for in shortstring

[bytebites]

Code: Pascal  [Select][+][-]

1. program short;
2. uses sysutils;
3. var s:string[20];
4.   x: Char;
5. begin
6.   s:=inttostr(12345);
7.   for x in s do write(ord(x):4);
8.   writeln;
9. end.
10.  

This code produces with 3.2
   49  50  51  52  53
but trunk version
   5  49  50  51  52  53   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0

[Bart]

Confirmed.
I would say: report that as a bug.
(Unless Delphi behaves the same, which I cannot test, but even then....)

Bart

[bytebites]

I added comment to the failing commit https://gitlab.com/freepascal.org/fpc/source/-/commit/68668c649e5aed88bac022a11a4e78bb3adb3a41

[Bart]

Not sure wether that gets noticed...
I dropped him a note in the original bugreport, hopefully he gets that.

Bart

[wp]

Delphi 11.3 CE, BTW, does it correctly:

Code: Pascal  [Select][+][-]

1. program Project1;
2. var
3.   s: String[10];
4.   ch: AnsiChar;
5. begin
6.   s := 'abc';
7.   for ch in s do
8.     Write(ord(ch), ' ');
9.   WriteLn;
10.   ReadLn;
11. end.

Output:

Code: [Select]

97 98 99

[Bart]

It's because "for .. in" for strings now uses Low and High instead of 1 and Length for it's boudaries.
It was introduced because "for .. in" for zero based strings (yes, we have a compiler directive for that) was off by one.
Obviousely it's a regression.

Bart

[AlexTP]

Posted to https://gitlab.com/freepascal.org/fpc/source/-/issues/40563