GitLab

The reasons for that can be found in the wiki:
https://gitlab.haskell.org/ghc/ghc/wikis/nested-cpr/split-off-cpr

We now run CPR after demand analysis (except for after the final demand
analysis run just before code gen). CPR got its own dump flags
(`-ddump-cpr-anal`, `-ddump-cpr-signatures`), but not its own flag to
activate/deactivate. It will run with `-fstrictness`/`-fworker-wrapper`.

As explained on the wiki page, this step is necessary for a sane Nested
CPR analysis. And it has quite positive impact on compiler performance:

Metric Decrease:
    T9233
    T9675
    T9961
    T15263

See ghc/ghc#17801 (comment 253330)

No regression test, as it's hard to trigger.

Closes #12926

Update haddock submodule

The function was added to base with base-4.11 (GHC 8.4)

There is no issue with nested splices as they do not require any compile
time code execution. All execution is delayed until the top-level
splice.

This documentation-only patch fixes #17793

Compl Yue noticed that the linker was dumping the link map on SmartOS. This is
because Smartos uses the Solaris linker, which uses the `-64` flag, not
`-m64` like Gnu ld, to indicate that it should link for 64-bits. Fix the
configure script to handle the Solaris linker correctly.

Previously we would assert that threads which are sending a
`MSG_THROWTO` message must have their blocking status be blocked on the
message. In the usual case of a thread throwing to another thread this
is guaranteed by `stg_killThreadzh`.  However, `throwToSelf`, used by
the GC to kill threads which ran out of heap, failed to guarantee this.

Noted while debugging #17785.

usleep was removed in POSIX.1-2008.

Previously, we would accidentally make constraints like
forall a. C a => forall b. D b => E a b c as we traversed
superclasses. No longer!

This patch also expands Note [Eagerly expand given superclasses]
to work over quantified constraints; necessary for T16502b.

Close #17202 and #16502.

test cases: typecheck/should_compile/T{17202,16502{,b}}

For record updates where the `record_expr` is a variable, as in #17783:

```hs
data PartialRec = No
                | Yes { a :: Int, b :: Bool }
update No = No
update r@(Yes {}) = r { b = False }
```

We should make use of long distance info in
`-Wincomplete-record-updates` checking. But the call to `matchWrapper`
in the `RecUpd` case didn't specify a scrutinee expression, which would
correspond to the `record_expr` `r` here. That is fixed now.

Fixes #17783.

Closes #13142

A little late to the game, but better late than never.

Configure intree GMP with `--with-pic` instead of patching it. Moreover
the correct patching was only done for x86_64/darwin (see #17799).

Once again make sure this dumps the STG used for codegen.

This patch implements the [sugggestion from Simon (PJ)](ghc/ghc#14628 (comment 146559)):
- Make `TcErrors.getSkolemInfo` return a `SkolemInfo` rather than an `Implication`.
- If `getSkolemInfo` gets `RuntimeUnk`s, just return a new data constructor in `SkolemInfo`, called `RuntimeUnkSkol`.
- In `TcErrors.pprSkols` print something sensible for a `RuntimeUnkSkol`.

The `getSkolemInfo` function paniced while formating suggestions to add type annotations (subfunction `suggestAddSig`)
to a *"Couldn't match type ‘x’ with ‘y’"* error message.
The `getSkolemInfo` function didn't find any Implication value and paniced.
With this patch the `getSkolemInfo` function does no longer panic, if it finds `RuntimeUnkSkol`s.

As the panic occured while processing an error message, we don't need to implement any new error message!

This commit implements a few Windows-specific fixes which get us from a CI
job that can't even get as far as starting the testsuite driver, to a state
where we can run the entire testssuite (but have test failures to fix).

- Don't forget about a potential extension for the haddock program, when
  preparing the bindist.
- Build the timeout program, used by the testsuite driver on Windows in place
  of the Python script used elsewhere, using the boot compiler. We could
  alternatively build it with the compiler that we're going to test but this
  would be a lot more tedious to write.
- Implement a wrapper-script less installation procedure for Windows, in
  `hadrian/bindist/Makefile.
- Make dependencies a bit more accurate in the aforementioned Makefile.
- Update Windows/Hadrian CI job accordingly.

This patch fixes #17486.

This link appears to have been forgotten in
0dad81ca .

This appears to be a cut-and-paste error.

Allowing the test metric output to be captured to a file, a la
the METRIC_FILE environment variable of the make build system.

When requesting more than BLOCKS_PER_MBLOCK blocks allocGroup can return a
different number of blocks than requested. Here we use the number of
requested blocks, however arenaFree will subtract the actual number of
blocks we got from arena_blocks (possibly) resulting in a negative value
and triggering ASSERT(arena_blocks >= 0).

The comment in GHC.Base claimed that ($) couldn't be used in that module
as it was wired-in. However, this is no longer true; ($) is merely known
key and is defined in Haskell (with a RuntimeRep-polymorphic type) in
GHC.Base.

The one piece of magic that ($) retains is that it a special typing
rule to allow type inference with higher-rank types
(e.g. `runST $ blah`; see Note [Typing rule for ($)] in TcExpr).

The bug seems to be fixed in the meantime, make sure it stays fixed.

Closes #12760

This is in preparation of backwards-incompatible changes in happy.
See https://github.com/simonmar/happy/issues/166

While single-line comments are supported by C99, dtrace on SmartOS
apparently doesn't support them yet.

This fixes #17763

cc/ @pepeiborra

This alternative is redundant and triggers no warning when building with 8.6.5

Closes #17094.