If we're doing heap profiling on an unprofiled executable we keep
allocating new space in initEra via nextEra on each profiler run but we
don't have a corresponding freeEra call.

We do free the last era in endHeapProfiling but previous eras will have
been overwritten by initEra and will never get free()ed.

Metric Decrease:
    space_leak_001

This refactors DictBinds into a data type rather than a pair.
No change in behaviour, just better code

Issue #17151 was a very tricky example of a bug in which the
specialiser accidentally constructs a recurive dictionary,
so that everything turns into bottom.

I have fixed variants of this bug at least twice before:
see Note [Avoiding loops].  It was a bit of a struggle
to isolate the problem, greatly aided by the work that
Alexey Kuleshevich did in distilling a test case.

Once I'd understood the problem, it was not difficult to fix,
though it did lead me a bit of refactoring in specImports.

Fixes #17947

When we have a ticky label for a proc, IdLabels for the ticky counter
and proc entry share the same Name. This caused overriding proc CafInfos
with the ticky CafInfos (i.e. NoCafRefs) during SRT analysis.

We now ignore the ticky labels when building SRTMaps. This makes sense
because:

- When building the current module they don't need to be in SRTMaps as
  they're initialized as non-CAFFY (see mkRednCountsLabel), so they
  don't take part in the dependency analysis and they're never added to
  SRTs.

  (Reminder: a "dependency" in the SRT analysis is a CAFFY dependency,
  non-CAFFY uses are not considered as dependencies for the algorithm)

- They don't appear in the interfaces as they're not exported, so it
  doesn't matter for cross-module concerns whether they're in the SRTMap
  or not.

See also the new Note [Ticky labels in SRT analysis].

This is necessary for certain record selectors with higher-rank
types, such as the examples in #18005. See
`Note [Impredicative record selectors]` in `TcTyDecls`.

Fixes #18005.

[ci skip]

This patch is joint work of Alexis King and Simon PJ.  It does some
significant refactoring of the type-class specialiser.  Main highlights:

* We can specialise functions with types like
     f :: Eq a => a -> Ord b => b => blah
  where the classes aren't all at the front (#16473).  Here we can
  correctly specialise 'f' based on a call like
     f @Int @Bool dEqInt x dOrdBool
  This change really happened in an earlier patch
     commit 2d0cf625
     Author: Sandy Maguire <sandy@sandymaguire.me>
     Date:   Thu May 16 12:12:10 2019 -0400
  work that this new patch builds directly on that work, and refactors
  it a bit.

* We can specialise functions with implicit parameters (#17930)
     g :: (?foo :: Bool, Show a) => a -> String
  Previously we could not, but now they behave just like a non-class
  argument as in 'f' above.

* We can specialise under-saturated calls, where some (but not all of
  the dictionary arguments are provided (#17966).  For example, we can
  specialise the above 'f' based on a call
     map (f @Int dEqInt) xs
  even though we don't (and can't) give Ord dictionary.

  This may sound exotic, but #17966 is a program from the wild, and
  showed significant perf loss for functions like f, if you need
  saturation of all dictionaries.

* We fix a buglet in which a floated dictionary had a bogus demand
  (#17810), by using zapIdDemandInfo in the NonRec case of specBind.

* A tiny side benefit: we can drop dead arguments to specialised
  functions; see Note [Drop dead args from specialisations]

* Fixed a bug in deciding what dictionaries are "interesting"; see
  Note [Keep the old dictionaries interesting]

This is all achieved by by building on Sandy Macguire's work in
defining SpecArg, which mkCallUDs uses to describe the arguments of
the call. Main changes:

* Main work is in specHeader, which marched down the [InBndr] from the
  function definition and the [SpecArg] from the call site, together.

* specCalls no longer has an arity check; the entire mechanism now
  handles unders-saturated calls fine.

