_GLIBCXX_ASSERT lossage

[Pétur Runólfsson]

Hi,

Recently a few testcases (for example
testsuite/22_locale/collate/compare/wchar_t/2.cc) were modified so they
contain the line

// { dg-options "-finput-charset=iso-8859-1" }

However, some of these testcases also use the VERIFY macro. The last
time I checked, VERIFY doesn't really do anything unless _GLIBCXX_ASSERT
is defined and _GLIBCXX_ASSERT is defined is defined by default for all
tests - except for those that happen to use dg-options. I don't think
the intent of the change was to disable VERIFY for these tests, at least
there is no mention of VERIFY or _GLIBCXX_ASSERT in the ChangeLog:

2004-01-29  Eric Christopher  <echristo@redhat.com>
            Zack Weinberg  <zack@codesourcery.com>

        * testsuite/22_locale/collate/compare/wchar_t/2.cc: Remove xfail. Use
        -finput-charset.
        * testsuite/22_locale/collate/compare/wchar_t/wrapped_env.cc: Ditto.
        * testsuite/22_locale/collate/compare/wchar_t/wrapped_locale.cc: Ditto
        * testsuite/22_locale/collate/hash/wchar_t/2.cc: Ditto.
        * testsuite/22_locale/collate/hash/wchar_t/wrapped_env.cc: Ditto.
        * testsuite/22_locale/collate/hash/wchar_t/wrapped_locale.cc: Ditto.
        * testsuite/22_locale/collate/transform/wchar_t/2.cc: Ditto.
        * testsuite/22_locale/collate/transform/wchar_t/wrapped_env.cc: Ditto.
        * testsuite/22_locale/collate/transform/wchar_t/wrapped_locale.cc:
        Ditto.

Regards,
Petur

P.S. Why is the definition of VERIFY so funky?