GitLab

Stylistic changes

Get all project info from configure.ac

Make it compile with a GHC 6.x. The layout rule seems to be a mystery
even for one of the "fathers" of Haskell...  :-}

Add HsPat.hi-boot

------------------------------------
	Add Generalised Algebraic Data Types
	------------------------------------

This rather big commit adds support for GADTs.  For example,

    data Term a where
 	  Lit :: Int -> Term Int
	  App :: Term (a->b) -> Term a -> Term b
	  If  :: Term Bool -> Term a -> Term a
	  ..etc..

    eval :: Term a -> a
    eval (Lit i) = i
    eval (App a b) = eval a (eval b)
    eval (If p q r) | eval p    = eval q
    		    | otherwise = eval r


Lots and lots of of related changes throughout the compiler to make
this fit nicely.

One important change, only loosely related to GADTs, is that skolem
constants in the typechecker are genuinely immutable and constant, so
we often get better error messages from the type checker.  See
TcType.TcTyVarDetails.

There's a new module types/Unify.lhs, which has purely-functional
unification and matching for Type. This is used both in the typechecker
(for type refinement of GADTs) and in Core Lint (also for type refinement).

Add -I- after the include paths when running gcc.  This prevents
accidental shadowing of system includes by putting a file called
eg. stdint.h in the current directory.

absCSyn directory doesn't exists in the HEAD

Look for getpwnam, getpwuid.

Mac OS X:
Fix a serious bug in the linker's handling of Mach-O's
GENERIC_RELOC_VANILLA relocations.

MERGE TO STABLE

Remove unused defnition

revise the description of rebindable syntax to match what it seems to do

rts_evalStableIO: set rtsApiCapability to NULL *before* calling
scheduleWaitThread, matching the way the other eval_* functions do
this.

The previous way lead to a suble race condition:
  - thread A calls rts_evalIO, enters scheduleWaitThread()
    (rtsApiCapability == NULL).
  - thread B calls rts_evalStableIO, creates a main thread and enters
    scheduleWaitThread() (rtsApiCapability == &MainCapability)
  - thread A exits scheduleWaitThread, sees that rtsApiCapability is
    non-NULL, and calls releaseCapability() on it.  This is bogus,
    because thread A doesn't actually hold the capability, and we've
    done a double-release.

This scenario leads to assertion failures in a debug threaded RTS, and
probably crashes in a non-debug threaded RTS.

MERGE TO STABLE

Avoid some warnings from gcc by appending an 'LL' suffix to 64-bit
integers on a 32-bit platform.

Suggested by: Sven Panne <Sven.Panne@aedion.de>

oops, revert version number to 6.3 again

Ignore RULES pragmas unless we're in -fglasgow-exts mode.

Mac OS X: Fix an incorrect ASSERT

Merge to STABLE

GHCi support for PowerPC Linux, courtesy of Ryan Lortie <desrt@desrt.ca>.

MERGE TO STABLE

A few updates to the local Emacs variables in these files.  For some
reason, I can't get Emacs' PSGML mode to work with the new XML files
(although I thought I had it working before... strange).

sed-magic for ProjectVersionInt: make it work for version numbers with
patchlevel components (ToDo: ProjectPatchLevel still isn't set
properly in version.mk, this has to be updated by hand).

export errMsgContext

PowerPC/Mac OS X:
Improve epilogue mangling.
This became necesssary because the new CG may "fall through" for cases in
a switch that are known to never happen, so GCC generates the epilogue code
again. With the previous CG, GCC never generated any epilog code, because
every function ended with an unconditional tail jump.

Limit ThreadLabel API use to DEBUG only.

* Replace obsolete macro AC_TRY_RUN with AC_RUN_IFELSE
* Same for AC_TRY_LINK and AC_LINK_IFELSE
* Minor cleanup

* Added new target for creating *.bz2 source distributions
* Cleanup *.tar.gz / *.tar.bz2 when things go wrong

Makefile.config is dead, so don't include it in SRC_DIST_FILES.

Make sure "docs" and "dist" exactly once in SRC_DIST_DIRS.

MERGE TO STABLE

Removed dependency of $(TOP)/mk/config.mk on $(TOP)/config.status, resurrecting
"make distclean". I'm not 100% sure that the autoconf-related rules are correct,
"./config.status --recheck" in conjunction with make is a real brain-twister...

Ooops, forgot to rename version.mk in my previous committ...

Make tests run on Cygwin hosted MinGW32 build.

Use version info from configure.ac

Use some sed magic to avoid PACKAGE_FOO clashes introduced by the autoconf
cleanup yesterday. Another option would have been using a plain "AC_INIT" in
fptools/configure.ac (and probably the same for fptools/libraries/configure.ac),
so only "leaf configure.ac"s have package info. Not quite sure what would be the
nicest way to handle this yet...

Make autoupdate 2.52 happy, mainly by using the new formats of AC_INIT and
AC_OUTPUT. This has the nice side effect that all "packages" have now a name, a
version, a bug-report address, and a tar name, yielding better output with
"configure --help=recursive". Nuked an unused AC_STRUCT_ST_BLKSIZE test on the
way.

Remove debugging trace that I left in when working on mk_switch.

New option, -P, turns off line pragma emission

Collect unknown chunks in new chkcat 'unknown' which gets treated like
'misc', but isn't mangled on sparc-*-solaris2 because of function
definitions in C-headers like <pwd.h>. This shouldn't affect any
other platform since this was already a border-case.

Closes SF Bug 1012521 "getUserEntryForID dies".
Please merge.

On Windows default to main.exe instead of a.out.

arrow notation: allow arrow applications (f -< a) to take a non-empty
command stack, as suggested by Sebastian Boldt <Sebastian.Boldt@arcor.de>.

- Remove configure tests for SIG{POLL,INFO,WINCH}:
  Testing via #ifdef SIGFOO should be sufficient.
- Change #if HAVE_SIGPOLL to #ifdef SIGPOLL
- Remove SIGINFO/WINCH from package base: they'll reappear in package unix
  in System/Posix/Signals.Exts.

Oops, fix bugs in previous commit.

Add a couple of cases to mk_switch to catch the case when we have a
tag range that has a lot of default cases at either end, and we're not
using a single switch.  In situations like this we want to eliminate
the default cases with an if-statement, before dealing with the rest
of the branches, which might then be suitable for a switch.

Also, ignore empty tag slots at either end of the range if there is no
default case.

This might work around a gcc 2.95 bug that we tickled with the
code being generated before.