GitLab

Removed 32-bit assumptions in info table macros (mostly replacing
StgWord32 with StgWord).

Brought the mangler up-to-date for the Alpha.

Fixed: The second-stage compilation in this script should use the compiler
built in the first stage.  Also, configure wants to find Happy, so we feed
it fake-happy (presuming that the ghc-less machine also does not have happy).

lex_demand: added a couple of missing cases for new demand type

Rearranged the usage help message to move --no-compile to the bottom.

More importantly-- Fixed: hsc2hs was translating

    #def inline ... function(...) { ... }

into (assuming defined(__GNUC__))

    [in the _hsc.h file:]  extern inline ... function(...)
    [in the _hsc.c file:]  inline ... function(...)

while it should really be translated into

    [in the _hsc.h file:]  extern inline ... function(...)
    [in the _hsc.c file:]  ... function(...)

(without the "inline" in the _hsc.c file).  Quoting the gcc documentation
for "inline":

>    If you specify both `inline' and `extern' in the function
> definition, then the definition is used only for inlining.  In no case
> is the function compiled on its own, not even if you refer to its
> address explicitly.  Such an address becomes an external reference, as
> if you had only declared the function, and had not defined it.
>
>    This combination of `inline' and `extern' has almost the effect of a
> macro.  The way to use it is to put a function definition in a header
> file with these keywords, and put another copy of the definition
> (lacking `inline' and `extern') in a library file.  The definition in
> the header file will cause most calls to the function to be inlined.
> If any uses of the function remain, they will refer to the single copy
> in the library.

Added documentation for hsc2hs "--no-compile" option.

Add the *.a.list "list of object files to archive" file
to MOSTLY_CLEAN_FILES.

Improved how we build library archives:

On Digital UNIX (or where otherwise available), use the "Z" option to
compress archives as we build them.

On Digital UNIX 5 (or where otherwise available), use the "-input" option,
which reads from a specified file the list of objects to archive.  This is
significantly faster than piping the list of objects to xargs when the
number of objects to archive is large.

Fixed: making dependencies for Haskell source files fails if -split-objs
is present in $(HC_OPTS).

Added "--enable-hc-boot-unregisterised" option to configure,
for bootstrapping from unregisterised HC files.

Teach configure.in to canonicalise new-fangled Alpha system names,
such as alphaev67-dec-osf5.0, to alpha-dec-osf3.

Added a helper script "fake-happy" that fakes "happy -v" output.
It is useful for bootstrapping on new platforms.

Grab the latest versions of config.guess and config.sub from GNU.

ranlib gmp/libgmp.a - needed for a.out plats (OpenBSD in particular).

Handle leading underscores for PrelMain undefineds

Get rid off DEBUG - prev commit didn't attack the real problem

Back out change that tried to sneak in with the new garbage collector.

Add a compacting garbage collector.

It isn't enabled by default, as there are still a couple of problems:
there's a fallback case I haven't implemented yet which means it will
occasionally bomb out, and speed-wise it's quite a bit slower than the
copying collector (about 1.8x slower).

Until I can make it go faster, it'll only be useful when you're
actually running low on real memory.

'+RTS -c' to enable it.

Oh, and I cleaned up a few things in the RTS while I was there, and
fixed one or two possibly real bugs in the existing GC.

Wibble

Mismatched module name in non-DEBUG case

Grrr.  Do not #include "Stg.h" in plain C code, because that enables
all the global register variables.  #include "Rts.h" directly instead.

This was correct in the original StgLongLong.c, why on earth it was
changed is beyond me.

add a couple of missing ';'

import qualified Demand

---------------------------------
	Switch to the new demand analyser
	---------------------------------

This commit makes the new demand analyser the main beast,
with the old strictness analyser as a backup.  When
DEBUG is on, the old strictness analyser is run too, and the
results compared.

WARNING: this isn't thorougly tested yet, so expect glitches.
Delay updating for a few days if the HEAD is mission critical
for you.

But do try it out.  I'm away for 2.5 weeks from Thursday, so
it would be good to shake out any glaring bugs before then.

block/unblockAsynchExns have arity 2

Small changes to improve GC performance slightly:

  - store the generation *number* in the block descriptor rather
    than a pointer to the generation structure, since the most
    common operation is to pull out the generation number, and
    it's one less indirection this way.

  - cache the generation number in the step structure too, which
    avoids an extra indirection in several places.

The real version of this file is in ghc/includes/.

merge in changes from rts/Linker.h

Small fixes to the help text for the -i heap profiling interval option.

Yet another newtype-squashing bug; this time TcType.unifyTyX

ELF_TARGET_386 has to be #defined for x86/Solaris.

Solaris system headers contain function definitions (not just prototypes) for
some functions; eg, stat().  WDP decided to toss (some of) them out during
mangling.  I think, this is wrong, as the list of these functions changes with
varying Solaris releases and it makes foreign importing these functions more
complicated; in particular, it broke Directory.hsc and PosixFiles on
x86/Solaris.

ghc-inplace.bat: COMMAND.COM doesn't support %*, fall back to using %[1-9]

make ==> $(MAKE)

This commit adds the very convenient function

  Subst.substTyWith :: [TyVar] -> [Type] -> Type -> Type

and uses it in various places.

------------------------
	More newtype squashing
	------------------------

Recursive newtypes were confusing the worker/wrapper generator.
This is because I originally got rid of opaque newtypes altogether,
then put them back for recursive ones only, and forgot to reinstate
the cunning stuff in the w/w stuff.

(Discovered by Sigbjorn; thanks!)

-----------------------
	Get rid of ArityAtLeast
	-----------------------

Now that we have CgInfo, with the exact code-generator arity
for the value, we don't need the distinction between ArityAtLeast
and ArityExactly in the ArityInfo field of an IdInfo.

This commit makes

	type ArityInfo = Maybe Arity

and propagates this change consistently through the compiler.

Third cut at the demand analyser; seems to work nicely now

--------------------------------------------
	Hack around an infelicity in the simplifier
	-------------------------------------------

Use 3 iterations of the simplifier (instead of the previous 2)
because otherwise we land up with huge dead expressions.
Reason: an infelicity in the simpifier.

      let k = BIG in foldr k z xs
 ==>  let k = BIG in letrec go = \xs -> ...(k x).... in go xs
 ==>  let k = BIG in letrec go = \xs -> ...(BIG x).... in go xs
 Don't stop now!


I couldn't see an easy way to make the simplifier avoid this
big intermediate, so I hacked round it like this.

--------------------------
	Correct a bug in exprArity
	--------------------------

This long-standing bug meant that exprArity gave over-pessimistic
answers.  e.g.

	\x -> f x

where nothing is known about f.  We were getting

	0 (for f)
	-1 (for the argument)
	+1 (for the lambda)
	-----
	0

The right answer is of course 1.