October 2006 - Cvs-nhc98

patch applied (hat): Correct bugs introduced by previous bug-fixes.
by Malcolm Wallace 10 Oct '06

10 Oct '06

Tue Apr 29 08:26:59 PDT 2003 olaf * Correct bugs introduced by previous bug-fixes. Malcolm Wallace reported that the hat-transformed prelude would no longer compile. There were two sorts of errors: Hat/PreludeBuiltin.hs:1: Type constructor or class not in scope: `Char' This was because import of the sort "TraceOrigPrelude" had been dropped by some inappropriate filtering of such imports. Now they are back again. Hat/PreludeBuiltin.hs:108: Variable not in scope: `Prelude.sempty' Calls in instances needed qualification when the class was imported qualified. However, the implementation always qualified the ids; when the class did not have a qualfifier it used Prelude. Now no qualfication happens in such a case. M ./src/hattrans/AuxFile.hs -6 +8 M ./src/hattrans/TraceTrans.hs -1 +3

1 0

patch applied (hat): Correct bugs in hx-file importing and renaming.
by Malcolm Wallace 10 Oct '06

10 Oct '06

Mon Apr 28 02:59:01 PDT 2003 olaf * Correct bugs in hx-file importing and renaming. All the fixes are however rather "hacky" and still leave loopholes. However, knowbody in their right mind should write Haskell programs that fall through the loopholes. Stephen Pitts reported the following problem with the Edison library. Transforming ListSeq and compiling it yields: Hat/ListSeq.hs:493: Ambiguous occurrence `anull' It could refer to either `Hat.ListSeq.anull', defined at Hat/ListSeq.hs:1261 or `Hat.PreludeBasic.anull', imported from Hat.Prelude at Hat/ListSeq.hs:23 ... and Hat/ListSeq.hs:996: Variable not in scope: `ssingle' ... The last errors are cause by code like instance Hat.S.Sequence (T.List) where gempty pempty p = T.constUse pempty p sempty sempty = T.constDef T.mkRoot a254v4empty (\ p -> gempty p254v12 p) The problem is that the class Sequence is only imported qualified, and hence even inside the instance declaration "sempty" is not in scope, only the qualified id "Hat.S.sempty". Now hat-trans qualfies such uses of shared variables with the same module id used in the qualification of the class id. In principal a method could be imported with a different qualifier than the class id... The first sort of error messages is caused by attaching the wrong hx-file information to identifiers in import (especially hiding) lists. First of all attaching of info is now done only with information from imported files, not with information of locally defined identifiers that may hide the imported ones (they do not hide them in the import declarations). Second, information for identifiers in "hiding"-lists should not be imported for attaching information in the program body, but is needed for attaching information to the identifiers in the hiding-list! So hidden identifiers are now imported separately. This is rather inefficient and probably scopes of imported identifiers are still not handled completely correctly. M ./src/hattrans/AuxFile.hs -6 +34 M ./src/hattrans/AuxLabelAST.hs -23 +28 M ./src/hattrans/TraceTrans.hs -10 +32

1 0

patch applied (hat): Remove spurious (non-H' 98) comma in named field structure.
by Malcolm Wallace 10 Oct '06

10 Oct '06

Fri Apr 25 06:34:44 PDT 2003 malcolm * Remove spurious (non-H'98) comma in named field structure. M ./src/hattools/SExp.hs -2 +2

1 0

patch applied (hat): Add a bunch of libraries that are (mainly) ghc-only, together with an
by Malcolm Wallace 10 Oct '06

10 Oct '06

Fri Apr 25 04:25:20 PDT 2003 malcolm * Add a bunch of libraries that are (mainly) ghc-only, together with an appropriate conditional in the Makefile. A ./src/hatlib/Control/Monad/Cont.hs A ./src/hatlib/Control/Monad/Error.hs A ./src/hatlib/Control/Monad/List.hs A ./src/hatlib/Control/Monad/RWS.hs A ./src/hatlib/Control/Monad/Reader.hs A ./src/hatlib/Control/Monad/State.hs A ./src/hatlib/Control/Monad/Trans.hs A ./src/hatlib/Control/Monad/Writer.hs A ./src/hatlib/Data/Monoid.hs M ./src/hatlib/Makefile -22 +18

1 0

patch applied (hat): Time to move the CVS version numbers up to nhc98-1.17 and hat-2.03.
by Malcolm Wallace 10 Oct '06

10 Oct '06

Fri Apr 25 03:57:46 PDT 2003 malcolm * Time to move the CVS version numbers up to nhc98-1.17 and hat-2.03. M ./Makefile.inc -2 +2

1 0

patch applied (hat): Store the actual value of a PackedString in the .hat file, not an
by Malcolm Wallace 10 Oct '06

10 Oct '06

Fri Apr 25 03:39:30 PDT 2003 malcolm * Store the actual value of a PackedString in the .hat file, not an abstract marker that just tells you it is a PS and nothing more. M ./src/hatlib/Hat/Data/PackedStringBuiltin.hs -1 +2

1 0

patch applied (hat): Comment out the Hat versions of unlinesPS, unwordsPS, joinPS,
by Malcolm Wallace 10 Oct '06

10 Oct '06

Fri Apr 25 03:38:22 PDT 2003 malcolm * Comment out the Hat versions of unlinesPS, unwordsPS, joinPS, because the released versions of ghc/nhc98 do not yet support them. M ./src/hatlib/Data/PackedString.hs +9

1 0

patch applied (hat): Correct bug in parent of traced application demanded by trusted function.
by Malcolm Wallace 10 Oct '06

10 Oct '06

Fri Apr 25 03:12:08 PDT 2003 olaf * Correct bug in parent of traced application demanded by trusted function. For example for main = print xs xs :: [Int] xs = map g [1,2,3] g x = 2 * x hat-trail showed Output: ----------------------------------------------------------------------- [2,4,6]\n Trail: ------- no src reference ----------------------------------------------- <- print [2,4,6] <- 2 * 1 <- g 1 <- xs That is, the parent of the traced application "g 1" is not as expected the application of the trusted function "map". The fault was an incorrect definition of the funi combinators, using the function parent as parent for the application instead of the context parent. Now it works correctly: Output: ----------------------------------------------------------------------- [2,4,6]\n Trail: ------- no src reference ----------------------------------------------- <- print [2,4,6] <- 2 * 1 <- g 1 <- map g [1,2,3] <- xs M ./src/hatlib/Hat/Hat.hs -15 +15

1 0

patch applied (hat): Add Control.Monad.Identity to the hat libraries (requested by Mark Carroll).
by Malcolm Wallace 10 Oct '06

10 Oct '06

Fri Apr 25 03:05:15 PDT 2003 malcolm * Add Control.Monad.Identity to the hat libraries (requested by Mark Carroll). A ./src/hatlib/Control/Monad/Identity.hs M ./src/hatlib/Makefile +1

1 0

patch applied (hat): Cover the case when a type synonym T is imported/exported with the T(..)
by Malcolm Wallace 10 Oct '06

10 Oct '06

Thu Apr 24 09:07:13 PDT 2003 malcolm * Cover the case when a type synonym T is imported/exported with the T(..) syntax. Treat it the same as when the entity is referred to as T or T(). M ./src/hattrans/TraceTrans.hs -3 +9

1 0