even more h98 updates, this time to codeGen tests

updates and tweaks

more h98 updates

h98 updates

Updated

Updated for H98

More Enum regression tests than you can shake a stick at.

Updated to work with current sources

Fixes to {word,int}64ToIntegerZh_fast

docfix

Added comment next to the I*_ macros to explain why having them
expand to use 'static' instead of 'extern' is a Really Bad Idea.

integerToInt64Zh: didn't handle negative integers particularly well

Bunch of bug fixes related to Enum instances. Code clean-up.

Move ST <--> IO coercion functions from IOExts to (Lazy)?ST

Removed workaround for bug - fixed (a while ago.)

Changed the various Ix.range methods, to specifically check whether we're
dealing with an empty range or not. As it was, empty array weren't handled
properly.

Tag all 'foreign import's as being unsafe.

do_if_stmt: insert extra (I_) to avoid minInt blues.

Parser didn't accept the (all-important) empty module. Does now.

Updated to work with current sources

ifaceKeywordsFM: adding missing __ccall_GC entry

Fix include typo.

Fix more uses of [n..m]

Fix a use of [ e1..e2 ] in the light of the new Haskell 98 semantics.

Fix two uses of [ e1 .. e2 ] in light of the new Haskell 98 semantics.

Enum instances updated to comply with the behaviour that
Haskell 98 specifies. Notable changes,

  * [a..b] is not the empty list when a>b any longer.
  * [x..] and [x,y..] for Enum Int are now bounded lists.

The first change is might be worth bearing in mind when converting
1.4 code to Haskell 98; functions may have made use of the
old behaviour.

Robustified derived toEnum methods; gives better error message when
passed enum tags < 0.

wibble

dsExpr: make sure special case for unboxed tuple patterns only
'fire' when the tup. pat. is unboxed.

Removed unused args from Id.mkIdVisible and Name.mkNameVisible.
Perpetuated Name.mkTopName hack in Name.mkNameVisible.

The temp file where cpp's output was output wasn't being properly removed

With Haskell98, Enum.enumFromThen can generate sequences that
are either increasing, decreasing or infinite.

Resurrect ticky-ticky profiling.  Not quite polished yet, but it
compiles and produces some reasonable-looking stats.

Add $(HC_OPTS) to $(CC_OPTS) - this is necessary to get the per-way
flags into the plain-C compilations.

Make use of -F<factor> flag in the two-space collector.

The BLACKHOLEs created when entering a CAF for the first time are now
CAF_BLACKHOLES, distinguishing them from the blackholes created by
lazy blackholing.

This enables the lazy blackholing algorithm to be optimised by not
blackholing a section of stack that has already been traversed.

Increase CmdLineOpts stack size to 4M.

hook in GetOpt docs

Back again - somehow got lost.

Dead code removal