Cleanup of non-blocking I/O

- file descriptors are now always set to non-blocking mode.

- we don't do an inputReady operation on descriptors before
  attempting to read from them any more.

- the non-blocking flag on Handles has gone.

- the {set,clear}[Conn]NonBlockingFlag() functions have gone.

- the socket operations have been made to work properly with threads:
  accept is now non-blocking (it does a threadWaitRead instead of
  blocking), and the file descriptors returned by accept are set to
  non-blocking mode.

Win32 will need some adjustments, no doubt.

Fix for DEBUG && !PROFILING

mingw32 support

NumExts.showListWith doc'ed

Main trunk is now 4.05.

Minor Win32 changes

Magic Constant Removal

Minor tweaks to do with DLL building

Disable timer ticks on 'pure' Win32 platforms unless really needed
(for profiling) -- no reason to enable it elsewhere as the Win32 bits
doesn't currently support thread{Delay,Wait{Read,Write}}#

Move DllMain() into separate file + it doesn't call startupHaskell()
any longer upon loading of the DLL. That is the task of clients of
the RTS.

Add cmp_thread

Code that trundles over stable_ptr_table need to start at index 1,
as index 0 is unused (and may contain garbage.)

When constructing vanilla modules from ModuleNames, consult the HiMaps
to make sure we're using the 'right' kind of Module (i.e., it has got
the DLL flag correctly set.)

mkVanillaModule: made DLL savvy

Cost-centre heap profiling and symbol table heap profiling are now
merged into one file: ProfHeap.c.

Heap Profiling
--------------

This commit adds simple hp2ps-style heap profiling.  Just the '-h'
option is supported so far.

Heap profiles will show the top two cost centres on each cost centre
stack.

Time/Alloc profiles have been extended to aggregate the costs on a
per-cost-centre basis and show the table of highest consuming cost
centres before the full table of cost centre stacks.

Profiles work best if '-caf-all' is used, especially when using
one of the '-auto' flags.

Extra words about unsafePerformIO

A couple of fixes and cleanups to ticky-ticky profiling:

	- remove UPD_EXISTING (doesn't make sense)
	- add UPD_CON_IN_PLACE, now that we have in-place updates
	- clean up the output a little.

gcc 2.95 on Sparc changed the assembly output slightly.  This should
fix it.

GHC's web pages have moved to http://www.haskell.org/ghc/.

updated

Only use initUserSignals() if you've got it defined

GMP only depends on libiberty for the implementation of random().
Avoid this dep. on a (pure) Win32 platform.

Win32 version of awaitEvent(). Easy :)

oops, back out last change.  It was already there.

Fix building with profiling.

In case of a partial writes, buffer pointer wasn't being adjusted/used.

DLL install support

If DLL support enabled, add import libs to INSTALL_LIBS + DLL to INSTALL_PROGS

inputReady() and mingw - always return 1, not 0

Export toInt and fromInt from Int as well as from the (now uncool) rag-bag
that is GlaExts, since Hugs supports them too.

s/ZONE/get_ZONE/g

Re-ordered code so that the POSIX-blessed time() is used in preference
to the plethora of other clock reading calls. This *should* be an OK and
non-upsetting change, but idiosyncracies of OS-supplied header files may
just force me to rethink. We'll see..

Let's try again - added get_ZONE(), which retrieves current
timezone string. Need to do this since cygwin doesn't initialise
its tzname[] properly (=> need to call tzset() explicitly.)

timezone.h

Experimental support for re-routeing I/O on a Handle to that of another
for the duration of an IO action. Used to provide the following three
(exported) actions

   withStdin, withStdout, withStderr :: Handle -> IO a -> IO a

which 'replaces' one of the standard handles with that of the supplied
handle while performing the second action.

Documented Select interface ; removed old Pretty doc

Added hSelect

Fix run queue corruption bug, introduced in recent threadWait commit.

This fixes the "memo002 goes off into outer space" bug.

fromInt stuff