GitLab

Trac #15009 showed that, for Given TyVar/TyVar equalities, we really
want to orient them with the deepest-bound skolem on the left. As it
happens, we also want to do the same for Wanteds, but for a different
reason (more likely to be touchable).  Either way, deepest wins:
see TcUnify Note [Deeper level on the left].

This observation led me to some significant changes:

* A SkolemTv already had a TcLevel, but the level wasn't really being
  used.   Now it is!

* I updated added invariant (SkolInf) to TcType
  Note [TcLevel and untouchable type variables], documenting that
  the level number of all the ic_skols should be the same as the
  ic_tclvl of the implication

* FlatSkolTvs and FlatMetaTvs previously had a dummy level-number of
  zero, which messed the scheme up.   Now they get a level number the
  same way as all other TcTyVars, instead of being a special case.

* To make sure that FlatSkolTvs and FlatMetaTvs are untouchable (which
  was previously done via their magic zero level) isTouchableMetaTyVar
  just tests for those two cases.

* TcUnify.swapOverTyVars is the crucial orientation function; see the
  new Note [TyVar/TyVar orientation].  I completely rewrote this function,
  and it's now much much easier to understand.

I ended up doing some related refactoring, of course

* I noticed that tcImplicitTKBndrsX and tcExplicitTKBndrsX were doing
  a lot of useless work in the case where there are no skolems; I
  added a fast-patch

* Elminate the un-used tcExplicitTKBndrsSig; and thereby get rid of
  the higher-order parameter to tcExpliciTKBndrsX.

* Replace TcHsType.emitTvImplication with TcUnify.checkTvConstraints,
  by analogy with TcUnify.checkConstraints.

* Inline TcUnify.buildImplication into its only call-site in
  TcUnify.checkConstraints

* TcS.buildImplication becomes TcS.CheckConstraintsTcS, with a
  simpler API

* Now that we have NoEvBindsVar we have no need of termEvidenceAllowed;
  nuke the latter, adding Note [No evidence bindings] to TcEvidence.

Happily, both of these issues appear to have been fixed in GHC 8.2.
Let's add regression tests for them to ensure that they stay fixed.

Summary:
See the new note.
This should fix cb5c2fe8 enough
to unbreak Windows and OS X builds.

Test Plan: manual testing with patched gdb

Reviewers: bgamari, simonmar, erikd

Reviewed By: bgamari

Subscribers: rwbarton, thomie, carter

Differential Revision: https://phabricator.haskell.org/D4694

Summary:
There were a number of leaks causing previously loaded modules to be
retained after a new `:load`.  This fixes enough leaks to get the
tests to pass from D4658.

Test Plan: See new tests in D4658

Reviewers: niteria, bgamari, simonpj, erikd

Subscribers: thomie, carter

GHC Trac Issues: #15111

Differential Revision: https://phabricator.haskell.org/D4659

Addressing review comments on D4637

Under certain circumstances, `GeneralizedNewtypeDeriving`
can emit code which uses unboxed tuple types, but if `UnboxedTuples`
wasn't enabled, the error message that GHC gave didn't make it very
clear that it could be worked around by explicitly enabling the
extension. Easily fixed.

Test Plan: make test TEST=T15073

Reviewers: bgamari

Reviewed By: bgamari

Subscribers: simonpj, thomie, carter

GHC Trac Issues: #15073

Differential Revision: https://phabricator.haskell.org/D4620

Fixes #15081.

Test Plan: cd libraries/base && make test TEST="enumNumeric"

Reviewers: hvr, bgamari

Reviewed By: bgamari

Subscribers: dfeuer, simonpj, thomie, carter

GHC Trac Issues: #15081

Differential Revision: https://phabricator.haskell.org/D4650

GHC previously had a handful of special cases for
pretty-printing equalities in a more user-friendly manner, but they
were far from comprehensive (see #15039 for an example of where this
fell apart).

This patch makes the pretty-printing of equalities much more
systematic. I've adopted the approach laid out in
https://ghc.haskell.org/trac/ghc/ticket/15039#comment:4, and updated
`Note [Equality predicates in IfaceType]` accordingly. We are now
more careful to respect the properties of the
`-fprint-explicit-kinds` and `-fprint-equality-relations` flags,
which led to some improvements in error message outputs.

