Inline markup is supported in image links but, as per the [commonmark
recommendation][0], it is stripped back to a plain text representation.

  [0]: https://spec.commonmark.org/0.28/#example-547

The parser for pictures hasn't been properly adjusted yet.

Perf only change:

  * use 'getParserState'/'setParserState' to make 'Text'-optimized
    parser combinators
  * minimize uses of 'Data.Text.{pack,unpack,cons,snoc}'

Perf only change:

  * don't look up type variable names (they're never in the environment)
  * use a difference list for accumulating missing names
  * more efficient 'Functor'/'Applicative' instances for 'RnM'

Perf only change:

  * Add a 'SPECIALIZE' pragma to help GHC optimize a 'Data a =>' constraint
  * Manually specialize the needlessly general type of 'specializeTyVarBndrs'

Perf only change:

  * avoid needlessly union-ing maps
  * avoid synify-ing instances twice

Took this opportunity to add some docs too

This should fix #929, as well as guard against future problems of this
sort in other places. Basically replaces 'writeFile' (which selects the
users default locale) with 'writeUtf8File' (which always uses utf8).

Fix & add to documentation

The 1.6.1 release should've been a major bump, since types in
the `Documentation.Haddock.Parser.Monad` module changed. This version
makes that module internal (as it morally should be).

haddock-api 2.21.0 and haddock-library 1.6.1 release

Fix some version bounds in haddock-library too.

* Update CHANGELOGS
* Update new versions in Cabal files
* Purge references to ghc-8.4/master branches in README

Also silence orphan warnings.

* Add documentation for a bunch of previously undocumented
  options (fixes #870)
* Extend the documentation of `--hoogle` considerably (see #807)
* Describe how to add docs to `deriving` clauses (fixes #912)
* Fix inaccurate docs about hyperlinking infix identifiers (fixes #780)

This should have been in the previous patch, but wasn't.

Properly synify promoted list types

The other types are still looked at when considering whether to make
a kind signature or not.

We reconstruct promoted list literals whenever possible. That means
that 'synifyType' produces

   '[Int, Bool, ()]

instead of

   (Int ': (() ': (Bool ': ([] :: [Type]))))

Better rendering of unboxed sums, unboxed tuples, promoted tuples.

When we have a fully applied promoted tuple, we can expand it out properly.

* adds space after/before the '#' marks
* properly reify 'HsSumTy' in 'synifyType'

Matthew Pickering authored 6 years ago and Ben Gamari committed 6 years ago

Fixes #900

(cherry picked from commit e6aa8fb4)

Alec Theriault authored 6 years ago and Ben Gamari committed 6 years ago

The old approach to fixing #469, while correct, consumes a lot of
memory. We ended up with a HUGE 'GblRdrEnv' in 'ic_rn_gbl_env'. However,
'getNameToInstancesIndex' takes that environment and compresses it down
to a much smaller 'ModuleSet'.

Now, we compute that 'ModuleSet' explicitly as we process modules. That
way we can just tell 'getNameToInstancesIndex' what modules to load
(instead of it trying to compute that information from the interactive
context).

(cherry picked from commit 5c7c596c)

Accumulate explicitly which modules to load for 'attachInstances'

Ben Sklaroff authored 6 years ago and Ben Gamari committed 6 years ago

This token is necessary for parsing #line pragmas inside nested comments.

Reviewers: bgamari

Reviewed By: bgamari

Differential Revision: https://phabricator.haskell.org/D4935

Alec Theriault authored 6 years ago and Alex Biehl committed 6 years ago

In particular: we report the tests that crashed seperately from the tests
that produced incorrect output. In order for tests to pass (and exit 0),
they must not crash and must produce the right output.

Fixes #900