GitLab

Fix *two* bugs in formatRealFloat:

The first one is in the Haskell 98 errata, namely that negative
exponents cause an infinite loop in the FFFixed case when no precision
is specified.  eg. `Numeric.showFFloat Nothing 0.02 ""'  I've modified
the code in the errata to properly handle the e == 0 case and to be
slightly more efficient when e > 0.

The second bug is this:

   Prelude> Numeric.showFFloat (Just 0) 0.02 ""
   "0."

FCode/Code is now a monad, and thus now also a constructed type, rather than a
type synonym.

This requires quite a lot of changes in quite a lot of files, but none of these changes should have changed the behaviour of anything.

Being a Monad allows code that used FCode to be IMHO rather more readable
as it can use do notation, and other common Monad idioms.

In addition, state has been abstracted away with getter and setter
functions, so that functions mess with the innards of FCode as little as
possible - making it easier to change FCode in future.

- use SET_HDR rather than initialising header.info directly (fixes
  potential bugs with profiling).

- add some masking to the Int32/Word32 cases to match the Int8-16 and
  Word8-16 cases (potential 64-bit bugs).

Add genRange method to class RandomGen, as per the revised Haskell 98
library report.

Make the Permissions type non-abstract as per the revised Haskell 98 report.

update maximumBy and minimumBy in line with the revised Haskell 98 report

Changes to the Ix class from the revised Haskell 98 report:

  - Ord is no longer a superclass of Ix.

  - rangeSize is now a class member, as there are cases when
    it is useful to be able to override it.  As a result, GHC's
    performance-improving "unsafeRangeSize" function also has to be
    a class method just in case the programmer has overriden
    rangeSize.  Of course, unsafeRangeSize isn't visible when just
    importing Ix.

  - Added unsafeRangeSize bindings to all our standard Ix instances.

  - Improved the Ix instances for Int{8,16,32,64} and
    Word{8,16,32,64} by defining unsafeIndex instead of index, and
    providing a definition of unsafeRangeSize.

I hope I haven't mucked anything up :) The array tests all pass
successfully, except for arr016 which depended on Ord being a
superclass of Ix.  I'll commit changes to this test shortly.

bring PprType (and its Outputable Type instance) into scope

Change the definition of recip in the instance of Fractional for
Ratio, to:

	recip (x:%y)	=  y % x

as per the revised Haskell 98 report.

Bring scanr1, scanl, and scanl1 into line with the revised H98 report.

Add instances of Eq & Ord up to size 15 (Bounded, Read & Show will
have to wait)

Correct the USE_REPORT_PRELUDE definition of (!!).

----------------------------------------
	Make isFFIArgumentTy understand newtypes
	----------------------------------------

This fixes the bug Manuel reported:

	newtype T = T (Ptr T)
	foreign import ccall foo :: T -> IO (Ptr T)

  test.hs:6:
      Unacceptable argument type in foreign declaration: T


On the way, I moved isFFIArgumentTy and friends out of TysWiredIn,
where they didn't really belong, into TcType.  That in turn force
me to move isStrictType, and isPrimitiveType.

Add pprEquation

Fix (read "Infinity") and (read "NaN") for RealFloat types.

fix problem with -fno-code

shut up shortMeOut traces

use qualified names to indicate external status of values

document 31-bit core output option settings

a couple of tweaks to the hc-file-bundle target:
  - Add -r option to ${RM} to blow away an existing binary-dist
    temporary directory if present.
  - Fix bogons in the bundle building code.

foldr is a good consumer (but not necessarily a good producer)

correct off-by-one error in hGetStringBuffer

Fill in details for 5.02.

Make when comparing old and new strictness information,
we were being a bit too generous with the old case.
When a function has (say) arity 2, we can't use strictness
info like SSS, because that only applies if the fn is
applied to 3 args.  So it's an unfair comparison.

This commit makes the old->new conversion function more
truthful in this regard, which should eliminate the
erroneously-claimed "worse" strictness results.

wibble

------------------------------
	Another demand analyser glitch
	------------------------------

This fixes, I hope, (in an unsatisfactory way) another demand-analyser
glitch, that resulted in a runtime "entered absent arg" error.

In DmdAnal.annotateLamIdBndr, we now have:

	-- This gross hack is needed because otherwise we label
	-- a lambda binder with demand 'B'.  But in terms of calling
	-- conventions that's Abs, because we don't pass it.  But
	-- when we do a w/w split we get
	--	fw x = (\x y:B -> ...) x (error "oops")
	-- And then the simplifier things the 'B' is a strict demand
	-- and evaluates the (error "oops").  Sigh

The hack zaps 'B' to 'Abs', just like 'lazify' does.

Need to talk to John about what the principled reasons are.

Improve error message for mis-matched contexts

Fix an obscure but easy bug in SpecConstr

- document :info
- bring up to date info about FFI support

Remove unused import of DmdType

First cut at new compiler + interpreter changes for 5.02.  Also rearranged
the order of sections.

Comment only

Improve error message for nullary class

* Added a brief outline of the handling of command line options
* Revised Reuben's explanation re -no-hs-main according to our email exchange

Added SPJ's new -fmax-worker-args to the flag reference

Correct typos and minor errors.
Reorganized most of the sections, and the interface subsection appears first in each section.
Made a small change to Figure 1, the overall architecture of the storage manager.
The description on the configuration of the nursery is now correct.
Added a short section on the state of the heap allocator.

Don't say that GHC doesn't do fixity resolution on the left hand side
of a binding before deciding which symbol is the function symbol.
This has been fixed.

Rephrased a few paragraphs.
Added a paragraph on the interface to retainerSet.
Added 'To do' paragraphs, which may be useful for maintenance.

Fix representation finding for recursive newtypes

Note that support for generics is currently broken.