Message-ID:	<3BAA1F02.7AB1E05D@yahoo.com.NOSPAM>
From:	CBFalconer <cbfalconer AT yahoo DOT com DOT NOSPAM>
Organization:	Ched Research
X-Mailer:	Mozilla 4.75 [en] (Win98; U)
X-Accept-Language:	en
MIME-Version:	1.0
Newsgroups:	comp.os.msdos.djgpp,comp.lang.pascal.misc,comp.lang.pascal.ansi-iso
Subject:	Testing GPC under DJGPP
Lines:	126
Date:	Thu, 20 Sep 2001 16:53:23 GMT
NNTP-Posting-Host:	12.90.167.21
X-Complaints-To:	abuse AT worldnet DOT att DOT net
X-Trace:	bgtnsc05-news.ops.worldnet.att.net 1001004803 12.90.167.21 (Thu, 20 Sep 2001 16:53:23 GMT)
NNTP-Posting-Date:	Thu, 20 Sep 2001 16:53:23 GMT
To:	djgpp AT delorie DOT com
DJ-Gateway:	from newsgroup comp.os.msdos.djgpp
Reply-To:	djgpp AT delorie DOT com

I saw some things in the GPC docs about not overlaying memory
areas in variant records, which worried me inasmuch as many
programs depend on this for interfacing.

So I wrote the following test, which shows that memory *IS*
overlaid.  However it also shows some other failings in GPC (I set
the option to standard-pascal).  It is being run from RHIDE under
DJGPP, if that matters.

I would appreciate hearing from the developers on this.  BTW,
according to standards a component of a packed array cannot be
accessed directly, but must be extracted or inserted with the pack
and unpack standard procedures.  My PascalP also has this failing.

PROGRAM gnuvarnt(input, output);
(* Check whether or not variant records access same memory *)

  TYPE
    varcd  = RECORD
      CASE 1..4 OF
1:      (c  : char);
2:      (s  : PACKED ARRAY[1..8] OF char);
3:      (a3 : ARRAY[1..3] OF integer);
4:      (a4 : ARRAY[1..4] OF integer);
      END;

  VAR
    vrcd   : varcd;
    i      : integer;

  PROCEDURE showrcd(kind : integer);

    VAR
      i   : integer;

    BEGIN (* showrcd *)
    write('Record type ', kind, ': ');
    CASE kind OF
1:    write(vrcd.c);
2:    write(vrcd.s);
3:    FOR i := 1 TO kind DO write(vrcd.a3[i] : 12);
4:    FOR i := 1 TO kind DO write(vrcd.a4[i] : 12);
    OTHERWISE
      write('Bad record kind ', kind);
      END;
    writeln;
    END; (* showrcd *)

  BEGIN (* gnuvarnt *)
  writeln; writeln('Testing variant records');
  FOR i := 1 TO 4 DO vrcd.a4[i] := i;
  showrcd(4);
  writeln;
  
  FOR i := 1 TO 3 DO vrcd.a3[i] := -i;
  showrcd(3);
  showrcd(4);
  writeln;

  vrcd.s := 'Eight ch';
  showrcd(2);
  showrcd(3);
  showrcd(4);
  writeln;

  vrcd.c := 'X';
  showrcd(1);
  showrcd(2);
  showrcd(3);
  showrcd(4);
  writeln;

  writeln('This should be a compile time error');
  vrcd.s[2] := 'Y';
  showrcd(2);
  writeln;

  writeln('This should fail on index range');
  FOR i := 1 TO 4 DO vrcd.a3[i] := i + 10;
  showrcd(4);
  writeln('DONE');

  END. (* gnuvarnt *)

Output, here, is:

Testing variant records
Record type 4:            1           2           3           4

Record type 3:           -1          -2          -3
Record type 4:           -1          -2          -3           4

Record type 2: Eight ch
Record type 3:   1751607621  1751326836          -3
Record type 4:   1751607621  1751326836          -3           4

Record type 1: X
Record type 2: Xight ch
Record type 3:   1751607640  1751326836          -3
Record type 4:   1751607640  1751326836          -3           4

This should be a compile time error
Record type 2: XYght ch

This should fail on index range
Record type 4:           11          12          13          14
DONE

And the compilation results under PascalP are:

[1] c:\p\gnuvarnt>pascalp gnuvarnt.pas gnuvarnt.lst nul
PASCALP (pasctext, pasclist, prr, ef, output) [parm] V 3.1.9T
    6000     0:d              CASE 1..4 OF
                 ****               ^2
  2. Identifier expected
   29000   189: 3           OTHERWISE
                 ****              ^118
118. ** WARNING ** Non-standard Pascal feature used

NO. ERRORS=1 WARNINGS=1 Program size(pcode bytes)=744

-- 
Chuck F (cbfalconer AT yahoo DOT com) (cbfalconer AT XXXXworldnet DOT att DOT net)
   (Remove "XXXX" from reply address. yahoo works unmodified)
   mailto:uce AT ftc DOT gov  (for spambots to harvest)

- Raw text -