when mangling gcc output, make sure we distinguish between a \ at EOL and a \ in DOS paths

updated

Reg. tests for h{Set,Get}Posn + hSeek

Testing the sanity of Show on Addr

Moved Show instance for Addr from PrelAddr to Addr & made it display
the Addr in hex (+ have it cope with ptrs with MSB set.)

Fix for compiling w/ ghc-2.10

Added 'updateIORef :: IORef a -> (a->a) -> IO ()'

- fix bug in setSocketOption__ which meant that trying to set
  SO_REUSEADDR on Linux (and possibly other OS's) didn't work.

- add rudimentary non-blocking connect support.

Need sys/types.h for off_t

exts/ is now also a _ccall_ free zone.

* Re-exported IO.HandlePosn, i.e.,

         type HandlePosition = Integer
	 data HandlePosn = HandlePosn Handle HandlePosition

* Added hTell :: Handle -> HandlePosition
  (merely a wrapper for IO.hGetPosn )

* Added hSetBinaryMode :: Handle -> Bool -> IO Bool
  for dynamically changing the 'translation mode' of a Handle.
  This stuff is only useful on platforms that make a distinction
  between text and binary files (e.g., Win32)

IOExts.hSetBinaryMode wrapper

updates

wibble

Allow seeking on devices and beyond EOF

Have sizeof_time_t report the size in *words*

unify two near-identical impls of toClockSec()

* drop the restriction that seeks cannot be made on devices & beyond
  EOFs. If the underlying lseek() doesn't like us doing either, it'll
  let us know.
* When asking for the current position under Win32, take into account
  that lseek() reports the _untranslated_ position, so adjust the resulting
  position by scanning input buffer looking for \n's (and treat them
  as if \r\n.)

re-organised the initialisation of the 'flags' a little; record whether a file handle is in 'binary' translation mode or not.

Drop the use of _ccall_, _casm_ and lit-lits in std/, "foreign import" is
the future.

nullAddr: avoid the use of a lit-lit

Consistently use .dll_o for toplevel object files that are to be included with all DLL-based apps

expected (std)output of show001

Testing Show for Maybes and Eithers

Fixed bogus Show instances for Maybe and Either

DLL building tweaks, including terrible hack to compile PrelMain

posix is not for Win32

Non-blocking I/O isn't supported on 'pure' Win32 platforms.

[This is the commit message that should have gone with the
 typechecker commit a few minutes ago, whose message said
	~/tmp/msg1
]

Improve location and context information on Insts (notably Dicts),
and thereby improve error messages that relate to dictionaries.

This arose from Marc van Doguen's suggestion.

This bunch of commits represents work in progress on inlining and
worker/wrapper stuff.

Currently, I think it makes the compiler slightly worse than 4.04, for
reasons I don't yet understand.  But it means that Simon and I can
both peer at what is going on.

* Substantially improve handling of coerces in worker/wrapper

* exprIsDupable for an application (f e1 .. en) wasn't calling exprIsDupable
  on the arguments!!  So applications with few, but large, args were being dupliated.

* sizeExpr on an application wasn't doing a nukeScrutDiscount on the arg of
  an application!!  So bogus discounts could accumulate from arguments!

* Improve handling of INLINE pragmas in calcUnfoldingGuidance.  It was really
  wrong before

/home/simonpj/tmp/msg1

Remove SST.lhs

Add drvrun007, a test that crashed Hugs

toInt wibble

Added expected stdout to reg. test for Float&Double exceptional IEEE values.
(the numbers were produced on a PII & with the MS C run-time -- don't be
surprised if you see slightly different numbers on your box.)

checking Main.main's type + newtype def that caused a ghc-X (X<4.03, I think) TC crash

Added Tree example from (older) versions of the Report

Extended to show why changing the defn of ReadS (and read) is a good idea

newtypes with a labelled field caused DS trouble sometime back

Bunch of H98 tests I had lying about