First part of support for hierarchical module names:

   - the Finder now searches all possible paths when looking for
     a source file or .hi file.  I've removed the caching because
     now we have to search in subdirectories of each path option,
     and it was dubious whether the cache was actually helping.

   - the compilation manager now outputs a warning if it can't find
     the source for a given module, only the .hi file.  Previously
     this caused a cryptic error message when we attempted to call
     getModificationTime on the non-existent source file.

Track current changes in the text-mode-seek-on-Windows saga.

Try to improve matters in the text-mode-seek swamp.  Only disallow
text-mode-seeks on Win32, rather than on all platforms.  That means
that seeks in text mode on Unix platforms should work as originally
(which was the correct H98 behaviour), whilst Win32 generates
exceptions if you do seeks in text mode.

yet more wibbles

wibbles

Another -L<dir> for ghci wibble.

RTS-side stuff for making -L<dir> work on the ghci command line.

Haskell-side stuff for making -L<dir> work on the ghci command line.

wibbles

Further improve the $(SRCS) story, allowing you to exclude certain
sources (and all of their derived components) by setting
$(EXCLUDED_SRCS).

Add prototype for inputReady().

Update with more instructions on getting the IS setup working and notes on
why some of the less obvious binaries are included in the Windows dist.

Add comment about why gmp is configured differently on Windows.

Say that building packages as DLLs is now no longer supported, but may still
work.

Note in the documentation for :load that you can't load a package
module, and describe how we decide what the context is set to after a
:load.

Clean up the SRCS story.  This still won't fix the problem with
GhcLibsWithReadline=NO, but it's headed in the right direction.

Make c_read visible on Win32.

Updates for 5.01.

Calculate the correct set of modules to link in the case of upsweep
partial failure.

MERGE INTO 5.00.2

Improve instance Outputable Linkable (purely cosmetic).

ghc/rts/gmp/configure hangs up altogether on this
rm command:
    rm -f mpn/${tmp_fn}.[Ssc] mpn/${tmp_fn}.asm
I added lots of print statements to show progress
(hence the more verbose output) and found, I think,
that expanding the [Ssc] fixes the problem.

Add str002 (recursive newtype test)

----------------------------------
	Make exprArity give the right answer
	----------------------------------

		MERGE INTO 5.00.2

This fixes a fairly long-standing bug in exprArity, dating from
the time we allowed arguments to be non-atoms. If f had arity
2, it was saying that

	f (factorial x)

had arity 2, as well, which is plain wrong.

----------------------------------
	Fix an existential-constructor bug
	----------------------------------

		MERGE INTO 5.00.2

This fixes a long-standing bug that made the strictness
analyser go into a loop if it met a recursive newtype:

	newtype Void = MkVoid Void

Import wibbles

Add test for Manuel's finalizer bug.

Try doing a major GC before declaring that we're deadlocked, since
there might be finalizers waiting to run that can unblock the system.

Noticed by: Manuel.

Remove unused files.

Hmm, thought I removed this file.  Anyway, it's now been replaced by
PrelPosix.

Add storage mgt doc

Don't use exitWith where exceptions are treated as errors.

Don't panic on ExitException.

added support for emiting external core format

More tracking of IO lib changes.

Track recent changes to lib/std for mingw.  Mostly, make all seeked-on
file handles be in binary mode.

More fixups to make the new IO lib work on mingw.
* Outlaw changing the file position on a text-mode file.  After
  consideration of the mingw translation semantics I cannot see
  how to make a correct implementation.
* Add a field to Handle__ to say whether or not the handle is in
  binary mode.
* Restrict seek operations on Handles to those in binary mode.
* Export hSetBinaryMode from IO.lhs.

perl config: Remove bogus test I introduced

correct mistake in the documentation of index<blah>Array# and
index<blah>OffAddr#.  The documentation on primitives is all wildly
out of date in other ways, though.

- document -hidir option
- move -ohi documentation to "redirecting output" section
- remove duplicate -ohi in the flags table.
- add a warning about -ohi confusing the recompilation checker.

- add -hidir flag to control the destination for .hi files when doing
  multiple compilations (matches -odir).

- change the ml_hi_file component of ModuleLocation from Maybe FilePath to
  FilePath.  We never made use of its Maybeness.

- clear out some unused code from the Hsc phase of run_phase and clean up
  a bit.  Fix a bug with the -ohi option at the same time (I don't think
  it works in 5.xx.x before this patch).