Numerous tweaks to remove out-of-date constructs and make tests Windows-
friendly.

fix bug in new -ofile stuff: we were missing a '.' between the tmp
file root and the suffix.

Use get_itbl instead of direct access to info field in
removeIndirections. This "worked" only because the probability of
hitting an indirection-like closure by chance is 5:65531 on 32bit
platforms. :-) But HOpenGL once again managed to expose this bug...

- revised for ghc 4.07; added suggestions from Pixel <pixel@mandrakesoft.com>
- added profiling package
  [these are named ghc-prof-<version>.<arch>.rpm]

Don't use coercion RULES with litlits (e.g. int2Word ``42''), once
again found by *the* GHC test, HOpenGL + Quake level viewer. :-)

Include Weak.h to remove a warning

I've bitten the bullet and cleaned up the mess around Readline. There
are actually questions to answer by autoconf:

   * Are the headers available for compilation?
   * Are the libraries (and which) available for linking?

Note that in the second case it should be the libraries on the machine
which GHC is *installed* on, not the ones on which it was compiled.
GhcLibsWithReadline is now automagically set if the headers are found,
but this can still be overridden.

/tmp/msg

Bumped the maximum size of tuples from 37 to 62. If this is not
wanted, use PrelTup_HC_OPTS+=-DNO_LARGE_TUPLES in build.mk.

Further steps towards an empty build.mk:
   * GHC is now built with -O by default
   * hslibs/tools now use GhcLibToolsHcOpts, again -O by default

Add loop test

Remove a way to make the simplifier go into an
infinite loop.   This has been there for some weeks;
and George's UniForm tickled it.  I'm amazed nothing
else has done so.  I'll add a test.

The bad case happened when you go

	let xs = 1:xs
	in
	foldr k z xs

Then we kept firing the foldr/cons rule.

Solution: we don't attach an unfolding *at all* to
loop breakers (Simplify.completeBinding)

First I forgot to commit all the files for the
Java support.  Then I spoke the wrong CVS command.
This commit just adds the log message!

--------------------------
Add support for Java generation, written in
a lightning day with Erik Meijer

	ghc -J Foo.hs

will do the business, generating Foo.java

The code is in a new directory, javaGen/, so
you'll need to cvs update -d.

I've reorganised main/CodeOutput quite a bit; it
is now much much tidier, and will accommodate new
languages quite easily.

I've also fiddled with the flags that communicate
between the driver and hsc.

GONE:	-S=  -C=

NEW:	-olang=xxx	output language xxx
			xxx can be: C, asm, java

	-ofile=xxx	put the output code in file xxx


BEWARE that I might have broken some of the more
cryptic stuff in ghc.lprl.

Simon

And here is the commit for the driver

/home/simonpj/tmp/msg

Add support for Java generation, written in
a lightning day with Erik Meijer

	ghc -J Foo.hs

will do the business, generating Foo.java

The code is in a new directory, javaGen/, so
you'll need to cvs update -d.

I've reorganised main/CodeOutput quite a bit; it
is now much much tidier, and will accommodate new
languages quite easily.

I've also fiddled with the flags that communicate
between the driver and hsc.

GONE:	-S=  -C=

NEW:	-olang=xxx	output language xxx
			xxx can be: C, asm, java

	-ofile=xxx	put the output code in file xxx


BEWARE that I might have broken some of the more
cryptic stuff in ghc.lprl.

Simon

Finish TcClassDecl

To avoid potential collisions between different modules, prefix the
names of the C stubs with the module name.

I'm not sure if the last commit of MkId.lhs was incorrect or
incomplete. Anyway, to build hsc I've commented out the now
superfluous parameter for mkDictSelId (marked with "SUP:???").

Package util now depends on package posix, too.
NOTE: To get Readline up and working, -lreadline and one of
-lncurses/-lcurses/-ltermcap has to be inserted, too, but this has to
be autoconf-ed somehow (later...)

Update this file to the one on the web site, and add Simon's latest
notes about using the cvs- lists for bug reports about the cvs
sources.

Print out cost centre names as C strings (using proper escaping for
special characters etc.).

Improved rules for stubs. Not perfect yet, but better than before.

add	pprStringInCStyle :: String -> SDoc
(we already have a FAST_STRING version but not one for plain Strings).

Fix for initModules when unregisterised.

Dont strip directory prefix when printing src-locs

- Fix bug in TcExpr.tcMonoExpr (RecordUpd ...), where I hadn't
  propagated the recent change (to be H98ish) that record selectors
  for types with a context are overloaded:
	data Eq a => T a { f1 :: a }

  Here	f1 :: Eq a => T a -> a

  I don't like this, but Mark persuaded me that this was the
  Right Thing if we are to have contexts in data decls at all
  (which we should not)

wibble

Space leak fix from Simon P.J.

- Fix foreign export dynamic wibble (reported by Sven)

- Don't rm the .o file when splitting.   Doing so
  forces grotesque amounts of recompilation

- Arrange that -ddump-simpl-stats works whether or
  not DEBUG is defines (Ross Paterson's request)

- generate the heap profiling into <prog>.hp, as advertised
- clean up RtsFlags a bit

Profiling info for all the thunks herein.

Profiling description info for PAP and AP_UPD.

MUT_VAR needs some profiling description info.

Corrected typo in nfib example (/ => `div`).
Added description for hp2ps's -c/-y flags.

Fix a bug in previous commit, some .globls were getting thrown away.

enumFrom<blah> on Ratios changed slightly recently.  The current
behaviour is correct.

Push directives over literal chunks when attempting to move them to
the following chunk on x86.  Occasionally gcc generates a .glob
directive some distance before the symbol it refers to, and we were
ending up with a whole load of .glob directives attached to strings,
and duplicated in each .o file when splitting.

This change reduces the size of my libHSstd_p.a from 43M (!!!) to 9M.
I think this problem must have appeared with gcc 2.95.2, but it's a
little strange that I didn't notice it until now.