GitLab

This is a tricky case exposed by Trac #9254.  I'm surprised it hasn't
shown up before, because it's happens when you use unsafePerformIO in
the right way.

Anyway, fixed now.  See Note [Analysing with absent demand]
in Demand.lhs

Summary: Check for integer overflow in allocate() (#9172)

Test Plan: validate

Reviewers: austin

Reviewed By: austin

Subscribers: simonmar, relrod, carter

Differential Revision: https://phabricator.haskell.org/D36

Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>

Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>

Summary:
This patch is a prelude to implementation of hi-to-hi compatibility
checking.
Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>

Test Plan: validate

Reviewers: simonpj, austin

Subscribers: simonmar, relrod, carter

Differential Revision: https://phabricator.haskell.org/D35

For parallelism with the existing code. I don't think it should
make any difference.

This is slightly hackish, but hp2ps is already convoluted enough
that I don't feel bad about it.

This is the second attempt to add this functionality. The first
attempt was reverted in 950fcae4, due
to register allocator failure on x86. Given how the register
allocator currently works, we don't have enough registers on x86 to
support cmpxchg using complicated addressing modes. Instead we fall
back to a simpler addressing mode on x86.

Adds the following primops:

 * atomicReadIntArray#
 * atomicWriteIntArray#
 * fetchSubIntArray#
 * fetchOrIntArray#
 * fetchXorIntArray#
 * fetchAndIntArray#

Makes these pre-existing out-of-line primops inline:

 * fetchAddIntArray#
 * casIntArray#

This reverts commit 2f8b4c93

.
Signed-off-by: Austin Seipp <austin@well-typed.com>

Programmers may provide a pragma immediately after the `instance` keyword
to control the overlap/incoherence behavior for individual instances.
For example:

    instance {-# OVERLAP #-} C a where ...

I chose this notation, rather than the other two outlined in the ticket
for these reasons:

   1. Having the pragma after the type looks odd, I think.
   2. Having the pragma after there `where` does not work for
       stand-alone derived instances

I have implemented 3 pragams:

   1. NO_OVERLAP
   2. OVERLAP
   3. INCOHERENT

These correspond directly to the internal modes currently supported by
GHC.  If a pragma is specified, it will be used no matter what flags are
turned on.   For example, putting `NO_OVERLAP` on an instance will mark
it as non-overlapping, even if `OVERLAPPIN_INSTANCES` is turned on for the
module.

This seems a bit cleaner conceptually because the overlap mode and running
in safety mode are quite orthogonal.

More pragmatically, it also makes it possible to use `OverlapMode` to let
programmers pick the overlap mode for individual instances.

I'm wondering whether it's sensible to omit so many typecheck testcases from
the default validate test target. As for instance, TcNullaryTC has been failing
since its introduction in c63a4650 (re #8993

) and
it seems to have gone unnoticed so far.
Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

Some numbers have decreased but the haddock numbers have generally
increased noticeable again (see also last update in 970e5d99)

This updates all numbers I noticed in the "fast" test-mode, *except* for
the T9203 test-case on 32bit, which needs more investigation before
bumping due to its significant increase:

  bytes allocated value is too high:
      Expected    bytes allocated: 50000000 +/-5%
      Lower bound bytes allocated: 47500000
      Upper bound bytes allocated: 52500000
      Actual      bytes allocated: 85093548
  *** unexpected failure for T9203(normal)

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

In 1c0b5fdc (re #9224

) `BinaryLiterals`
was temporarily added to T4437's `expectedGhcOnlyExtensions` list. This
can now reverted as Cabal has been made aware of `BinaryLiterals`
(see haskell/cabal#1970 and haskell/cabal#1972).

updates Cabal submodule
Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

Now that the `libraries/dph` submodule is checked out always we need
a different way to disable building DPH to save compile-time while
developing GHC.

This commit adds a new YES/NO Make variable `BUILD_DPH` that can be used
inside mk/build.mk to control whether to build libraries/dph or not.
The default setting is `BUILD_DPH=YES` (via `mk/config.mk.in`).

This also changes `validate`'s flag `--no-dph` to explicitly disable DPH
for the current validation run.
Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

Test Plan: successful validates with `--fast --no-dph`

Differential Revision: https://phabricator.haskell.org/D31

It's a bit confusing to have .gitignore files spread all over the
filesystem. This commit tries to consolidate those into one .gitignore
file per component. Moreover, we try to describe files to be ignored which
happen to have a common identifying pattern by glob patterns.
Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

Summary:
What this fix does is reorder how we look for hi-boot files: we
unconditionally check for an hi-boot file, and if we don't find one, we
check the import graph to see if there was circularity.  This is as
opposed to the previous scheme (check for circularity, then load hi-boot
file).

This costs us an extra file system access every typecheck, which
is not the best.
Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>

Test Plan: Validate and check for compiler regressions in nofib

Reviewers: simonpj, austin

Subscribers: simonmar, relrod, carter

Differential Revision: https://phabricator.haskell.org/D30

Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>

Haskell2010 supports

- base-10 (prefix-less),
- base-8 (via `0[oO]`-prefix), and
- base-16 (via `0[xX]`-prefix) integer literals.

This commit adds syntax support for base-2 integer literals via the new `0[bB]`
prefix. The use of a `0b` prefix for indicating binary literals is known
from popular programming languages such as C++14, Perl, Python, Ruby, and Java.

This syntax extension is disabled by default and can be enabled via the
new `{-# LANGUAGE BinaryLiterals #-}` pragma and/or the new `-XBinaryLiterals`

This new extensions requires to upgrade the `ExtsBitmap` type from
`Word` to `Word64` as this adds a 33th flag which is not guaranteed to
fit into a `Word`.
Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

Differential Revision: https://phabricator.haskell.org/D22

This reverts commit 05120ecd.

Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>

This introduces an Enum type `ExtBits` for the currently used 32 flags
and introduces a type-synonym `ExtsBitmap` for representing a set over
`ExtBits`. While at it, the current `Int` was replaced by `Word` to
have the compiler catch any missed use-sites.

This will make it easy to swap the `Word`-representation of `ExtsBitmap`
by something different, such as e.g. a `Word64`

Test Plan: successful validate

Differential Revision: https://phabricator.haskell.org/D23

Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>

This was shown up by Trac #9222.

Just a minor refactoring

Also set `submodule.<name>.ignore=none` explicitly for the recently
converted submodules, as those are not supposed to have untracked/unignored
files lying around.
Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

Summary:
Normally, -fno-code does not generate interface files.
However, if you want to use it to type check over multiple
runs of GHC, you will need the interface files to check
source files further down the dependency chain; -fwrite-interface
does this for you.
Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>

Test Plan: clean validate, and a new test-case

Reviewers: simonpj

Subscribers: simonmar, relrod, carter

Differential Revision: https://phabricator.haskell.org/D27