GitLab

Update the documentation on using DLL's from Windows, fixing several errors, in particular those relating to bug 3605

(1) More lenient kind checking, (2) Fixed orientation problems and avoiding double unifications, (3) Comments

We weren't doing fundeps for derived superclasses

Fixes a loop in the compiler, when running the dph tests

Contrary to the comment, for the module in #4367 at least, it is a big
improvement. Without it we get a huge spike of drag.

A bit yukky; see Note [Runtime skolems] in TcErrors.
But it works, and the debugger just is yukky in places.

In particular there is much less fiddly skolemisation now
Things are not *quite* right (break001 and 006 still fail), 
but they are *much* better than before.

Major pass through type checker:(1) prioritizing equalities, (2) improved Derived mechanism, (3) bugfixes

(1) Caching FD improvements for efficiency, (2) preventing cascading deriveds from entering the inert, (3) Fixing bugs in the creation of FlexiTcS variables

The bitmap type wasn't big enough to hold large bitmaps on 64 bit
platforms. Profiling GHC was segfaulting when retainStack was handling a
size 33 bitmap.

Reason: avoid having to link the entire ghc package in modules
that use compile-time annotations:

     import GHC.Exts( SpecConstrAnnotation )
     {-# ANN type T ForceSpecConstr #-}

It's a kind of bug that the package exporting SpecConstrAnnotation
is linked even though it is only needed at compile time, but putting
the data type declaration in GHC.Exts is a simple way to sidestep
the problem

See See Note [SpecConstrAnnotation] in SpecConstr

It now gets created by configure, rather than trying to work out the
version number at runtime.

This puts the ghc version into the package name so they are considered separate
packages.

We don't install dph etc, so don't put them in the doc index.

We weren't ignoring comment lines

struct _ObjectCode should be able to retain the name of archive members.
Though currently the only use of those names are for debugging outputs.

It's on by default (which matches the previous behaviour).

Motivation:
GLUT on OS X needs to run on the main thread. If you
try to use it from another thread then you just get a
white rectangle rendered. For this, or anything else
with such restrictions, you can turn the GHCi sandbox off
and things will be run in the main thread.

(which is the case in sdists)