Along the way, I also tweaked the error-reporting machinery not to
print out the type of a typed hole when the type is an unlifted
equality, since it's kind (`TYPE ('TupleRep '[])`) was more
confusing than anything.

Test Plan: make test TEST="T15039a T15039b T15039c T15039d"

Reviewers: simonpj, goldfire, bgamari

Reviewed By: simonpj

Subscribers: rwbarton, thomie, carter

GHC Trac Issues: #15039

Differential Revision: https://phabricator.haskell.org/D4696

Test Plan: ci, using it in monadic code.

Reviewers: bgamari, mpickering

Reviewed By: mpickering

Subscribers: rwbarton, thomie, carter

Differential Revision: https://phabricator.haskell.org/D4697

Ignore any errors thrown by finalizers when running them.

This prevents a faulty finalizer from stopping the rest being called.

Test Plan: ./validate, new test T13167

Reviewers: hvr, bgamari, simonmar

Reviewed By: bgamari, simonmar

Subscribers: rwbarton, thomie, carter

GHC Trac Issues: #13167

Differential Revision: https://phabricator.haskell.org/D4693

Summary:
The idea here is to save a little code size and some work in the GC,
by collapsing FUN_STATIC closures and their SRTs.

This is (4) in a series; see D4632 for more details.

There's a tradeoff here: more complexity in the compiler in exchange
for a modest code size reduction (probably around 0.5%).

Results:
* GHC binary itself (statically linked) is 1% smaller
* -0.2% binary sizes in nofib (-0.5% module sizes)

Full nofib results comparing D4634 with this: P177 (ignore runtimes,
these aren't stable on my laptop)

Test Plan: validate, nofib

Reviewers: bgamari, niteria, simonpj, erikd

Subscribers: thomie, carter

Differential Revision: https://phabricator.haskell.org/D4637

Summary:
An info table with an SRT normally looks like this:

    StgWord64 srt_offset
    StgClosureInfo layout
    StgWord32 layout
    StgWord32 has_srt

But we only need 32 bits for srt_offset on x86_64, because the small
memory model requires that code segments are at most 2GB. So we can
optimise this to

    StgClosureInfo layout
    StgWord32 layout
    StgWord32 srt_offset

saving a word.  We can tell whether the info table has an SRT or not,
because zero is not a valid srt_offset, so zero still indicates that
there's no SRT.

Test Plan:
* validate
* For results, see D4632.

Reviewers: bgamari, niteria, osa1, erikd

Subscribers: thomie, carter

Differential Revision: https://phabricator.haskell.org/D4634

Summary:
This change makes it possible to generate a static 32-bit relative label
offset on x86_64. Currently we can only generate word-sized label
offsets.

This will be used in D4634 to shrink info tables.  See D4632 for more
details.

Test Plan: See D4632

Reviewers: bgamari, niteria, michalt, erikd, jrtc27, osa1

Subscribers: thomie, carter

Differential Revision: https://phabricator.haskell.org/D4633

Summary:
- Previously we would hvae a single big table of pointers per module,
  with a set of bitmaps to reference entries within it. The new
  representation is identical to a static constructor, which is much
  simpler for the GC to traverse, and we get to remove the complicated
  bitmap-traversal code from the GC.

- Rewrite all the code to generate SRTs in CmmBuildInfoTables, and
  document it much better (see Note [SRTs]). This has been something
  I've wanted to do since we moved to the new code generator, I
  finally had the opportunity to finish it while on a transatlantic
  flight recently :)

There are a series of 4 diffs:

1. D4632 (this one), which does the bulk of the changes

2. D4633 which adds support for smaller `CmmLabelDiffOff` constants

3. D4634 which takes advantage of D4632 and D4633 to save a word in
   info tables that have an SRT on x86_64. This is where most of the
   binary size improvement comes from.

