October 2006 - Cvs-nhc98

patch applied (hat): Treat a ForeignObj returned via FFI correctly. Normally, it is disallowed
by Malcolm Wallace 10 Oct '06

10 Oct '06

Thu Oct 19 09:07:17 PDT 2000 malcolm * Treat a ForeignObj returned via FFI correctly. Normally, it is disallowed altogether, but we need one instance so that 'makeForeignObj' works. M ./src/compiler98/Foreign.hs -9 +3

1 0

patch applied (hat): Oops, need to restrict import of PackedString to avoid clashes with
by Malcolm Wallace 10 Oct '06

10 Oct '06

Thu Oct 19 06:19:35 PDT 2000 malcolm * Oops, need to restrict import of PackedString to avoid clashes with redefined concat, map, foldr, etc in nhc98's own implementation of the PackedString library. M ./src/compiler98/FFITrans.hs -1 +1

1 0

patch applied (hat): A big re-organisation of I/O. Two main changes:
by Malcolm Wallace 10 Oct '06

10 Oct '06

Wed Oct 18 05:43:27 PDT 2000 malcolm * A big re-organisation of I/O. Two main changes: (1) Foreign functions of IO type (going via the primitive FFI) are now handled via wrapper functions _mkIOokN (where N is the arity) rather than by the special bytecode MKIORETURN. (2) The internal runtime system representations of file handles and ForeignObjs have been made simpler. ForeignObjs are now the basic mechanism, with file handles as just one instance of a ForeignObj. Between them, these solve several bugs: * a shared IO computation was only executed once * an IO CAF (i.e. zero-arity) was only executed once whether shared or not * tracing of foreign IO was also suspect for the same reasons * hClose'd files were not being deallocated quickly enough, leading to some I/O-intensive programs hitting the file allocation limit unnecessarily * it was too difficult for users to get hold of a file handle as an ordinary value to use in foreign functions M ./src/compiler98/Case.hs -1 +1 M ./src/compiler98/DbgDataTrans.hs -22 +10 M ./src/compiler98/DbgTrans.hs -2 M ./src/compiler98/Extract.hs -2 +1 A ./src/compiler98/FFITrans.hs M ./src/compiler98/Foreign.hs -23 +27 M ./src/compiler98/Main.hs -14 +47 M ./src/compiler98/Makefile -1 +1 M ./src/compiler98/Need.hs +2 M ./src/compiler98/PPSyntax.hs -1 +1 M ./src/compiler98/Parse.hs -11 +9 M ./src/compiler98/ParseLex.hs -1 +2 M ./src/compiler98/PosCode.hs -1 +2 M ./src/compiler98/SccModule.hs -8 +10 M ./src/compiler98/StrPos.hs -1 +1 M ./src/compiler98/Syntax.hs -5 +8 M ./src/compiler98/TokenId.hs -2 +7 M ./src/compiler98/TokenInt.hs -3 +7 M ./src/compiler98/Type.hs -113 +209 M ./src/old/runtime/io.c -10 +9

1 0

patch applied (hat): Fix bug whereby the Unit type () was mistakenly interpreted as a tuple:
by Malcolm Wallace 10 Oct '06

10 Oct '06

Wed Oct 4 02:17:20 PDT 2000 malcolm * Fix bug whereby the Unit type () was mistakenly interpreted as a tuple: for the primitive FFI, IO () is special. M ./src/compiler98/Foreign.hs -2 +2

1 0

patch applied (hat): First step toward allowing a hierarchical module namespace. We now accept
by Malcolm Wallace 10 Oct '06

10 Oct '06

Tue Oct 3 09:13:16 PDT 2000 malcolm * First step toward allowing a hierarchical module namespace. We now accept an import line of the form import X.Y.Z which is (for now) treated identically to import Z M ./src/compiler98/Parse2.hs -8 +14

1 0

patch applied (hat): Some re-formatting of definitions. No real changes.
by Malcolm Wallace 10 Oct '06

10 Oct '06

Tue Oct 3 09:11:37 PDT 2000 malcolm * Some re-formatting of definitions. No real changes. M ./src/compiler98/RenameLib.hs -150 +175

1 0

patch applied (hat): Modify primitive FFI to permit tuple return types. (No-one should
by Malcolm Wallace 10 Oct '06

10 Oct '06

Tue Oct 3 09:09:40 PDT 2000 malcolm * Modify primitive FFI to permit tuple return types. (No-one should use tuple types directly - they should only be machine generated e.g. by GreenCard.) M ./src/compiler98/Foreign.hs -28 +79 M ./src/compiler98/TokenId.hs +5

1 0

patch applied (hat): Removed #if defined(TRACING) from src/prelude/*/*.hs where it was used before,
by Malcolm Wallace 10 Oct '06

10 Oct '06

Tue Oct 3 04:16:23 PDT 2000 olaf * Removed #if defined(TRACING) from src/prelude/*/*.hs where it was used before, because the tracing transformation couldn't cope with type synonyms. Discovered a bug in the tracing transformation of type synonyms and removed it. The function unwrapNT needs a special expand function, the one in TypeUnify is not suitable because it uses the idempotent closure of the substitution. Also added some type declarations and comments. M ./src/compiler98/DbgTrans.hs -10 +22 M ./src/compiler98/TypeSubst.hs -5 +33 M ./src/compiler98/TypeUnify.hs -1 +25

1 0

patch applied (hat): added some files needed by my recent commits
by Malcolm Wallace 10 Oct '06

10 Oct '06

Mon Oct 2 08:27:02 PDT 2000 sparud * added some files needed by my recent commits A ./src/old/trail/Events.java A ./src/old/trail/ScriptException.java

1 0

patch applied (hat): script cleanup
by Malcolm Wallace 10 Oct '06

10 Oct '06

Mon Oct 2 04:31:48 PDT 2000 sparud * script cleanup M ./src/old/trail/DbgPanel.java -10 +10 M ./src/old/trail/OutputPanel.java -2 +2 M ./src/old/trail/Script.java -216 +22 M ./src/old/trail/ScriptEngine.java -23 +23 M ./src/old/trail/TraceFrame.java -3 +3

1 0