[Bug libstdc++/110653] Support std::stoi etc. without C99 APIs

["redi at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110653

--- Comment #12 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to dave.anglin from comment #11)
> Yes, this works.

Great, thanks.

> hppa64-hpux does not have have strtof.  Could std::stof be implemented using
> strtod in this case?

Maybe something like this:

--- a/libstdc++-v3/include/bits/basic_string.h
+++ b/libstdc++-v3/include/bits/basic_string.h
@@ -4153,6 +4153,25 @@ _GLIBCXX_BEGIN_NAMESPACE_CXX11
   inline float
   stof(const string& __str, size_t* __idx = 0)
   { return __gnu_cxx::__stoa(&std::strtof, "stof", __str.c_str(), __idx); }
+#else
+  inline float
+  stof(const string& __str, size_t* __idx = 0)
+  {
+    double __d = std::stod(__str, __idx);
+    if (__builtin_isfinite(__d))
+      {
+       double __abs_d = __builtin_fabs(__d);
+       if (__abs_d < __FLT_MIN__)
+         errno = ERANGE;
+       else if (__abs_d > __FLT_MAX__)
+         {
+           errno = ERANGE;
+           float __f = __builtin_huge_valf();
+           return __d > 0.0 ? __f : -__f;
+         }
+      }
+    return __d;
+  }
 #endif

 #if _GLIBCXX_USE_C99_STDLIB || _GLIBCXX_HAVE_STRTOLD


> I'm thinking a test to check the presence and maybe compliance of these
> routines might be good.

We have some:
testsuite/21_strings/basic_string/numeric_conversions/char/stof.cc
testsuite/21_strings/basic_string/numeric_conversions/char/stold.cc

But they depend on { dg-require-string-conversions "" } which is only true if
we have the full set of them:

    return [check_v3_target_prop_cached et_string_conversions {
        set cond "_GLIBCXX_USE_C99_STDIO && _GLIBCXX_USE_C99_STDLIB"
        set cond "$cond && _GLIBCXX_USE_C99_WCHAR"
        set cond "$cond && !defined _GLIBCXX_HAVE_BROKEN_VSWPRINTF"
        return [v3_check_preprocessor_condition string_conversions $cond]
    }]