4. D4637 which makes a further optimisation to merge some SRTs with
   static FUN closures.  This adds some complexity and the benefits
   are fairly modest, so it's not clear yet whether we should do this.

Results (after (3), on x86_64)

- GHC itself (staticaly linked) is 5.2% smaller

- -1.7% binary sizes in nofib, -2.9% module sizes. Full nofib results: P176

- I measured the overhead of traversing all the static objects in a
  major GC in GHC itself by doing `replicateM_ 1000 performGC` as the
  first thing in `Main.main`.  The new version was 5-10% faster, but
  the results did vary quite a bit.

- I'm not sure if there's a compile-time difference, the results are
  too unreliable.

Test Plan: validate

Reviewers: bgamari, michalt, niteria, simonpj, erikd, osa1

Subscribers: thomie, carter

Differential Revision: https://phabricator.haskell.org/D4632

Summary: A better alternative to D4657.

Test Plan:
```
cd testsuite/tests/codeGen/should_run
../../../../inplace/bin/ghc-stage2 -debug cgrun001
nm cgrun001 | grep findPtr
```

Reviewers: bgamari, Phyx, erikd

Subscribers: rwbarton, thomie, carter

Differential Revision: https://phabricator.haskell.org/D4683

Simplifier depends on typechecker in two points: `thNameToGhcName`
(`lookupThName_maybe`, in particular)  and `lookupGlobal`. We want to
cut the ties in two steps.

1. (Presented in this commit), reimplement both functions in a way that
doesn't use typechecker.

2. (Should follow), do code moving: a) `lookupGlobal` should go in some
typechecker-free place; b) `thNameToGhcName` should leave simplifier,
because it is not used there at all (probably, it should be placed
somewhere where `GhcPlugins` can see it -- this is suggested by Joachim
on Trac).

Details
=======

We redesigned lookup interface a bit so that it exposes some
`IO`-equivalents of `Tc`-features in use.

First, `CoreMonad.hs` still calls `lookupGlobal` which is no longer
bound to the typechecker monad, but still resides in `TcEnv.hs` — it
should be moved out of Tc-land at some point (“Phase 2”) in the
future in order to achieve its part of the #14391's goal.

Second, `lookupThName_maybe` is eliminated from `CoreMonad.hs`
completely; this already achieves its part of the goal of #14391. Its
client, though, `thNameToGhcName`, is better to be moved in the future
also, for it is not used in the `CoreMonad.hs` (or anywhere else)
anyway. Joachim suggested “any module reexported by GhcPlugins (or
maybe even that module itself)”.

As a side goal, we removed `initTcForLookup` which was instrumental for
the past version of `lookupGlobal`. This, in turn, called for pushing
some more parts of the lookup interface from the `Tc`-monad to `IO`,
most notably, adding `IO`-version of `lookupOrig` and pushing
`dataConInfoPtrToName` to `IO`. The `lookupOrig` part, in turn,
triggered a slight redesign of name cache updating interface: we now
have both, `updNameCacheIO` and `updNameCacheTc`, both accepting `mod`
and `occ` to force them inside, instead of more error-prone outside
before. But all these hardly have to do anything with #14391, mere
refactoring.

Reviewers: simonpj, nomeata, bgamari, hvr

Reviewed By: simonpj, bgamari

Subscribers: rwbarton, thomie, carter

GHC Trac Issues: #14391

Differential Revision: https://phabricator.haskell.org/D4503

Previously, the `{add,sub}{Int,Word}C#` PrimOps weren't handled
in PrelRules (constant folding and algebraic simplification) at all.
This implements the necessary logic, so that using these primitives
isn't too punishing compared to their well-optimised, overflow-unaware
counterparts.

This is so that using these primitives in `enumFromThenTo @Int` can
be optimized by constant folding, reducing closure sizes.

