GitLab

Oops, I bungled the -no-recomp fix.  This probably had something to do
with the fact that -no-recomp is perhaps the worst-named option in the
history of compilers, and the internal variable used to store the
value of -no-recomp actually stores (not -no-recomp), perhaps in an
attempt to reverse the confusion generated by the obsurely named
option.  In the end I figured I had a 50/50 chance of getting it
right...

Look for the source file first, so we don't miss it if there's an hi
file in the current directory.

Handle the R_SPARC_UA32 relocation type, as reported by Armin
Groesslinger.

/* According to the Sun documentation:
   R_SPARC_UA32
   This relocation type resembles R_SPARC_32, except it refers to an
   unaligned word. That is, the word to be relocated must be treated
   as four separate bytes with arbitrary alignment, not as a word
   aligned according to the architecture requirements.

   (JRS: which means that freeloading on the R_SPARC_32 case
   is probably wrong, but hey ...)
*/

update output

Test various forms of :info

Test that System.exitWith doesn't kill the interpreter

Go back to using a relative path name for the config.h include.
This is for the same reason that we do so in ghc/compiler/*, which
is that saying -I$(GHC_INCLUDE_DIR) confuses 4.08 series compilers.

Make local bindings work on the GHCi command line again.

make -no-recomp work with --make

comment reformatting only

-----------------------
	Fix a tyvar scoping bug
	-----------------------

This program:

	data T = forall a. T a (a->Int)

	f :: T -> T
	f (T (x::a) f) = T (undefined::a) f

gave the error

    Inferred type is less polymorphic than expected
    Quantified type variable `a' is unified with another quantified type variable `a'
    When checking a pattern that binds f :: a -> Int
    In the definition of `f': f (T (x :: a) f) = T (undefined :: a) f

This is of course bogus.  The fix is in TcMatches.tcMatchPats, where the
in-scope tyvars should be un-extended before calling tcCheckExistentialPat.

Use -ignore-dot-ghci for the GHCi tests.

Fix include of config.h

Further corrections to type validity checking

Teach it about MUT_VAR.

Another case of parenthesising operator names, this time in a ConDecl.

wibbles

Fix tycon assert failure

document the requirement for source files in --make.

document the requirement for source files in :load.

Accept The Truth and disallow loading any home modules which don't
have a source file.

The reason is that the compilation manager can't figure out
dependencies unless it has the source file to browse through -
previously there was a half-hearted attempt to let you get away with
just an object + interface file, but it didn't work in general.
Better just to disallow it altogether.

Note that this applies to --make too.

Remove unused function isSlash

Fix ghc-pkg to find package.conf when installed on Windows

Add a case so that tyConTheta works for tuples (made this change
earlier but forgot to commit it - it fixes the panic when you do :info
() in GHCi).

Allow renameFile to rename dangling symlinks.  We use lstat, but check
for HAVE_LSTAT first.

-------------------------------------
	Make NOINLINE zap the strictness info
	-------------------------------------

Make a NOINLINE pragma zap strictness information.
Reasons given in the WorkWrap comment:

	-- Furthermore, zap the strictess info in the Id.  Why?  Because
	-- the NOINLINE says "don't expose any of the inner workings at the call
	-- site" and the strictness is certainly an inner working.
	--
	-- More concretely, the demand analyser discovers the following strictness
	-- for unsafePerformIO:  C(U(AV))
	-- But then consider
	--	unsafePerformIO (\s -> let r = f x in
	--			       case writeIORef v r s of (# s1, _ #) ->
	--			       (# s1, r #)
	-- The strictness analyser will find that the binding for r is strict,
	-- (becuase of uPIO's strictness sig), and so it'll evaluate it before
	-- doing the writeIORef.  This actually makes tests/lib/should_run/memo002
	-- get a deadlock!
	--
	-- Solution: don't expose the strictness of unsafePerformIO.

This fixes the memo002 deadlock.

Add assertion

- be more informative about primitive types: they are identified as
  such, and a declaration is printed out as if the thing was an
  abstract data type (this at least tells you what the arity of the
  tycon is).

- print 'infix 4 elem' as 'infix 4 `elem`'.

Do something vaguely useful in ifaceTyCls when presented with a
primitive type constructor.  We pretend it's an abstract data type for
now.

sigemptyset: may be implemented as a macro on some plats,
so wrap it up in an inline function to get the C types
right. (Thomas Nordin, using OpenBSD, spotted this one).

This StablePtr test is old, most of the features it uses are
deprecated and have been replaced by newer mechanisms (eg. foreign
export) which is tested elsewhere.

Amplify comment on mkMachInt a little

Remove the in-range assertions on mkMachInt/mkMachWord.  They clearly
aren't true, because there's nothing stopping you from writing an
out-of-range Int# literal (although that's the only way I can see for
these to arise).

The wider issue is what should be done about out-of-range Int#
literals; I vaguely remember that at some point we disallowed them,
but I can't find anything in the logs.  The case which triggered the
assertion, namely "intToWord# 0xffff0000" would appear to be a
legitimate use for an out-of-range Int# literal though, given that
you can't write Word# literals directly.

I have braved the insanity of the Sparc calling conventions and lived
to tell the tale.  This commit adds support for foreign import
{static,dynamic} for Sparc in GHCi.

Commit a new autoheader-generated version of this file.

It isn't necessary to put HAVE_DLOPEN in here, autoheader can figure
that out for itself.

dunno what DATA_SECTION_WIBBLE is supposed to be, but it doesn't
appear anywhere else in the source tree so I'm removing it.

check for lstat()

Added two links to CVS

Remove the identity-substitution "optimisation" from zip_ty_env.

	-- There used to be a special case for when
	--	ty == TyVarTy tv
	-- (a not-uncommon case) in which case the substitution was dropped.
	-- But the type-tidier changes the print-name of a type variable without
	-- changing the unique, and that led to a bug.   Why?  Pre-tidying, we had
	-- a type {Foo t}, where Foo is a one-method class.  So Foo is really a newtype.
	-- And it happened that t was the type variable of the class.  Post-tiding,
	-- it got turned into {Foo t2}.  The ext-core printer expanded this using
	-- sourceTypeRep, but that said "Oh, t == t2" because they have the same unique,
	-- and so generated a rep type mentioning t not t2.
	--
	-- Simplest fix is to nuke the "optimisation"