GitLab

When a new closure identifier is being established to a
local or exported closure already emitted into the same
module, refrain from adding an IND_STATIC closure, and
instead emit an assembly-language alias.

Inter-module IND_STATIC objects still remain, and need to be
addressed by other measures.

Binary-size savings on nofib are around 0.1%.

Document the ':set local-config' command and add a warning about
sourcing untrusted local .ghci scripts.

Related: ghc/ghc#6017
Related: ghc/ghc#14250

Add the ':set local-config { source | ignore }' setting to control
whether .ghci file in current directory will be sourced or not.  The
directive can be set in global config or $HOME/.ghci, which are
processed before local .ghci files.

The default is "source", preserving current behaviour.

Related: ghc/ghc#6017
Related: ghc/ghc#14250

Update users guide to match the new startup script order.  Also
clarify that -ignore-dot-ghci does not apply to scripts specified
via the -ghci-script option.

Part of: ghc/ghc#14689

Directives in .ghci files in the current directory ("local .ghci")
can be overridden by global files.  Change the order in which the
configs are loaded: global and $HOME/.ghci first, then local.

Also introduce a new field to GHCiState to control whether local
.ghci gets sourced or ignored.  This commit does not add a way to
set this value (a subsequent commit will add this), but the .ghci
sourcing routine respects its value.

Fixes: ghc/ghc#14689
Related: ghc/ghc#6017
Related: ghc/ghc#14250

    -O2 is the highest value of optimization.
    -O3 will be reverted to -O2.

I suspect this is why @simonmar wasn't notified of !706.

[skip ci]

Lint returns a pair (Maybe a, WarnsAndErrs). The Maybe monad
allows to handle an unrecoverable failure.
In case of such a failure, the error should be added to the second
component of the pair. If this is not done, Lint will silently
accept bad programs. This situation actually happened during
development of linear types. This adds a safeguard.

[ci skip]

This patch adds an Hadrian target "binary-dist-dir". Compared to
"binary-dist", it only builds a binary distribution directory without
creating the Tar archive. It makes the use/test of the bindist
installation script easier.

Fixes test T3807 when building with Hadrian.

Trac #16370

As noted in #16546 timeout uses asynchronous exceptions internally, an
implementation detail which can leak out in surprising ways.  Note this
fact.

Also expose the `Timeout` tycon.

[skip ci]

This trivial MR updates the CI badge in the readme to point to the
new CI on gitlab, rather than the very out-of-date badge from
Travis.

This should have used `grep -E`, not `grep -e`

This broke Marge by creating a second pipeline (consisting of only the
`lint-submods` job). Marge then looked at this pipeline and concluded
that CI for her merge branch passed. However, this is ignores the fact
that the majority of the CI jobs are triggered on `merge_request` and
are therefore in another pipeline.

Not only is it reliably failing due to #16574 but all of the quickly
failing builds also causes the Windows runners to run out of disk space.

When -a or --test-accept is passed, and if one runs the 'test' target, then
any test failing because of mismatching output and which is not expected to
fail will have its expected output adjusted by the test driver, effectively
considering the new output correct from now on.

When this flag is passed, hadrian's 'test' target becomes sensitive to the
PLATFORM and OS environment variable, just like the Make build system:
- when the PLATFORM env var is set to "YES", when accepting a result, accept it
  for the current platform;
- when the OS env var is set to "YES", when accepting a result, accept it
  for all wordsizes of the current operating system.

This can all be combined with `--only="..."` and `TEST="..." to only accept
the new output of a subset of tests.

Intermediate STG does not contain free variables which can be useful
sometimes. So adding a flag to dump that info.

Point users to the right URL

Hadrian's "binary-dist" target must check that the "docs" directory
exists (it may not since we can disable docs generation).

* simplifies registers to have GPR, Float and Double, by removing the SSE2 and X87 Constructors
* makes -msse2 assumed/default for x86 platforms, fixing a long standing nondeterminism in rounding
behavior in 32bit haskell code
* removes the 80bit floating point representation from the supported float sizes
* theres still 1 tiny bit of x87 support needed,
for handling float and double return values in FFI calls  wrt the C ABI on x86_32,
but this one piece does not leak into the rest of NCG.
* Lots of code thats not been touched in a long time got deleted as a
consequence of all of this

all in all, this change paves the way towards a lot of future further
improvements in how GHC handles floating point computations, along with
making the native code gen more accessible to a larger pool of contributors.

Using `ghc-prim <= 0.6.1` is somewhat dodgy from a PVP point of view,
as it makes it awkward to support new minor releases of `ghc-prim`.
Let's instead use `< 0.7`, which is the idiomatic way of expressing
PVP-compliant upper version bounds.

It should be entirely the responsibility of make/Hadrian to ensure that
everything that needs this flag gets it. GHC shouldn't be hardcoded to
assist with bootstrapping since it builds other things besides itself.

Reviewers:

Subscribers: TerrorJack, rwbarton, carter

GHC Trac Issues: #15548 -- progress towards but not fix

Differential Revision: https://phabricator.haskell.org/D5082 -- extract
from that

Hopefully fixes Windows Hadrian build.

GHC now works out-of-the-box (i.e. without any wrapper script) by
assuming that @bin@ and @lib@ directories sit next to each other. In
particular, its RUNPATH uses $ORIGIN-based relative path to find the
libraries.

However, to be good citizens we want to support the case where @bin@ and
@lib@ directories (respectively BINDIR and LIBDIR) don't sit next to
each other or are renamed. To do that the install script simply creates
GHC specific @bin@ and @lib@ siblings directories into:

   LIBDIR/ghc-VERSION/{bin,lib}

Then it installs wrapper scripts into BINDIR that call the appropriate
programs into LIBDIR/ghc-VERSION/bin/.

The issue fixed by this patch is that libraries were not installed into
LIBDIR/ghc-VERSION/lib but directly into LIBDIR.

Fixes #16052

When the offset in `setByteArray#` is statically known, we can provide
better alignment guarantees then just 1 byte.

Also, memset can now do 64-bit wide sets.

The current memset intrinsic is not optimal however and can be
improved for the case when we know that we deal with

(baseAddress at known alignment) + offset

For instance, on 64-bit

`setByteArray# s 1# 23# 0#`

given that bytearray is 8 bytes aligned could be unrolled into
`movb, movw, movl, movq, movq`; but currently it is
`movb x23` since alignment of 1 is all we can embed into MO_Memset op.

As noted in #16309 this somehow went undocumented.

Currently, the `hpc` submodule is pinned against the `wip/final-mfp`
branch, not against `master`. This pins it back against `master`.

We're might be about to switch to generating it in Hadrian/Make. This
reformat makes it easier to programmingmatically generate and end up
with the exact same thing, which is good for diffing to ensure no
regressions.

I had this as part of !712, but given the difficulty of satisfying CI, I
figured I should break things up even further.