[Bug tree-optimization/90905] missing -Wreturn-local-addr returning a local std::string::c_str()

[glisse at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90905

--- Comment #7 from Marc Glisse <glisse at gcc dot gnu.org> ---
(In reply to Martin Sebor from comment #6)
> With str being a local (non-reference) variable this should be diagnosed
> because of the str.D.28972._M_local_buf(12):
> 
> # _47 = PHI <_59(9), &str.D.28972._M_local_buf(12), _59(8)>
>   str ={v} {CLOBBER};
>   return _47;

the dump above was obtained from your testcase, compiled with a slightly hacked
libstdc++

+++ basic_string.h      2019-06-18 17:40:06.435533019 +0200
@@ -214,7 +214,9 @@
       _M_set_length(size_type __n)
       {
        _M_length(__n);
+       auto X = _M_data();
        traits_type::assign(_M_data()[__n], _CharT());
+       if(X!=_M_data())__builtin_unreachable();
       }

       bool

+++ basic_string.tcc    2019-06-18 17:38:00.755534757 +0200
@@ -221,8 +221,12 @@
          }

        // Check for out_of_range and length_error exceptions.
+       auto X = _M_data();
        __try
-         { this->_S_copy_chars(_M_data(), __beg, __end); }
+         {
+           this->_S_copy_chars(_M_data(), __beg, __end);
+           if(X!=_M_data())__builtin_unreachable();
+         }
        __catch(...)
          {
            _M_dispose();

and I am really not seeing a warning. But that's probably because the
path-isolation pass is too early (from a warning PoV), the IL doesn't look that
nice at that time yet.