Disable HPC in REPL for GHC

GHC doesn't support HPC in interpreter, see
https://downloads.haskell.org/~ghc/7.10.2/docs/html/users_guide/hpc.html#idp22264416

Fixes #2757

Fix deadlock in detailed-0.9 test suites

Fixes #2489. The detailed-0.9 test suite runner would deadlock because
the IO manager was not draining the output pipe of the child
process. Thanks to @r0ml for the patch.

This reverts commit f5020d41, reversing
changes made to 8d1cc528.

Fix windows build

Add frameworks when linking a dynamic library

This fixes the Cabal side of [GHC trac #10568](https://ghc.haskell.org/trac/ghc/ticket/10568)

Rename a bunch of functions of type 'Foo -> String' from 'debugFoo' to
'showFoo'.

Improve error message for unsatisfiable package constraints (issue #2643)

This commit adds the sources of constraints to debugging and error messages,
e.g., "main config file" or "command line flag".

Refactoring installplan

Support haddock response files

This change is for adding the ehancement to support haddock
response files.  This is what was used to build HP 7.10.2
for Windows, but it does need a corresponding change in
haddock to utilize it. Hopefully, this and its corresponding
haddock change can make it into the next ghc release and
eliminate much gnashing of teeth, error prone, and time
consuming manual steps in the HP build process.

Modified:

* Cabal/Distribution/Simple/Haddock.hs
  * renderArgs function: I put a couple of things into locals
    since I needed another use for UTF8 support check, plus I
    added another check based on version; the temp file logic
    was just as the prologue case above but I did need to
    repeat the invocation of the 'k' function in order to
    keep the cases separate and allow proper handling of the
    temp file automatic (or not, per --keep-temp-files)
    deletion.  Important: the haddock version being check
    against for response file support, greater than 2.16.1,
    is a placeholder and may or may not be the actual value
    since that will depend on the as-yet-unreleased haddock
    (which looks like it may be >2.16.1 but at this moment is
    not released).

Duncan Coutts authored 9 years ago and Mikhail Glushenkov committed 9 years ago

It wasn't used within the InstallPlan, but it had accessors and those
were used in a few places. Just pass them into those few places that
need it.

Duncan Coutts authored 9 years ago and Mikhail Glushenkov committed 9 years ago

So rather than the concrete InstalledPackageInfo and ConfiguredPackage,
the InstallPlan and PlanPackage are parameterised by the type of the
installed package, the source package and the types used for successful
and unsucessful build results.

Duncan Coutts authored 9 years ago and Mikhail Glushenkov committed 9 years ago

Fewer shared types makes the code easier to grok and navigate.

Duncan Coutts authored 9 years ago and Mikhail Glushenkov committed 9 years ago

No longer needed anywhere else.

Duncan Coutts authored 9 years ago and Mikhail Glushenkov committed 9 years ago

Removed from the InstallPlan and solver interface.
The InstalledPackage is really just there for the benefit of the old
TopDown solver which needs to know source deps.

Duncan Coutts authored 9 years ago and Mikhail Glushenkov committed 9 years ago

Rather than directly reusing the InstallPlan.PlanPackage type which has
more cases and which we'd like to generalise somewhat.

Duncan Coutts authored 9 years ago and Mikhail Glushenkov committed 9 years ago

The InstallPlan can be generalised by abstracting over the specific
package types. The only thing that really relies on a lot of the
details of the concrete ConfiguredPackage type is the bit that
validates them individually (as opposed to validating packages
within the plan in relation to other packages, graph structure etc).

So as a prelude to generalising the InstallPlan, move the checks
on the ConfiguredPackage into the Dependency module and use them
when checking the output of the solver.

Duncan Coutts authored 9 years ago and Mikhail Glushenkov committed 9 years ago

It was in the PackageIndex module but it was unused there and it
being there was just confusing.

This reverts commit bd7bbd70, reversing
changes made to d8fa4618.

Ensure that OS X framework headers are in include path

This fixes #2689.