- Cvs-nhc98 - Haskell.org

patch applied (hat): Portable Hat: Fixed < < loop> > bug.
by Malcolm Wallace 10 Oct '06

10 Oct '06

Tue May 28 09:08:43 PDT 2002 olaf * Portable Hat: Fixed <<loop>> bug. Even for local constants uConstDef has to record something to avoid a black hole, because useDef pattern matches on its result. Hence use uwrapForward there. Reintroduced distinction between wrapForward and uwrapForward. The former always records a Forward. Know that for type conversion combinators the trace will always be demanded. Also replaced unFun by pattern matching on Fun. It was stupid previously to remove pattern matching. Because Fun is a newtype it doesn't force any evaluation and it is nicer and probably more efficient than unFun. Experimented with different definition of uwrapForward. The alternative is nicer, but slightly less efficient and hence only included in a comment. M ./src/hatlib/Hat.hs -51 +87

1 0

patch applied (hat): Allow pattern " f in g" to match even if g is inside a lambda within the
by Malcolm Wallace 10 Oct '06

10 Oct '06

Tue May 28 07:27:19 PDT 2002 malcolm * Allow pattern "f in g" to match even if g is inside a lambda within the definition of f. M ./src/hattools/observeutils.c -11 +28

1 0

patch applied (hat): In all hat tools:
by Malcolm Wallace 10 Oct '06

10 Oct '06

Tue May 28 07:24:04 PDT 2002 malcolm * In all hat tools: * Replace the `verbose' option with three separate new options: uneval, stringSugar, and listSugar. * Do projections correctly. * Right-associative operators (e.g. :) are now correctly pretty-printed without extra parens. (The decoding of L/R fixity from an Int was wrong in SExp). In hat-trail: * Fix the problem of textual overflow in the error/output panes. M ./src/hattools/HatObserve.hs -25 +50 M ./src/hattools/HatTrail.hs -66 +104 M ./src/hattools/Pattern.hs -3 M ./src/hattools/SExp.hs -123 +74 M ./src/hattools/TExp.hs -7 M ./src/hattools/artutils.c -1 +2

1 0

patch applied (hat): Slight tidy-up of list-sugaring.
by Malcolm Wallace 10 Oct '06

10 Oct '06

Wed May 22 06:35:39 PDT 2002 malcolm * Slight tidy-up of list-sugaring. M ./src/hattools/SExp.hs -3 +5

1 0

patch applied (hat): Reformat to 80 columns only.
by Malcolm Wallace 10 Oct '06

10 Oct '06

Wed May 22 05:21:50 PDT 2002 malcolm * Reformat to 80 columns only. M ./src/compiler98/Case.hs -59 +69

1 0

patch applied (hat): Do `extractExp' on field update expressions, to ensure that local
by Malcolm Wallace 10 Oct '06

10 Oct '06

Tue May 21 04:08:22 PDT 2002 malcolm * Do `extractExp' on field update expressions, to ensure that local let-expressions get their arity recorded. This avoids a compiler failure later on. Test case that previously broke the compiler: module Apa where data A = A{ field :: Bool } apa = A{ field = let s = False in s } M ./src/compiler98/Extract.hs -1 +6

1 0

patch applied (hat): Bugfix for negative literal number patterns, where the instance of
by Malcolm Wallace 10 Oct '06

10 Oct '06

Mon May 20 10:51:35 PDT 2002 malcolm * Bugfix for negative literal number patterns, where the instance of Num to be used is at a user-defined type. Must ensure that the Num dictionary is passed through to the pattern-compilation phase (for selection of Eq.==), rather than selecting the negate method from it here and immediately discarding the dict. M ./src/compiler98/FixSyntax.hs -1 +7

1 0

patch applied (hat): In hat-trans, if a module has no export list, produce an export list
by Malcolm Wallace 10 Oct '06

10 Oct '06

Mon May 20 07:16:17 PDT 2002 malcolm * In hat-trans, if a module has no export list, produce an export list containing only the stuff that should be exported. This avoids the problem of overlapping definitions of internally-generated identifiers (e.g. source positions) in a multi-module program. A little bit of a hack but should work. (Also noticed that a default declaration should always be deleted by the transformation). M ./src/compiler98/TraceTrans.hs -19 +37 M ./src/hattrans/TraceTrans.hs -19 +37

1 0

patch applied (hat): With recent fixes for when Hidden/Forward chains are generated,
by Malcolm Wallace 10 Oct '06

10 Oct '06

Fri May 17 07:28:55 PDT 2002 malcolm * With recent fixes for when Hidden/Forward chains are generated, getResult() could fall into an infinite loop caused by a cycle in the result chain. The fix is to remember the most recent Hidden node encountered, and if we trip over it again, stop the cycle and return the Hidden. M ./src/hattools/artutils.c -8 +25 M ./src/hattools/artutils.h +1

1 0

patch applied (hat): Explicit projections in the SExp structure are a nuisance, so don' t
by Malcolm Wallace 10 Oct '06

10 Oct '06

Fri May 17 07:25:36 PDT 2002 malcolm * Explicit projections in the SExp structure are a nuisance, so don't generate them any more. M ./src/hattools/SExp.hs -4 +7

1 0