[PATCH] PR libstdc++/80624 satisfy invariant for char_traits<char16_t>::eof()

[Jonathan Wakely]

On 08/05/17 13:37 +0200, Stephan Bergmann via libstdc++ wrote:
>On 05/08/2017 12:19 PM, Jonathan Wakely via libstdc++ wrote:
>>Changing the semantics of to_int_type for U+FFFF is far less likely to
>>affect any ABIs (it's a constexpr function so it's possible somebody
>>is using the value of to_int_type(char_type(-1)) as a template
>>argument, but it seems unlikely. It's a much smaller change, "allowed"
>>by http://www.unicode.org/faq/private_use.html#nonchar10 and it only
>>affects a noncharacter that is not intended for interchange anyway.
>
>I'm not convinced that a C++ implementation should be allowed to 
>rewrite U+FFFF -> U+FFFD in this manner.  Noncharacters not being 
>suitable for interchange IMO doesn't imply that an application 
>shouldn't be able to write them out and read them back in.  (And 
><http://www.unicode.org/versions/corrigendum9.html> "Corrigendum #9: 
>Clarification About Noncharacters" seems to support that view.)
>
>To me, it feels like this change is not an improvement over keeping 
>living with the current defect.

So rather than transforming U+FFFF into U+FFFD you'd prefer to write
nothing instead?

#include <sstream>
#include <cassert>

int main()
{
  std::basic_ostringstream<char16_t> s;
  s.put(u'\uFFFF');
  assert( s.str().length() == 1 );
}

a.out: ex.cc:8: int main(): Assertion `s.str().length() == 1' failed.
Aborted (core dumped)

With my patch it inserts U+FFFD.