Win32Forth defect in (LOCAL)?

While trying to validate my new ANS Forth implemention of StrongForth on
Win32Forth, I stumbled over an issue with (LOCAL). The generated code didn't
work, and I finally found out that even the sample definition of LOCALS|
from appendix A15 of the ANS Forth spec doesn't work correctly:

Win32Forth: a Forth for Windows 95/98/ME/NT/2K/XP
Version: 6.12.00 Build: 2
Compiled: Samstag, August 23 2008, 4:18
Platform: Windows XP, 821k bytes free
2,885 Words in Application dictionary
2,332 Words in System dictionary
5,217 Words total in dictionaries
20,312 Windows Constants available
Current directory: C:\Programme\Win32Forth\extsrc\wincon\winconcpp
Loading Win32For.CFG
Loading Win32ForUser.CFG

Program path is: C:\Programme\Win32Forth\Win32For.EXE
Forth path is: C:\Programme\Win32Forth\
Application path is:

System Editor is: Win32Forth IDE

WARNING OFF SYS-WARNING-OFF ok
: EXAMPLE ( n -- n n**2 n**3 )
LOCALS| N | N DUP N * DUP N * ; ok
5 EXAMPLE . . . 125 25 5 ok
: LOCALS| ( "name...name |" -- )
BEGIN
BL WORD COUNT OVER C@
[CHAR] | - OVER 1 - OR
WHILE
(LOCAL)
REPEAT 2DROP 0 0 (LOCAL)
; ok
IMMEDIATE ok
: EXAMPLE ( n -- n n**2 n**3 )
LOCALS| N | N DUP N * DUP N * ; ok
5 EXAMPLE . . . 0 0 0
^
Error(-4): . stack underflow
4 5 EXAMPLE . . . 64 16 4 ok
..S empty ok

It seems the runtime code of this redefined version of LOCALS| swallows one
cell from the stack. Is this a known defect?

Regards,
Stephan

On Aug 23, 3:52*pm, "Stephan Becher" <stephan.remove-this.bec...@t-
online.de> wrote:
> While trying to validate my new ANS Forth implemention of StrongForth on
> Win32Forth, I stumbled over an issue with (LOCAL). The generated code didn't
> work, and I finally found out that even the sample definition of LOCALS|
> from appendix A15 of the ANS Forth spec doesn't work correctly:
>
> Win32Forth: a Forth for Windows 95/98/ME/NT/2K/XP
> Version: 6.12.00 Build: 2
> Compiled: Samstag, August 23 2008, 4:18
> Platform: Windows XP, 821k bytes free
> *2,885 Words in Application dictionary
> *2,332 Words in System dictionary
> *5,217 Words total in dictionaries
> 20,312 Windows Constants available
> Current directory: C:\Programme\Win32Forth\extsrc\wincon\winconcpp
> Loading Win32For.CFG
> Loading Win32ForUser.CFG
>
> Program path is: * * C:\Programme\Win32Forth\Win32For.EXE
> Forth path is: * * * C:\Programme\Win32Forth\
> Application path is:
>
> System Editor is: * *Win32Forth IDE
>
> WARNING OFF SYS-WARNING-OFF *ok
> : EXAMPLE *( n -- n n**2 n**3 )
> * * LOCALS| N | * N *DUP N * *DUP N * ; *ok
> 5 EXAMPLE . . . 125 25 5 *ok
> : LOCALS| *( "name...name |" -- )
> * * BEGIN
> * * * BL WORD * COUNT OVER C@
> * * * [CHAR] | - OVER 1 - OR
> * * WHILE
> * * * (LOCAL)
> * * REPEAT 2DROP * 0 0 (LOCAL)
> ; *ok
> IMMEDIATE *ok
> : EXAMPLE *( n -- n n**2 n**3 )
> * * LOCALS| N | * N *DUP N * *DUP N * ; *ok
> 5 EXAMPLE . . . 0 0 0
> * * * * * * * ^
> Error(-4): . stack underflow
> 4 5 EXAMPLE . . . 64 16 4 *ok
> .S *empty *ok
>
> It seems the runtime code of this redefined version of LOCALS| swallows one
> cell from the stack. Is this a known defect?
>
> Regards,
> Stephan

Yes, it looks like a bug. However, it's probably not going to get
fixed for some time I suspect, as (LOCALS) isn't used in user code.

: LOCALS| ( -- ) \ ANS standard locals
LOCALS-INIT
FALSE TO LOCDIR \ reversed stack order
BEGIN BLNEXTWORD
2DUP S" |" STR= INVERT
WHILE
(LOCAL) \ declare a local
REPEAT 2DROP
PARMS, \ compile runtime code
(equiv of 0 0 (LOCAL))
; IMMEDIATE

There's a bit of jiggery pokery goes on in the standard definition of
LOCALS| as you can see.

--
Regards
Alex McDonald

> "Alex McDonald" <blog@rivadpm.com> schrieb im Newsbeitrag
> news:0939f2ce-9f21-4b5d-b99b-3b14665e2e9f@56g2000hsm.googlegroups.com...

