Rodrigo Mesquita authored 8 months ago and Marge Bot committed 8 months ago

The `mi_usages` field of the interface files must use a deterministic
list of `Usage`s to guarantee a deterministic interface. However, this
list was, in its origins, constructed from a `ModuleEnv` which uses a
non-deterministic ordering that was leaking into the interface.

Specifically, ImportedMods = ModuleEnv ... would get converted to a list and
then passed to `mkUsageInfo` to construct the Usages.

The solution is simple. Back `ImportedMods` with a deterministic map.
`Map Module ...` is enough, since the Ord instance for `Module` already
uses a stable, deterministic, comparison.

Fixes #25131

Rodrigo Mesquita authored 8 months ago and Marge Bot committed 8 months ago

Sorting the Avails in DocStructures is required to produce fully
deterministic interface files in presence of re-exported modules.

Fixes #25104

This commit adds support for multiline strings, proposed at
https://github.com/ghc-proposals/ghc-proposals/pull/569.
Multiline strings can now be written as:

    myString =
      """
      this is a
      multiline string
      """

The multiline string will have leading indentation stripped away.
Full details of this post-processing may be found at the new
GHC.Parser.String module.

In order to cleanly implement this and maximize reusability, I
broke out the lexing logic for strings out of Lexer.x into a
new GHC.Parser.String module, which lexes strings with any
provided "get next character" function. This also gave us the
opportunity to clean up this logic, and even optimize it a bit.
With this change, parsing string literals now takes 25% less
time and 25% less space.

Andrei Borzenkov authored 1 year ago and Marge Bot committed 8 months ago

This patch extends the grammar of expressions with syntax that is
typically found only in types:
  * function types (a -> b), (a ->. b), (a %m -> b)
  * constrained types (ctx => t)
  * forall-quantification (forall tvs. t)

The new forms are guarded behind the RequiredTypeArguments extension,
as specified in GHC Proposal #281. Examples:

  {-# LANGUAGE RequiredTypeArguments #-}
  e1 = f (Int    -> String)          -- function type
  e2 = f (Int %1 -> String)          -- linear function type
  e3 = f (forall a. Bounded a => a)  -- forall type, constraint

The GHC AST and the TH AST have been extended as follows:

   syntax        | HsExpr   | TH.Exp
  ---------------+----------+--------------
   a -> b        | HsFunArr | ConE (->)
   a %m -> b     | HsFunArr | ConE FUN
   ctx => t      | HsQual   | ConstrainedE
   forall a. t   | HsForAll | ForallE
   forall a -> t | HsForAll | ForallVisE

Additionally, a new warning flag -Wview-pattern-signatures has been
introduced to aid with migration to the new precedence of (e -> p :: t).

Co-authored-by: Vladislav Zavialov <vlad.z.4096@gmail.com>

CLC proposals 231 and 261:

- Add exception type metadata to SomeException's displayException.
- Add "Exception" header to default exception handler.

See:

https://github.com/haskell/core-libraries-committee/issues/231
https://github.com/haskell/core-libraries-committee/issues/261

Update stm submodule for test fixes.

sheaf authored 8 months ago and Marge Bot committed 8 months ago

This commit modifies GHC.HsToCore.Pmc.Solver.generateInhabitingPatterns
to prioritise reporting COMPLETE sets in which all of the ConLikes
are in scope. This avoids suggesting out of scope constructors
when displaying an incomplete pattern match warning, e.g. in

  baz :: Ordering -> Int
  baz = \case
    EQ -> 5

we prefer:

  Patterns of type 'Ordering' not matched:
      LT
      GT

over:

  Patterns of type 'Ordering' not matched:
      OutOfScope

Fixes #25115

Matthew Pickering authored 8 months ago and Marge Bot committed 8 months ago

It was assumed that module names were unique but that isn't true with
multiple units.

The fix is quite simple, maintain a set of `(ModuleName, UnitId)` and
query that to see whether the module has been specified.

Fixes #25122

Sylvain Henry authored 8 months ago and Marge Bot committed 8 months ago

The code was introduced in 3549c952.
It used `getPackageIncludePath` which name doesn't convey that it looks
into all include paths of the preload units too. So this behavior is
probably unintentional and it should be ok to change it.

Fix #25106

Rodrigo Mesquita authored 8 months ago and Marge Bot committed 8 months ago

Fixes #25091

jeffrey young authored 9 months ago and Marge Bot committed 8 months ago

Dump better debugging information when a symbol address is null.

Part of the haskell.nix patches upstream project

Co-authored-by: Sylvain Henry <sylvain@haskus.fr>

jeffrey young authored 9 months ago and Marge Bot committed 8 months ago

part of the upstream haskell.nix patches

Rodrigo Mesquita authored 9 months ago and Marge Bot committed 8 months ago

A small refactor to use deriving via GHC.Utils.Monad.State.Strict
Monad instances for state Monads with unboxed/strict results which all
re-implemented the one-shot trick in the instance and used unboxed
tuples:

* CmmOptM in GHC.Cmm.GenericOpt
* RegM in GHC.CmmToAsm.Reg.Linear.State
* UniqSM in GHC.Types.Unique.Supply

Matthew Pickering authored 8 months ago and Marge Bot committed 8 months ago

* Incorrect version range on the alpine bindists
* Missing underscore in "unknown_versioning"

Fixes #25119

Rodrigo Mesquita authored 2 years ago and Marge Bot committed 8 months ago

Also migrate Fixity from GHC.Hs to Language.Haskell.Syntax
since it no longer uses any GHC-specific data types.

