A question regarding "Set type"

[RuudB]

I was playing with his set in V3.0.2:
 
const
  s = [47, '0'];

I noticed that TP7 doesn't accept this: all members have to be of the same type. NP for me, I can live with that

The next one:

  s = ['0', 49];

was NOT accepted, with the error that a char was expected. The documentation at https://www.freepascal.org/docs-html/current/ref/refsu16.html#x40-580003.3.3 doesn't mention anything about the possibility of mixing ordinal types.
If it is not possible, the first example should have raised an error, if it is possible, the second one should have been accepted. Unless I have missed something.

Can anybody shed some light on this, please? Many thanks!

FYI, I'm not complaining, just pure curiosity.

[Handoko]

To me, that is a bug. Please report to Bugtracker.
http://bugs.freepascal.org/set_project.php?project_id=6

[Roland57]

IMHO it isn't a problem.

Code: Pascal  [Select][+][-]

1. const
2.   s = [47, '0'];
3.  
4. begin
5.   WriteLn('/' in s); // TRUE
6. end.

But I don't know why this isn't accepted:

Code: Pascal  [Select][+][-]

1. const
2.   s = ['0', 47];

[marcov]

I first thought it was a bug, but strictly speaking since "S" is untyped, and if you take "S" as a array of variant, it is correct syntax

[Handoko]

But why I got no error on the first one but compile-time error on the second one?

[marcov]

But why I got no error on the first one but compile-time error on the second one?

No idea. I thought maybe because char is not a variant type, but it doesn't accept '01' either.

[lucamar]

But why I got no error on the first one but compile-time error on the second one?

Just a guess but it may be that the compiler finds the first member is a char and supposes it is a set of char/string (and waits to see which of the two). Whether that makes sense is anyones guess ...