* mkCallUDs decides on an argument-by-argument basis whether to
  specialise a particular dictionary argument; this is new.
  See mk_spec_arg in mkCallUDs.

It looks as if there are many more lines of code, but I think that
all the extra lines are comments!

The newlines break the command on windows.

In !2959 we noticed that there was some redundant code (in GHC.Cmm.Utils
and GHC.Cmm.StgToCmm.Utils) used to deal with `CmmStatics` datatype
(before SRT generation) and `RawCmmStatics` datatype (after SRT
generation).

This patch removes this redundant code by using a single GADT for
(Raw)CmmStatics.

Move handling of big literal strings from CmmToAsm to StgToCmm. It
avoids the use of `sdocWithDynFlags` (cf #10143). We might need to move
this handling even higher in the pipeline in the future (cf #17960):
this patch will make it easier.

We were mistakenly measuring program stats

This adds a definition to construct a singleton non-empty list
(Data.List.NonEmpty) according to issue #17851.

We now differentiate three cases of constructor bindings:

1)Bindings which we can "replace" with a reference to
  an existing closure. Reference the replacement closure
  when accessing the binding.
2)Bindings which we can "replace" as above. But we still
  generate a closure which will be referenced by modules
  importing this binding.
3)For any other binding generate a closure. Then reference
  it.

Before this patch 1) did only apply to local bindings and we
didn't do 2) at all.

flag to dump pretty printed contents of the .hie file

Metric Increase:
   hie002

Because of the regression on i386:

compile_time/bytes allocated increased from i386-linux-deb9 baseline @ HEAD~10:
    Expected    hie002 (normal) compile_time/bytes allocated: 583014888.0 +/-10%
    Lower bound hie002 (normal) compile_time/bytes allocated:   524713399
    Upper bound hie002 (normal) compile_time/bytes allocated:   641316377
    Actual      hie002 (normal) compile_time/bytes allocated:   877986292
    Deviation   hie002 (normal) compile_time/bytes allocated:        50.6 %
*** unexpected stat test failure for hie002(normal)

Two `ASSERT`s in `reifyDataCon` were always using `arg_tys`, but
`arg_tys` is not meaningful for GADT constructors. In fact, it's
worse than non-meaningful, since using `arg_tys` when reifying a
GADT constructor can lead to failed `ASSERT`ions, as #17305
demonstrates.

This patch applies the simplest possible fix to the immediate
problem. The `ASSERT`s now use `r_arg_tys` instead of `arg_tys`, as
the former makes sure to give something meaningful for GADT
constructors. This makes the panic go away at the very least. There
is still an underlying issue with the way the internals of
`reifyDataCon` work, as described in
#17305 (comment 227023), but we
leave that as future work, since fixing the underlying issue is
much trickier (see
#17305 (comment 227087)).

Broken by 57b888c0

GHC 8.10 ships with `Cabal-3.2.0.0`, so it would be convenient to
make Hadrian supporting building against 3.2.* instead of having to
rebuild the entirety of `Cabal-3.0.0.0`. There is one API change in
`Cabal-3.2.*` that affects Hadrian: the `synopsis` and `description`
functions now return `ShortText` instead of `String`. Since Hadrian
manipulates these `String`s in various places, I found that the
simplest fix was to use CPP to convert `ShortText` to `String`s
where appropriate.

Not only is this a reasonable efficiency measure but it avoids making
reentrant calls into ncurses, which is not thread-safe. See #17922.

Fix #13380 and #17676 by

1. Changing `raiseIO#` to have `topDiv` instead of `botDiv`
2. Give it special treatment in `Simplifier.Util.mkArgInfo`, treating it
   as if it still had `botDiv`, to recover dead code elimination.

This is the first commit of the plan outlined in
!2525 (comment 260886).

The binder-swap transformation is implemented by the occurrence
analyser -- see Note [Binder swap] in OccurAnal. However it had
a very nasty corner in it, for the case where the case scrutinee
was a GlobalId.  This led to trouble and hacks, and ultimately
to #16296.

