July 2002 - Hugs-Bugs

hierarchical libraries
by Ross Paterson 25 Jul '02

25 Jul '02

I've hacked up an experimental arrangement to make most of the hierarchical libraries distributed with GHC work with Hugs too. The aim is to provide source-level compatibility with both GHC 5.04 and current Hugs. The main omission is modules related to the FFI, as well as various GHC-specific stuff. People accustomed to building Hugs from CVS may wish to try it out. See docs/libraries-notes.txt for more -- you'll need to get libraries and hslibs from CVS too.

1 0

power cables
by sales 25 Jul '02

25 Jul '02

Dear Sir or Madam, It's an honor to know your company through the website. We are a large manufacturer of which specialize in manufacturing and exporting all kinds of power cords, plugs, cordsets, extension cords, sockets, and so on. Now we are planning to incorporate our business activities in your market, and we will be much obliged for your introduction to a most reliable importer handling this field. We've exported a large quantity of power cords to many countries, and enjoying a high reputation through excellent quality and price. We believe you have a great demand of power cords, please don't hesitate to contact us if you are interested in. Please advise us the following: 1) Cable specification: VDE approved H05VV-F, H03VV-F, H03VVH2-F, H05RN-F,…UL approved SJT, SJTW, HPN 18AWG, 16AWG, with USA plug. 2) Length, Color, requirement of the end. 3) Which types about plug? 4) Quantity per order? Please send us the drawings for these items so that we can quote you the best offer immediately. Please transfer this letter to Purchasing Department of your company when you received it. Your kind attention to this matter will be much appreciated. With thanks and best regards, Shanghai Foundoor Electrical Co., Ltd. Steven Xu Contact information: Shanghai Foundoor Electrical Co., Ltd. Office Address: RM 204, No. 54, Lane 250, Zhenjin Road, Shanghai, China 200333 Tel: 86-21-56348070 Fax: 86-21-56348069 E-mail: sales(a)foundoor.com http://www.foundoor.com

1 0

transitive imports
by Ross Paterson 24 Jul '02

24 Jul '02

Given these modules: module A(Stuff(Foo, Bar)) where data Stuff = Foo | Bar module B(Stuff(..)) where import A module C where import B loading C gives "INTERNAL ERROR: importEntity" Changing the import in C gives different internal errors. The particular form of the exports in A and B and the import in B seems necessary to trigger the bug.

2 2

Hugs on Linux/PPC
by Michele Campeotto 23 Jul '02

23 Jul '02

Hi, I'm trying to run Hugs on Linux/PPC, it seems to build fine but it just segfaults on startup. Is it a known problem or would you like me to send a stack trace? Cheers -- Michele - www.micampe.it "Smile, it makes people wonder what you are thinking"

1 1

expandSynFully in subst.c
by Ross Paterson 23 Jul '02

23 Jul '02

expandSynFully() is declared and defined inside #ifdef TREX but used outside of it (in getProgType).

1 0

standard input not a tty
by Ross Paterson 18 Jul '02

18 Jul '02

When hugs's standard input is not a tty, this patch stops it prompting or attempting to use readline. It also disables $$ in that situation; this could be changed if desired, but is probably the right thing. If it is put it in, the --with-readline option of configure can probably be ditched (i.e. set USE_READLINE if editline or readline is available). Index: src/input.c =================================================================== RCS file: /home/cvs/root/hugs98/src/input.c,v retrieving revision 1.43 diff -u -r1.43 input.c --- src/input.c 2002/06/14 14:41:10 1.43 +++ src/input.c 2002/07/18 15:58:05 @@ -264,6 +264,7 @@ #define SCRIPTFILE 2 /* - script file */ #define PROJFILE 3 /* - project file */ #define STRING 4 /* - string buffer? */ +#define NOKEYBOARD 5 /* - standard input, but not a tty */ static Int reading = NOTHING; @@ -322,6 +323,12 @@ hereState = START; #endif +#ifdef HAVE_ISATTY + if (!isatty(fileno(stdin))) { /* not reading from a tty: */ + reading = NOKEYBOARD; /* don't prompt or try readline */ + return; + } +#endif #if USE_READLINE /* Paranoid freeing code supplied by Sverker Nilsson (sverker(a)opq.se) * avoids accidentally freeing currentLine twice. @@ -607,6 +614,9 @@ c1 = ' '; } } + } + else if (reading==NOKEYBOARD) { + c1 = c0=='\n' ? EOF : getc(stdin); } else if (reading==STRING) { c1 = (unsigned char) *nextStringChar++;