Fixed arrow desugaring bug. (This was dead code before.)
Remove mkOpFormRn, it is also dead code, only used in the arrow
desugaring now removed.

Co-authored-by: Fabian Kirchner <kirchner@posteo.de>
Co-authored-by: Alan Zimmerman <alan.zimm@gmail.com>

Arnaud Spiwack authored 9 months ago and Marge Bot committed 8 months ago

Fixes #25081

Ben Gamari authored 10 months ago and Krzysztof Gogolewski committed 8 months ago

This has not existed for over a decade.

Simon Peyton Jones authored 8 months ago and Marge Bot committed 8 months ago

As #25096 showed, the occurrence analyser was getting one-shot info
flat out wrong.

This commit does two things:

* It fixes the bug and actually makes the code a bit tidier too.
  The work is done in the new function
     GHC.Core.Opt.OccurAnal.mkRhsOccEnv,
  especially the bit that prepares the `occ_one_shots` for the RHS.

  See Note [The OccEnv for a right hand side]

* When floating out a binding we must be conservative about one-shot
  info.  But we were zapping the entire demand info, whereas we only
  really need zap the /top level/ cardinality.

  See Note [Floatifying demand info when floating]
  in GHC.Core.Opt.SetLevels

For some reason there is a 2.2% improvement in compile-time allocation
for CoOpt_Read.  Otherwise nickels and dimes.

Metric Decrease:
    CoOpt_Read

It seems we reserve 8 registers instead of four for global regs
based on the layout in Note [AArch64 Register assignments].

I'm not sure it's neccesary, but for now we just accept this state of
affairs and simple update -fregs-graph to account for this.

Simon Peyton Jones authored 9 months ago and Marge Bot committed 8 months ago

See Note [Case-of-case and full laziness]
in GHC.Driver.Config.Core.Opt.Simplify

Sergey Vinokurov authored 10 months ago and Marge Bot committed 8 months ago

This way the Generically1 newtype could be used to derive Eq1 and Ord1
for user types with DerivingVia.

The CLC proposal is https://github.com/haskell/core-libraries-committee/issues/273.

The GHC issue is #24312.

Simon Peyton Jones authored 9 months ago and Marge Bot committed 8 months ago

As #25094 showed, the previous tests for forgetfulness was
plain wrong, when there was a forgetful synonym in the RHS
of a synonym.

This broke configure in subtle ways resulting in #25076 where hadrian
didn't end up the boot compiler it was configured to use.

This reverts commit 209d09f5.

Matthew Pickering authored 9 months ago and Marge Bot committed 9 months ago

I made some mistakes in 20383006

* Syntax error
* The aarch-deb11 bindist doesn't exist

I tested against the latest nightly pipeline locally:

```
nix run .gitlab/generate-ci#generate-job-metadata
nix shell -f .gitlab/rel_eng/ -c ghcup-metadata --pipeline-id 98286 --version 9.11.20240715 --fragment --date 2024-07-17 --metadata=/tmp/meta
```

Torsten Schmits authored 9 months ago and Marge Bot committed 9 months ago

!12347 and !12582 introduced breaking changes to these two constructors
and mitigated that with pattern synonyms.

Matthew Pickering authored 9 months ago and Marge Bot committed 9 months ago

The tests which relied on find_so would fail when `test` was run
before the tree was built. This was because `find_so` was evaluated too
eagerly.

We can fix this by waiting to query the location of the libraries until
after the compiler has built them.

Matthew Pickering authored 9 months ago and Marge Bot committed 9 months ago

The arch argument was ignored when making the jobname, which lead to
failures when generating metadata for the alpine_3_18-aarch64 bindist.

Fixes #25089

Since debian 10 is now EOL we migrate onwards to debian 12 as the basis
for most platform independent validation jobs.

Matthew Pickering authored 9 months ago and Marge Bot committed 9 months ago

The groups are now split up so it's easier to see which jobs are
generated for each platform

No change in behaviour, just refactoring.

Zubin authored 9 months ago and Marge Bot committed 9 months ago

source locations on out of package instances

Fixes #24929

Zubin authored 9 months ago and Marge Bot committed 9 months ago

and use this to propagate these on items re-exported from dependent packages.

Fixes #25037

Zubin authored 9 months ago and Marge Bot committed 9 months ago

foreign imports and template haskell.

Fixes #24964

sheaf authored 9 months ago and Marge Bot committed 9 months ago

This patch ensures we taken into account COMPLETE pragmas when we
compute whether a pattern is irrefutable. In particular, if a pattern
synonym is the sole member of a COMPLETE pragma (without a result TyCon),
then we consider a pattern match on that pattern synonym to be irrefutable.

This affects the desugaring of do blocks, as it ensures we don't use
a "fail" operation.

Fixes #15681 #16618 #22004

Matthew Pickering authored 9 months ago and Marge Bot committed 9 months ago

The TEST_CC variable should be set based on the test compiler, which may
be different to the compiler which is set to CC on your system (for
example when cross compiling).

Fixes #24946

sheaf authored 9 months ago and Marge Bot committed 9 months ago

This replaces the rather hacky logic we had in place for checking
CPU features. In particular, this means that feature availability now
works properly on Windows.

Hécate Kleidukos authored 9 months ago and Marge Bot committed 9 months ago

This reduces the number of operations done on each Interface, because
supported languages and extensions are determined from architecture and
operating system of the build host. This information remains stable
across Interfaces, and as such doesn not need to be recovered for each
Interface.