GitLab

Following a suggestion of Claus Reinke, this patch improves the error
messages involving type functions.  Here's the relevant note from TcTyFuns.

Note [Non-injective type functions]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
It's very confusing to get a message like
     Couldn't match expected type `Depend s'
            against inferred type `Depend s1'
so pp_open_tc adds:
       NB: `Depend' is a (non-injective) type function

Currently we add this independently for each argument, so we also get
     Couldn't match expected type `a'
            against inferred type `Dual (Dual a)'
       NB: `Dual' is a (non-injective) type function
which is arguably redundant.  But on the other hand, it's probably
a good idea for the programmer to know the error involves type functions
so I've left it in for now.  The obvious alternative is to only add
this NB in the case of matching (T ...) ~ (T ...). 

See comment for details

                        MUT time (elapsed)       GC time  (elapsed)
  Task  0 (worker) :    0.00s    (  0.00s)       0.00s    (  0.00s)
  Task  1 (worker) :    0.00s    (  0.01s)       0.00s    (  0.00s)
  Task  2 (worker) :    0.00s    (  0.03s)       0.00s    (  0.00s)
  Task  3 (bound)  :    0.00s    (  0.00s)       0.02s    (  0.04s)

We still need to sort out the "install-docs" story to get this working
properly.

Now we have variables for whether or not to build the docbook docs as
HTML, as PS, and as PDF. The configure script output now matches what
the build system will do (except it cannot take account of any
mk/build.mk settings, of course).

We only need the forwardslash mode now, so always use that mode and
don't accept any arguments.

Also fixes it so that it really does pass (we weren't substituting the
GMP variables in the RTS package config).

configure looks for install-sh in . .. ../.. and we don't want it to
find the build system's install-sh.

No idea how the last patch managed to pass validate.