Unexpected treatment of tags in variants
lazarus / 62-General

----------------------------------------------------------------------
440bx · 2022-02-27 18:11

Hello,

consider the following type declaration Code: Pascal
 [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i type 
SomeRecord          : record    case integer of      1 : (           a
             : integer;          );      1 : (           b           
 : integer;          );      1 : (           c              :
integer;          );      1 : (           d              : integer;  
      );  end; FPC doesn't complain that multiple variants are tagged
with the same integer (strictly speaking the compiler doesn't need the
tags to be different) but, Delphi (I've only tested with version 2)
doesn't accept it (which seems to be the reasonable/strict action.)

Thoughts and comments welcome.

----------------------------------------------------------------------
howardpc · 2022-02-27 21:34

Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i SomeRecord
= recordcase Integer of   1: (a: Integer);   1: (b: Integer);   1: (c:
Integer);   1: (d: Integer);end;is exactly equivalent to
Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i ThisRecord
= record   z: Integer;end;
The anonymous integer tags of 1, and the four apparently different
fields a, b, c, d are all tautological.

So although Delphi appears stricter, I would say the FPC is not
incorrect to ignore the value of the case selector which is
meaningless in this case, since although the four fields are given
different names they all refer to the same memory location in the
record.

----------------------------------------------------------------------
Bart · 2022-02-27 21:45

Quote from: howardpc on February 27, 2022, 10:34:15 pmThe anonymous
integer tags of 1, and the four apparently different fields a, b, c, d
are all tautological.

This will also compile though:
Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i type 
SomeRecord = record  case Integer of    1: (a: Integer);    1: (b:
char);    1: (c: double);    1: (d: boolean);  end;
Bart

----------------------------------------------------------------------
howardpc · 2022-02-27 21:56

@Bart

True enough. However in this case the Pascal syntax introduces
redundancy, since no case selectors are needed, but the syntax
requires them.
The compiler constructs the record according to the maximum field size
required, and the programmer's field naming is sufficient to identify
which type to select at runtime. The case selector can be ignored
altogether.

Nevertheless I certainly admit that it does look less than ideal that
the compiler swallows this without complaint, even though no harm
comes of it.

----------------------------------------------------------------------
440bx · 2022-02-27 23:29

Just for the record, I used integer everywhere for copy/paste
simplicity.

I noticed the problem with a much more complicated structure that
involved many types.

----------------------------------------------------------------------
PascalDragon · 2022-02-28 13:49

Quote from: 440bx on February 27, 2022, 07:11:34 pmThoughts and
comments welcome.

Please file a bug report.

----------------------------------------------------------------------
440bx · 2022-02-28 15:49

Quote from: PascalDragon on February 28, 2022, 02:49:30 pmQuote from:
440bx on February 27, 2022, 07:11:34 pmThoughts and comments welcome.

Please file a bug report.

done.  https://gitlab.com/freepascal.org/fpc/source/-/issues/39605

----------------------------------------------------------------------
PascalDragon · 2022-03-01 08:10

Quote from: 440bx on February 28, 2022, 04:49:11 pmQuote from:
PascalDragon on February 28, 2022, 02:49:30 pmQuote from: 440bx on
February 27, 2022, 07:11:34 pmThoughts and comments welcome.

Please file a bug report.

done.  https://gitlab.com/freepascal.org/fpc/source/-/issues/39605

Thank you.

----------------------------------------------------------------------
Bart · 2022-03-02 13:36

Just for fun: TP 6.0 actually accepts this.

Bart