This patch re-engineers how the occurrence analyser implements
the binder-swap, by actually carrying out a substitution rather
than by adding a let-binding.  It's all described in
Note [The binder-swap substitution].

I did a few other things along the way

* Fix a bug in StgCse, which could allow a loop breaker to be CSE'd
  away.  See Note [Care with loop breakers] in StgCse.  I think it can
  only show up if occurrence analyser sets up bad loop breakers, but
  still.

* Better commenting in SimplUtils.prepareAlts

* A little refactoring in CoreUnfold; nothing significant
  e.g. rename CoreUnfold.mkTopUnfolding to mkFinalUnfolding

* Renamed CoreSyn.isFragileUnfolding to hasCoreUnfolding

* Move mkRuleInfo to CoreFVs

We observed respectively 4.6% and 5.9% allocation decreases for the following
tests:

Metric Decrease:
    T9961
    haddock.base

Fixes #17983.

In #17977, we ran into the reduction depth limit of the typechecker.
That was only a symptom of a much broader issue: The recursion depth
of the coverage checker for trying to instantiate strict fields in the
`nonVoid` test was far too high (100, the `defaultMaxTcBound`).

As a result, we were performing quite poorly on `T17977`.
Short of a proper termination analysis to prove emptyness of a type,
we just arbitrarily default to a much lower recursion limit of 3.

Fixes #17977.

Metric Decrease:
    T13035
    T1969

Fixes #17979.

[ci skip]

Before, there were two distinct Notes named
"Eta reduction for data families". This renames one of them to
"Implementing eta reduction for data families" to disambiguate the
two and fixes references in other parts of the codebase to ensure
that they are pointing to the right place.

Fixes #17313.

[ci skip]

Required in order to build hadrian using ghc-8.10

As noted in #17970, these (e.g. `getFileSystemEncoding` and
`setFileSystemEncoding`) previously had unfoldings, which would
break their global-ness.

While not strictly necessary, I also add a NOINLINE on
`initLocaleEncoding` since it is used in `System.IO`, ensuring that we
only system's query the locale encoding once.

Fixes #17970.

The combinator andM is used only once, and the code is shorter and
simpler if you inline it.

The panic in #17963 happened to be fixed by commit
e3c374cc. This patch adds a
regression test to ensure that it remains fixed.

Fixes #17963.

This allows us to remove several bits of CPP that are either always
true or no longer reachable. As an added bonus, we no longer need to
worry about importing `Control.Monad.Fail.fail` qualified to avoid
clashing with `Control.Monad.fail`, since the latter is now the same
as the former.

Since 54250f2d we expected T4267 to
fail, but it passes on CI.

- Simplify mkBuildExpr, the function newTyVars was called
  only on a one-element list.
- TTG: use noExtCon in more places. This is more future-proof.
- In zonkExpr, panic instead of printing a warning.

Within `checkValidDataCon`, we used to run `checkValidType` on the
argument types of a newtype constructor before running
`checkNewDataCon`, which ensures that the user does not attempt
non-sensical things such as newtypes with multiple arguments or
constraints. This works out in most situations, but this falls over
on a corner case revealed in #17955:

```hs
newtype T = Coercible () T => T ()
```

`checkValidType`, among other things, peforms an ambiguity check on
the context of a data constructor, and that it turn invokes the
constraint solver. It turns out that there is a special case in the
constraint solver for representational equalities (read: `Coercible`
constraints) that causes newtypes to be unwrapped (see
`Note [Unwrap newtypes first]` in `TcCanonical`). This special case
does not know how to cope with an ill formed newtype like `T`, so
it ends up panicking.

The solution is surprisingly simple: just invoke `checkNewDataCon`
before `checkValidType` to ensure that the illicit newtype
constructor context is detected before the constraint solver can
run amok with it.

Fixes #17955.