Re: [PATCH] Complete __gnu_debug::basic_string Standard conformity

[Jonathan Wakely <jwakely@redhat.com>]

On Sun, 10 Jul 2022 at 14:57, François Dumont via Libstdc++
<libstdc++@gcc.gnu.org> wrote:
>
> Here is a first patch to complete __gnu_debug::basic_string Standard
> conformity.
>
> I prefer to submit this before checking for more missing stuff to check
> that my proposal of having a testsuite_string.h header is ok.

I think this change means some testcases will never test std::string
with _GLIBCXX_DEBUG defined, because if that is defined they test
__gnu_debug::string instead. That means assertions in
std::basic_string like this one will not get tested:

  template<typename _CharT, typename _Traits, typename _Alloc>
    _GLIBCXX_STRING_CONSTEXPR
    typename basic_string<_CharT, _Traits, _Alloc>::size_type
    basic_string<_CharT, _Traits, _Alloc>::
    find(const _CharT* __s, size_type __pos, size_type __n) const
    _GLIBCXX_NOEXCEPT
    {
      __glibcxx_requires_string_len(__s, __n);

Are we OK with never testing those assertions?

Currently they don't get tested by default because of the extern
template declarations for std::string, so they would only be tested
with -std=c++20 -D_GLIBCXX_DEBUG -D_GLIBCXX_DEBUG_PEDANTIC, which I
don't test routinely. So I suppose this change doesn't make things any
worse in practice.

> I also noticed some problems with _GLIBCXX_DEBUG_PEDANTIC.
>
>      libstdc++: Complete __gnu_debug::string Standard conformity
>
>      Add testsuite/testsuite_string.h header to help testing
> __gnu_debug::basic_string like
>      std::basic_string depending on _GLIBCXX_DEBUG.
>
>      Add using of base type methods in __gnu_debug::basic_string to make
> use of the method
>      overloads when there is no debug version.
>
>      Fix _GLIBCXX_DEBUG_PEDANTIC assertions in <debug/string>. This
> header has to be used directly
>      like __gnu_debug::string, it is not included by _GLIBCXX_DEBUG. It
> means that
>      _GLIBCXX_DEBUG_PEDANTIC is not considered to define
> __glibcxx_check_string and
>      __glibcxx_check_string_len which are then empty macros. Now those
> macros are defined
>      directly in <debug/string> and properly consider
> _GLIBCXX_DEBUG_PEDANTIC.

Nice catch.

OK for trunk, thanks.