benefit in specialising data constructor /wrappers/.
  (They can appear overloaded because they are given a dictionary
  to store in the constructor.)  Small guard in canSpecImport.

* There was a buglet in the UnspecArg case of specHeader, in the
  case where there is a dead binder. We need a LitRubbish filler
  for the specUnfolding stuff.  I expanded
  Note [Drop dead args from specialisations] to explain.

There is a 4% increase in compile time for T13056, because we generate
more specialised code.  This seems OK.

Metric Increase:
    T13056

Otherwise, if the marge batch has decreased metrics, they will fail on
master which will result in the pipeline being cut short and the
expected metric values for the other jobs will not be updated.

This makes it more robust to people running it with `quick` flavour and
so on.

Fixes #19615

Otherwise, errors can go missing which arise when running the splices.

Fixes #19470

We want an accurate SrcSpan for redundant constraints:

        • Redundant constraint: Eq a
        • In the type signature for:
               f :: forall a. Eq a => a -> ()
      |
    5 | f :: Eq a => a -> ()
      |      ^^^^

This patch adds some plumbing to achieve this

* New data type GHC.Tc.Types.Origin.ReportRedundantConstraints (RRC)

* This RRC value is kept inside
  - FunSigCtxt
  - ExprSigCtxt

* Then, when reporting the error in GHC.Tc.Errors, use this SrcSpan
  to control the error message: GHC.Tc.Errors.warnRedundantConstraints

Quite a lot of files are touched in a boring way.

The problem was that ghci inserts some ticks around the crucial bit of
the expression. Just like in some of the other rules we now strip the
ticks so that the rule fires more reliably.

It was possible to defeat magicDict by using -fhpc as well, so not just an
issue in ghci.

Fixes #19667 and related to #19673

See https://docs.gitlab.com/ee/ci/variables/predefined_variables.html

This allows to start iserv by passing port 0 to startSlave,
which in turns allows to get an available port when no
port is known to be free a priori.

It isn't used for anything anymore

With this patch we switch from reading the globally installed
platformConstants file to reading the DerivedConstants.h header file
that is bundled in the RTS unit.  When we build the RTS unit itself, we
get it from its includes directories.

The new parser is more efficient and strict than the Read instance for
PlatformConstants and we get about 2.2MB less allocations in every
cases. However it only really shows in tests that don't allocate much,
hence the following metric decreases.

Metric Decrease:
    Naperian
    T10421
    T10547
    T12150
    T12234
    T12425
    T13035
    T18304
    T18923
    T5837
    T6048
    T18140

deriveConstants utility now generates a Haskell parser for
DerivedConstants.h. It can be used to replace the one used to read
platformConstants file.

A few refactorings made after looking at Core/STG

* Use Doc instead of SDoc in pprASCII to avoid passing the SDocContext
  that is never used.

* Inline every SDoc wrappers in GHC.Utils.Outputable to expose Doc
  constructs

* Add text/[] rule for empty strings (i.e., text "")

* Use a single occurrence of pprGNUSectionHeader

* Use bangs on Platform parameters and some others

Metric Decrease:
    ManyAlternatives
    ManyConstructors
    T12707
    T13035
    T13379
    T18698a
    T18698b
    T1969
    T3294
    T4801
    T5321FD
    T783

- This allows specialized mconcat implementations an opportunity to combine
  elements efficiently in a single pass.

- Inline the default implementation of `mconcat`, this
  may result in list fusion.

- In Monoids with strict `mappend`, implement `mconcat` as a strict left fold:

  * And (FiniteBits)
  * Ior (FiniteBits)
  * Xor (FiniteBits)
  * Iff (FiniteBits)
  * Max (Ord)
  * Min (Ord)
  * Sum (Num)
  * Product (Num)
  * (a -> m) (Monoid m)

- Delegate mconcat for WrappedMonoid to the underlying monoid.

Resolves: #17123

Per the discussion in !4890, we expect some stat changes:

    * T17123(normal) run/alloc   403143160.0     4954736.0 -98.8% GOOD

      This is the expected improvement in `fold` for a long list of
      `Text` elements.

    * T13056(optasm) ghc/alloc   381013328.0   447700520.0 +17.5% BAD

      Here there's an extra simplifier run as a result of the new methods
      of the Foldable instance for List.  It looks benign.  The test is
      a micro benchmark that compiles just the derived foldable instances
      for a pair of structures, a cost of this magnitude is not expected
      to extend to more realistic programs.

    * T9198(normal) ghc/alloc   504661992.0   541334168.0  +7.3% BAD

      This test regressed from 8.10 and 9.0 back to exponential blowup.
      This metric also fluctuates, for reasons not yet clear.  The issue
      here is the exponetial blowup, not this MR.

Metric Decrease:
    T17123
Metric Increase:
    T9198
    T13056

No changes to code; no changes to theory. Just better
explanation.

Close #17672.

By scratching our heads quite hard, we realized that
we should never kick out Given/Nominal equalities. This
commit tweaks the kick-out conditions accordingly.

See also Note [K4] which describes what is going on.

This does not fix a known misbehavior, but it should be
a small improvement in both practice (kicking out is bad,
and we now do less of it) and theory (a Given/Nominal should
behave just like a filled-in metavariable, which has no notion
of kicking out).

Kick out condition K2b really only makes sense for
inerts with a type variable on the left. This updates
the commentary and the code to skip this check for
inerts with type families on the left.

Also cleans up some commentary around solver invariants
and adds Note [K2b].

Close #19042.

test case: typecheck/should_compile/T19042

Also includes small unrelated type fix

The test max memory usage improves dramatically with the fixes to
memory usage in demand analyser from #15455

There used to be some cases were kinds were not generalised properly
before being printed in GHCi. This seems to have changed in the past so
now it's uncessary to tidy before printing out the test case.

```
> :set -XPolyKinds
> data A x y
> :k A
k1 -> k2 -> A
```

This tidying was causing issues with an attempt to increase sharing by
making `mkTyConApp` (see !4762)

The approach taking in this patch is that the tcl_bndrs in TcLclEnv are
zonked and tidied eagerly, so that work can be shared across multiple
calls to `relevant_bindings`.

To test this patch I tried without the `keepThisHole` filter and the
test finished quickly.

Fixes #14766