Reviewers: bgamari, simonpj, hsyl20

Reviewed By: bgamari, simonpj

Subscribers: AndreasK, thomie, carter

GHC Trac Issues: #8763

Differential Revision: https://phabricator.haskell.org/D4605

For the threaded RTS, putting a private copy of the throwErrno
series in GHC.IO.FD which gets if the operation was on a socket,
so that we can call c_maperrno if need be.
For the non-threaded RTS, if memory serves we call GetLastError()
in case of an error on socket IO. However, we don't do the translation
ErrCode ↔ Errno currently (and besides, it's a primop) so we do it if
needed through c_maperrno_func in the asynchronous read/write
functions.
Signed-off-by: ARJANEN Loïc Jean David <arjanen.loic@gmail.com>

Reviewers: ekmett, hvr, bgamari

Reviewed By: bgamari

Subscribers: thomie, carter

GHC Trac Issues: #12012

Differential Revision: https://phabricator.haskell.org/D4639

Test Plan: Validate on Windows

Subscribers: thomie, carter

GHC Trac Issues: #14697, #15072

Differential Revision: https://phabricator.haskell.org/D4619

Test Plan: Validate with Hadrian and `libnuma` support

Reviewers: snowleopard, hvr, erikd, simonmar

Subscribers: izgzhen, alpmestan, thomie, carter

Differential Revision: https://phabricator.haskell.org/D4616

This lead to some rather significant performance regressions.
Specifically,

    bytes allocated value is too high:
        Expected    T5631(normal) bytes allocated: 1106015512 +/-5%
        Lower bound T5631(normal) bytes allocated: 1050714736
        Upper bound T5631(normal) bytes allocated: 1161316288
        Actual      T5631(normal) bytes allocated: 1164953208
        Deviation   T5631(normal) bytes allocated:        5.3 %
    *** unexpected stat test failure for T5631(normal)
    max_bytes_used value is too high:
        Expected    T9630(normal) max_bytes_used: 35324712 +/-15%
        Lower bound T9630(normal) max_bytes_used: 30026005
        Upper bound T9630(normal) max_bytes_used: 40623419
        Actual      T9630(normal) max_bytes_used: 43490984
        Deviation   T9630(normal) max_bytes_used:     23.1 %
    *** unexpected stat test failure for T9630(normal)

This reverts commit 7271db46.
This reverts commit b750dcc5.
This reverts commit 33de71fa.

These were both improved by d92c7556 but the
changes were lost due to merge silliness.

Trac #15152 showed that when a flag turned an error into a warning, we
were still (alas) suppressing subequent errors; includign their
essential addTcEvBind.  That led (rightly) to a Lint error.

This patch fixes it, and incidentally tidies up an ad-hoc special
case of out-of-scope variables (see the old binding for
'out_of_scope_killer' in 'tryReporters').

No test, because the problem was only shown up when turning
inaccessible code into a warning.

commit b2ff5dde "Fix #15038"
added new stable closure 'absentSumFieldError_closure' to
base package. This closure is used in rts package.

Unfortunately the symbol was not explicitly exported and build
failed on windows as:

```
"inplace/bin/ghc-stage1" -o ...hsc2hs.exe ...
rts/dist/build/libHSrts.a(RtsStartup.o): In function `hs_init_ghc':

rts/RtsStartup.c:272:0: error:
     undefined reference to `base_ControlziExceptionziBase_absentSumFieldError_closure'
    |
272 |     getStablePtr((StgPtr)absentSumFieldError_closure);
    | ^

```

This change adds 'absentSumFieldError_closure' to explicit export
into libHSbase.def.
Signed-off-by: Sergei Trofimovich <slyfox@gentoo.org>

See #15019. When removing an unnecessary type equality check in #14737,
several regression tests failed. The cause was that some coercions that
are actually Refl coercions weren't passed in as such, which made the
equality check needlessly complex (Refl coercions can be discarded in
this particular check immediately, without inspecting the types at all).

We fix that, and get additional performance improvements for free.

Reviewers: goldfire, bgamari, simonpj

Reviewed By: bgamari, simonpj

Subscribers: simonpj, thomie, carter

Differential Revision: https://phabricator.haskell.org/D4635

This reverts commit d4abd031.

