GitLab

This fixes Trac #1204.  There's quite a delicate interaction of
GADTs, type families, records, and in particular record updates.

Test is indexed-types/should_compile/Records.hs

The stack of breakpoint resume contexts is now part of the
InteractiveContext and managed by the GHC API.  This prevents misuse
of the resume context by the client (e.g. resuming a breakpoint that
isn't the topmost, which would lead to a confused IC at the least).

I changed the TypeEnv in the IC to a [Id].  It only contained Ids
anyway, and this allows us to have shadowing, which removes an ugly
and annoying restriction.

The parts of the GHC API which deal with interactive evaluation are
now in a module of their own, InteractiveEval.

This was accidentally changed to 1 in the HEAD a while ago, the
behaviour is now the same as 6.6.x again.

See Trac #1310

Fixes Trac #1313

- Fix two distinct bugs, one in MkId.mkDataConIds, one in DataCon.mkDataCon
- Add more comments
- Add a little assertion checking in TyCon

Type-family tests now work.

Tom S found this bug by inspection; we've never seen it
fail in the wild, but it was obviosly Utterly Wrong

This was being done already for each binding added by :print,
but :sprint does not add any new binding, so we take care of it 
separately 

 - .tix files are now a list of MixModule, which contain a hash of the contents of the .mix file.
 - .mix files now have (the same) hash number.

This changes allow different binaries that use the same module compiled in the same way
to share coverage information.

It wasn't a good idea to disable it

Now we require that configure fails, and use our own signalling rather
than Cabal's buildable field.

This fixes a problem where patches altering the darcs-all script break
on Windows as the file is open. The script is now also slightly nicer,
on balance.

Sometimes parts of a source file come from different places due to
preprocessing, and the lines can overlap.

Now I can single step through Happy-generated parsers