• Arcane Jill (8/11) Jun 05 2004 However, we see a related bug in the following example:
• Walter (7/17) Jun 05 2004 The

Arcane Jill <Arcane_member pathlink.com> writes:

The following is correct behavior, and is implemented correctly. Nice one! The
compiler correctly correctly rejects the following line.

       char c = 'ß';  // compile error - invalid UTF-8 sequence


However, we see a related bug in the following example:

       char c = 0xC3; // first byte of a UTF-8 sequence
       wchar w = c;

This auto-promotion should fail, throwing a runtime exception (because 0xC3 by
itself is an invalid UTF-8 sequence). Current behavior is that the cast
succeeds, as though c had contained an ISO-8859-1 character instead of a UTF-8
fragment.

Arcane Jill

"Walter" <newshound digitalmars.com> writes:

"Arcane Jill" <Arcane_member pathlink.com> wrote in message
news:c9s7nu$1255$1 digitaldaemon.com...
 The following is correct behavior, and is implemented correctly. Nice one!

The
 compiler correctly correctly rejects the following line.

       char c = 'ß';  // compile error - invalid UTF-8 sequence


 However, we see a related bug in the following example:

       char c = 0xC3; // first byte of a UTF-8 sequence
       wchar w = c;

 This auto-promotion should fail, throwing a runtime exception (because

0xC3 by
 itself is an invalid UTF-8 sequence). Current behavior is that the cast
 succeeds, as though c had contained an ISO-8859-1 character instead of a

UTF-8
 fragment.

I see what you're saying. Doing such would require a runtime test; not sure
about the tradeoffs.