Broken char through Format() with ThousandSeparator = NBSP

[anse]

I have some users with ru-RU locale, which has ThousandSeparator = #160 (no-break-space, looks like a space but allows no line break).
Formatting numbers in their locale shows a broken ThousandSeparator, see attached images.
First one is done on a ru-RU Linux system, second one on my own de-DE Windows where I testwise set that nbsp char:

Code: Pascal  [Select][+][-]

1. DefaultFormatSettings.ThousandSeparator:= chr(160);

Both are generated through:

Code: Pascal  [Select][+][-]

1. Format('%.0n', [aValue + 0.0])

Confusing to me is that nbsp is even in the ASCII and LATIN1 charset.

Is it known that Format cannot handle certain diacritical characters?

There's a 9 year old post saying this is not supported by FPC: https://forum.lazarus.freepascal.org/index.php/topic,32159.msg207284.html#msg207284

[paweld]

Lazarus treats strings as UTF-8 so you need to convert formatted value to UTF8:

Code: Pascal  [Select][+][-]

1. uses
2.   LazUTF8;
3.  
4. var
5.   fs: TFormatSettings;
6. begin
7.   fs.ThousandSeparator := #160;
8.   Caption := WinCPToUTF8(Format('%.0n', [123456.0], fs)); //or CP1251ToUTF8 from LConvEncoding unit
9. end;

[anse]

Thanks a lot @paweld. LazUTF8.WinCPToUTF8 showed up another (broken) character instead of the question mark. But it looks like LConvEncoding.CP1251ToUTF8 is doing the right thing - at least on Windows, but I guess the Linux version does the same.

[paweld]

LazUTF8.WinCPToUTF8 showed up another (broken) character instead of the question mark.

probably because the Windows code page is not Latin1