Locales

[Paolo Carlini]

Hi,

>I'd appreciate some clarification here and will gladly file a bug if
>appropriate.
>
I think that the best way to clarify the matter is showing what we are
actually doing when a named locale is needed:

  std::locale
  try_named_locale(const char* name)
  {
    try
      {
        return std::locale(name);
      }
    catch (std::runtime_error& ex)
      {
        // Thrown by generic and gnu implemenation if named locale fails.
        if (std::strstr(ex.what(), "name not valid"))
          exit(0);
        else
          throw;
      }
  }

Therefore, the current behavior is definitely not accidental.

That said, I agree that the present situation is not optimal, but I'm also
sure that we do *not* want hundreds of XPASSes!

In my opinion, very-very roughly, testcases involving not available named
locales should not be run at all. This would require a preliminary check
before running the testsuite. But, if we decide to go this way, we should
make sure is obvious from the results *why* some tests are not run...

Benjamin??

Thanks,
Paolo.