The value indicating if the carry is useful wasn't taken into account.

On big-endian platforms executing

  import GHC.Exts
  data Foo = Foo Float# deriving Show
  foo = Foo 42.0#
  foo
  :print foo

results in an arithmetic overflow exception which is caused by function
index where moveBytes equals
  word_size - (r + item_size_b) * 8
Here we have a mixture of units. Both, word_size and item_size_b have
unit bytes whereas r has unit bits.  On 64-bit platforms moveBytes
equals then
  8 - (0 + 4) * 8
which results in a negative and therefore invalid second parameter for a
shiftL operation.

In order to make things more clear the expression
  (word .&. (mask `shiftL` moveBytes)) `shiftR` moveBytes
is equivalent to
  (word `shiftR` moveBytes) .&. mask
On big-endian platforms the shift must be a left shift instead of a
right shift. For symmetry reasons not a mask is used but two shifts in
order to zero out bits. Thus the fixed version equals
  case endian of
    BigEndian    -> (word `shiftL` moveBits) `shiftR` zeroOutBits `shiftL` zeroOutBits
    LittleEndian -> (word `shiftR` moveBits) `shiftL` zeroOutBits `shiftR` zeroOutBits

Fixes #16548 and #14455

The function `pprHsForAllExtra` was called only on `Nothing`
since 2015 (1e041b73).

Noticed by @Ericson2314

Before this patch BigNat names were confusing because we had:

* GHC.Num.BigNat.BigNat: unlifted type used everywhere else
* GHC.Num.BigNat.BigNatW: lifted type only used to share static constants
* GHC.Natural.BigNat: lifted type only used for backward compatibility

After this patch we have:

* GHC.Num.BigNat.BigNat#: unlifted type
* GHC.Num.BigNat.BigNat: lifted type (reexported from GHC.Natural)

Thanks to @RyanGlScott for spotting this.

Previously it wasn't uncommon to see +/-1% fluctuations in compiler
allocations on this test.

rts linker: teach the linker about GLIBC's special handling of *stat, mknod and atexit functions #7072

By default we don't link the RTS when linking shared libraries because in the
most usual mode a shared library is an intermediary product, for example a
Haskell library, that will be linked into some executable in the end. So we
wish to defer the RTS flavour to link to the final link.

However sometimes the final product is the shared library, for example when
writing a plugin for some other system, so we do wish the shared library to
link the RTS.

For consistency we also make -staticlib honor this flag and its inversion.
-staticlib currently implies -flink-shared.

Mostly symbols needed for aarch64/armv7l
and in combination with musl, where we have
to rely on loading *all* objects/archives

- __stack_chk_* only when not DYNAMIC

We shouldn't directly call 'genericWordMul2Op' in genericIntMul2Op
because a target may provide a faster primop for 'WordMul2Op': we'd
better use it!

The old name was and terrible and became worse after the previous
commit's refactor moved non-trivial funcationlity into its body.

Follow @simonpj's suggestion of pushing the "into regs" logic into
`emitPrimOp`. With the previous commit getting rid of the recursion in
`genericIntMul2Op`, this is now an easy refactor.

This unblocks a refactor, and removes partiality. It might be a PowerPC
regression but that should be fixable.

[ci skip]

Following a long conversation with Richard, this patch tidies up the
handling of return kinds for data/newtype declarations (vanilla,
family, and instance).

I have substantially edited the Notes in TyCl, so they would
bear careful reading.

Fixes #18300, #18357

In GHC.Tc.Instance.Family.newFamInst we were checking some Lint-like
properties with ASSSERT.  Instead Richard and I have added
a proper linter for axioms, and called it from lintGblEnv, which in
turn is called in tcRnModuleTcRnM

New tests (T18300, T18357) cause an ASSERT failure in HEAD.

d535ef00 allowed the use of up to 10
vanilla registers but didn't update LLVM backend to support them. This
patch fixes it.

* Mention missing flags that are not actually enabled by -Wall (docs/users_guide/using-warnings.rst)
* Additionally remove -Wmissing-monadfail-instances from the list of flags enabled by -Wcompat, as it is not the case since 8.8

Previously when not LDV profiling we would repeatedly reinitialise
`censuses[0]` with `initEra`. This failed to free the `Arena` and
`HashTable` from the old census, resulting in a memory leak.

Fixes #18348.

When not LDV profiling there is no reason to allocate 32 Censuses; one
will do. This is a very small memory footprint optimisation, but it
comes for free.

SCC profiling was enabled in a convoluted way: if WayProf was enabled,
Opt_SccProfilingOn general flag was set (in
`GHC.Driver.Ways.wayGeneralFlags`), and then this flag was queried in
various places.

There is no need to go via general flags, so this patch defines a
`sccProfilingEnabled :: DynFlags -> Bool` helper function that just
checks whether WayProf is enabled.

This is a leftover from ef63ff27

Before this patch we could generate addresses of this form:

   movzbl cP0_str+-9223372036854775808,%eax

The linker can't handle them because the offset is too large:

   ld.lld: error: Main.o:(.text+0xB3): relocation R_X86_64_32S out of range: -9223372036852653050 is not in [-2147483648, 2147483647]

With this patch we detect those cases and generate:

   movq $-9223372036854775808,%rax
   addq $cP0_str,%rax
   movzbl (%rax),%eax

I've also refactored `getAmode` a little bit to make it easier to
understand and to trace.

[ci skip]

Fixes #18279. Bumps the `text` submodule.

The way that `GHC.HsToCore.Quote` desugared quoted `via` types (e.g.,
`deriving via forall a. [a] instance Eq a => Eq (List a)`) and
explicit type annotations in signatures (e.g.,
`f = id @a :: forall a. a -> a`) was completely wrong, as it did not
implement the scoping guidelines laid out in
`Note [Scoped type variables in bindings]`. This is easily fixed.

While I was in town, I did some minor cleanup of related Notes:

* `Note [Scoped type variables in bindings]` and
  `Note [Scoped type variables in class and instance declarations]`
  say very nearly the same thing. I decided to just consolidate the
  two Notes into `Note [Scoped type variables in quotes]`.
* `Note [Don't quantify implicit type variables in quotes]` is
  somewhat outdated, as it predates GHC 8.10, where the
  `forall`-or-nothing rule requires kind variables to be explicitly
  quantified in the presence of an explicit `forall`. As a result,
  the running example in that Note doesn't even compile. I have
  changed the example to something simpler that illustrates the
  same point that the original Note was making.

Fixes #18388.