2 1

Garbage collection does not fully recover from error
by Raymond Boute 18 Jul '02

18 Jul '02

L.S., Consider the module definition: // module Infinite where x = 1 : map (+1) x y = 1 : map (+1) y \\ and the following session // Hugs session for: C:\hugs98\lib\Prelude.hs D:\TopFolder\Haskexper\Misc\Infinite.hs Type :? for help Infinite> take 5 x [1,2,3,4,5] Infinite> take 5 x == take 5 y True Infinite> x == y ERROR - Garbage collection fails to reclaim sufficient space Infinite> take 5 x == take 5 y True ERROR - Garbage collection fails to reclaim sufficient space Infinite> \\ The first evaluation of take 5 x == take 5 y causes no problems, but after the stack overflow caused by evaluating x == y the second evaluation of take 5 x == take 5 y returns the correct answer but also an error message. Please let me know (i) whether this should be expected, (ii) if so, why, (iii) if not, how this can be corrected. Regards, Raymond Boute

2 1

semicolons
by Ross Paterson 16 Jul '02

16 Jul '02

Haskell 98 allows things separated by semicolons to be empty, unlike earlier versions. Hugs implements this for decls and ldecls, but not other contexts, e.g. it rejects stuff like module Foo where {import A;;x = 1;;y = 2;} do {;;;x <- [1];;;return x;;} The attached patch fixes this for impDecls, topDecls and stmts. That leaves alts, I think. ------------------------------------------------------------------- Index: src/parser.y =================================================================== RCS file: /cvs/hugs98/src/parser.y,v retrieving revision 1.33 diff -u -r1.33 parser.y --- src/parser.y 2002/06/14 14:41:10 1.33 +++ src/parser.y 2002/07/15 11:21:15 @@ -163,7 +163,8 @@ } } ; -modBody : topDecls {$$ = $1;} +modBody : /* empty */ {$$ = gc0(NIL);} + | topDecls {$$ = gc1($1);} | impDecls chase {$$ = gc2(NIL);} | impDecls ';' chase topDecls {$$ = gc4($4);} ; @@ -203,6 +204,7 @@ /*- Import declarations: --------------------------------------------------*/ impDecls : impDecls ';' impDecl {imps = cons($3,imps); $$=gc3(NIL);} + | impDecls ';' {$$=gc2(NIL);} | impDecl {imps = singleton($1); $$=gc1(NIL);} ; chase : /* empty */ {if (chase(imps)) { @@ -260,13 +262,9 @@ /*- Top-level declarations: -----------------------------------------------*/ -topDecls : /* empty */ {$$ = gc0(NIL);} - | ';' {$$ = gc1(NIL);} - | topDecls1 {$$ = $1;} - | topDecls1 ';' {$$ = gc2($1);} - ; -topDecls1 : topDecls1 ';' topDecl {$$ = gc2($1);} - | topDecls1 ';' decl {$$ = gc3(cons($3,$1));} +topDecls : topDecls ';' {$$ = gc2($1);} + | topDecls ';' topDecl {$$ = gc2($1);} + | topDecls ';' decl {$$ = gc3(cons($3,$1));} | topDecl {$$ = gc0(NIL);} | decl {$$ = gc1(cons($1,NIL));} ; @@ -879,11 +877,9 @@ | guardAlt {$$ = gc1(cons($1,NIL));} ; guardAlt : '|' exp0 ARROW exp {$$ = gc4(pair($3,pair($2,$4)));} - ; -stmts : stmts1 ';' {$$ = gc2($1);} - | stmts1 {$$ = $1;} ; -stmts1 : stmts1 ';' stmt {$$ = gc3(cons($3,$1));} +stmts : stmts ';' stmt {$$ = gc3(cons($3,$1));} + | stmts ';' {$$ = gc2($1);} | stmt {$$ = gc1(cons($1,NIL));} ; stmt : exp_err FROM exp {$$ = gc3(ap(FROMQUAL,pair($1,$3)));}

2 1

Hugs doesn't follow latest layout specification
by Mark Tullsen 15 Jul '02

15 Jul '02

In the following code snippet: f (x:xs) = "a" where f [] = "b" We have f [] => "b" -- in ghc and nhc98 and /latest/ Report f [] => Program error: {f []} -- in hugs It appears that Hugs has not been updated to follow the most current Haskell98 Report which explicates the layout rules in more detail. (Which specifies that the above 'where' construct should bind nothing rather than binding the "f [] = ...".) - Mark

2 1

Darling
by iSrerOECmtblt 13 Jul '02

13 Jul '02

1 0