strtof is not defined anymore in std=c++11

[Laurent Alfonsi]

Jon,
>    || (defined (__cplusplus) && __cplusplus >= 201103L)
I agree. That would be perfect.

> Yes.  I think that all of the protoypes I touched: strtof(), strtoll(),
> strtoull() and strtold() need this attention.
Yes. It is the same for these 4 also.

> As currently written, it doesn't test if __STDC_VERSION__ is defined (and so
> uses the assumed value of 0 when compiling C++), so do we actually need to
> test if __cplusplus is defined?
When using gcc, a macro that is not defined is always expanded as 0. see :
     http://gcc.gnu.org/onlinedocs/cpp/If.html#If

So __STDC_VERSION__ >= 199901L is FALSE when __STDC_VERSION__ is not 
defined,
same for __cplusplus >= 201103L. (in the context of gcc preprocessor)

Thanks,
Laurent

On 02.04.2013 21:10, Jon TURNEY wrote:
> On 02/04/2013 13:00, Corinna Vinschen wrote:
>> On Apr  2 12:37, Laurent Alfonsi wrote:
>>> The strtof function is now rejected when selecting the gcc C++11
>>> standard option.
>>> This regression has been introduced in this patch
>>>      http://sourceware.org/ml/newlib/2012/msg00425.html
>>>
>>> J.Turney patch is fine regarding the C standard side, but when using
>>> from C++ :
>>>      $ cat a.cpp
>>>      #include <cstdlib>
>>>
>>>      float  f(const char *s, char **endptr) {
>>>        return strtof(s, endptr);
>>>      }
>>>
>>> It fails with the message :
>>>      $ g++ -std=c++11 a.cpp
>>>      a.cpp: In function 'float f(const char*, char**)':
>>>      a.cpp:3:30: error: 'strtof' was not declared in this scope
>>>
>>> Whereas this function strtof is well included in the cstdlib header
>>> file defined in C++11 ISO.
>>>
>>> Please advice.
>> Not sure.  Enabling c++11 implies defining __STRICT_ANSI__ with gcc.
>> Maybe we have to add something like this to the #if's guarding the
>> declarations:
>>
>>    || (defined (__cplusplus) && __cplusplus >= 201103L)
> Yes.  I think that all of the protoypes I touched: strtof(), strtoll(),
> strtoull() and strtold() need this attention.
>
> As currently written, it doesn't test if __STDC_VERSION__ is defined (and so
> uses the assumed value of 0 when compiling C++), so do we actually need to
> test if __cplusplus is defined?
>
> Would you like me to write a patch?
>