Reviewers: bgamari

Reviewed By: bgamari

Subscribers: thomie, carter

Differential Revision: https://phabricator.haskell.org/D4661

Fix cross-compilation failure from linux to windows:
```
  CC utils/unlit/dist-install/build/fs.o
utils/unlit/fs.c:24:10: fatal error: sys\stat.h: No such file or directory
 #include <sys\stat.h>
          ^~~~~~~~~~~~
```
Signed-off-by: Sergei Trofimovich <slyfox@gentoo.org>

* `length` is improved by using the default definition,
  while `foldr1` is improved by using a custom one.

* Several methods had useless lazy pattern matches
  (i.e., the functions were actually strict in those arguments).
  Remove `~`s to clarify.

Reviewers: hvr, bgamari, mpickering, nomeata

Reviewed By: bgamari

Subscribers: ygale, rwbarton, thomie, carter

GHC Trac Issues: #15131

Differential Revision: https://phabricator.haskell.org/D4677

A rather detailed summary can be found at:

    https://gist.github.com/alpmestan/be82b47bb88b7dc9ff84105af9b1bb82

This doesn't fix all expectation mismatches yet, but we're down to about
20 mismatches with my previous patch and this one, as opposed to ~150
when I got started.

Test Plan: ./validate --slow

Reviewers: bgamari, erikd, simonmar

Reviewed By: simonmar

Subscribers: thomie, carter

GHC Trac Issues: #14890

Differential Revision: https://phabricator.haskell.org/D4636

This implements the first part proposed in #13753:

Define a special magic "null" environment, which instructs GHC to ignore
any package environment files. To this end, I propose to use the name
`-` (i.e. a single dash), as that is more portable than using the empty
string for environment variables. In other words, a

- `-package-env -` CLI flag, or a
- `GHC_ENVIRONMENT=-` env var (unless a `-package-env` flag is present)

would inhibit GHC from looking up and interpreting any package
environment files from the filesystem; this is the equivalent of
`-ignore-dot-ghci` for package environment files.

Reviewers: bgamari

Reviewed By: bgamari

Subscribers: rwbarton, thomie, carter

GHC Trac Issues: #13753

Differential Revision: https://phabricator.haskell.org/D4690

This fixes #15021 using a the same approach as was used to fix the issue
in ghc (#10762).

Test Plan: Validate on Windows as user whose username contains
non-ASCII characters

Reviewers: simonmar

Reviewed By: simonmar

Subscribers: lehins, thomie, carter

GHC Trac Issues: #15021

Differential Revision: https://phabricator.haskell.org/D4642

Trying to determine when to insert parentheses during TH
conversion is a bit of a mess. There is an assortment of functions
that try to detect this, such as:

* `hsExprNeedsParens`
* `isCompoundHsType`
* `hsPatNeedsParens`
* `isCompoundPat`
* etc.

To make things worse, each of them have slightly different semantics.
Plus, they don't work well in the presence of explicit type
signatures, as #14875 demonstrates.

All of these problems can be alleviated with the use of an explicit
precedence argument (much like what `showsPrec` currently does). To
accomplish this, I introduce a new `PprPrec` data type, and define
standard predences for things like function application, infix
operators, function arrows, and explicit type signatures (that last
one is new). I then added `PprPrec` arguments to the various
`-NeedsParens` functions, and use them to make smarter decisions
about when things need to be parenthesized.

A nice side effect is that functions like `isCompoundHsType` are
now completely unneeded, since they're simply aliases for
`hsTypeNeedsParens appPrec`. As a result, I did a bit of refactoring
to remove these sorts of functions. I also did a pass over various
utility functions in GHC for constructing AST forms and used more
appropriate precedences where convenient.

Along the way, I also ripped out the existing `TyPrec`
data type (which was tailor-made for pretty-printing `Type`s) and
replaced it with `PprPrec` for consistency.

Test Plan: make test TEST=T14875

Reviewers: alanz, goldfire, bgamari

Reviewed By: bgamari

Subscribers: rwbarton, thomie, carter

GHC Trac Issues: #14875

Differential Revision: https://phabricator.haskell.org/D4688

Reviewers: sjakobi, bgamari

Reviewed By: sjakobi

Subscribers: rwbarton, thomie, carter

Differential Revision: https://phabricator.haskell.org/D4687