> On Aug 23, 3:52 pm, "Stephan Becher" <stephan.remove-this.bec...@t-
> online.de> wrote:
> > While trying to validate my new ANS Forth implemention of StrongForth on
> > Win32Forth, I stumbled over an issue with (LOCAL). The generated code
> > didn't
> > work, and I finally found out that even the sample definition of LOCALS|
> > from appendix A15 of the ANS Forth spec doesn't work correctly:
> >
> > Win32Forth: a Forth for Windows 95/98/ME/NT/2K/XP
> > Version: 6.12.00 Build: 2
> > Compiled: Samstag, August 23 2008, 4:18
> > Platform: Windows XP, 821k bytes free
> > 2,885 Words in Application dictionary
> > 2,332 Words in System dictionary
> > 5,217 Words total in dictionaries
> > 20,312 Windows Constants available
> > Current directory: C:\Programme\Win32Forth\extsrc\wincon\winconcpp
> > Loading Win32For.CFG
> > Loading Win32ForUser.CFG
> >
> > Program path is: C:\Programme\Win32Forth\Win32For.EXE
> > Forth path is: C:\Programme\Win32Forth\
> > Application path is:
> >
> > System Editor is: Win32Forth IDE
> >
> > WARNING OFF SYS-WARNING-OFF ok
> > : EXAMPLE ( n -- n n**2 n**3 )
> > LOCALS| N | N DUP N * DUP N * ; ok
> > 5 EXAMPLE . . . 125 25 5 ok
> > : LOCALS| ( "name...name |" -- )
> > BEGIN
> > BL WORD COUNT OVER C@
> > [CHAR] | - OVER 1 - OR
> > WHILE
> > (LOCAL)
> > REPEAT 2DROP 0 0 (LOCAL)
> > ; ok
> > IMMEDIATE ok
> > : EXAMPLE ( n -- n n**2 n**3 )
> > LOCALS| N | N DUP N * DUP N * ; ok
> > 5 EXAMPLE . . . 0 0 0
> > ^
> > Error(-4): . stack underflow
> > 4 5 EXAMPLE . . . 64 16 4 ok
> > .S empty ok
> >
> > It seems the runtime code of this redefined version of LOCALS| swallows
> > one
> > cell from the stack. Is this a known defect?
> >
> > Regards,
> > Stephan
>
> Yes, it looks like a bug. However, it's probably not going to get
> fixed for some time I suspect, as (LOCALS) isn't used in user code.
>

> ...



On the other hand, if someone wishes to implement and use an alternative
word for LOCALS|, he or she has to rely on (LOCAL) to work as specified.
Here's an example:



\ LOCALS( ... ) is an alternative to LOCALS| ... | that accepts

\ the local names in the "correct" order. For example,

\ ... ( n1 n2 n3 -- ) LOCALS( A B C ) ... assigns n1 to A, n2 to

\ B and n3 to C.



: (LOCALS)

PARSE-WORD 2DUP S" )" COMPARE

IF RECURSE (LOCAL) ELSE 2DROP THEN ;



: LOCALS(

(LOCALS) 0 0 (LOCAL) ; IMMEDIATE



The actual problem I have with Win32Forth is that my implementation of
StrongForth in ANS Forth is system code. It uses the host system's version
of (LOCAL) in order to implement StrongForth's type-save version of LOCALS|
that allows arbitrarily mixing single-cell, double-cell and floating-point
locals. So far, I successfully validated StrongForth on GForth and
SwiftForth, but as long as (LOCAL) remains broken, it will definitely not
run on Win32Forth.



Regards,

Stephan

On Sep 1, 9:57*pm, Alex McDonald <b...@rivadpm.com> wrote:
> On Aug 23, 3:52*pm, "Stephan Becher" <stephan.remove-this.bec...@t-
>
>
>
> Yes, it looks like a bug. However, it's probably not going to get
> fixed for some time I suspect, as (LOCALS) isn't used in user code.
>

Actually it's already been fixed in the win32forth CVS.

> : LOCALS| * * * ( -- ) * * * * * * * * * * * * \ ANS standard locals
> * * * * LOCALS-INIT
> * * * * FALSE TO LOCDIR * * * * * * * * * * * *\ reversed stack order
> * * * * BEGIN BLNEXTWORD
> * * * * * * * 2DUP S" |" STR= INVERT
> * * * * WHILE
> * * * * * * * * (LOCAL) * * * * * * * * * * * *\ declare a local
> * * * * REPEAT *2DROP
> * * * * PARMS, * * * * * * * * * * * * * * * * \ compile runtime code
> (equiv of 0 0 (LOCAL))
> * * * * ; IMMEDIATE
>
> There's a bit of jiggery pokery goes on in the standard definition of
> LOCALS| as you can see.
>

it just needed the first two lines moving to (LOCAL) and made
conditional on only being executed for the first local. The
original (LOCAL) is now {LOCAL} since it's used (with different
initialisation) by {.

> --
> Regards
> Alex McDonald

George Hubert

The next definition is LOCALS|...| based :


: reverse-params| ( -- )
0
BEGIN
>IN @ TOKEN S" )" COMPARE
WHILE
SWAP 1+
REPEAT
DROP >IN @ >R
0 ?DO
>IN !
TOKEN PAD APPEND
S" " PAD APPEND
LOOP
S" |" PAD APPEND
R> >IN ! ;


: LOCALS( ( -- )
S" LOCALS| " PAD PLACE reverse-params|
PAD COUNT EVALUATE ; immediate

"Stephan Becher" <stephan.remove-this.becher@t-online.de> a écrit dans
le message de news: g8p874$7n1$00$1@news.t-online.com...
> While trying to validate my new ANS Forth implemention of StrongForth
> on Win32Forth, I stumbled over an issue with (LOCAL). The generated
> code didn't work, and I finally found out that even the sample
> definition of LOCALS| from appendix A15 of the ANS Forth spec doesn't
> work correctly: