Bug in cmath namespace

Bug in cmath namespace

[Jens Vollinga]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset="us-ascii", Size: 803 bytes --]

Hi,

I want to define classes called sin,cos,etc. and include the cmath header at
the same time. I thought this would work since I don't say "using namespace
std;" anywhere. But it seems that by including cmath these functions appear
not only in namespace std, but also in the global namespace. The following
stripped-down program compiles with gcc 3/4, but it shouldn't (according to
the standard, since it doesn't say std::sin):

#include <cmath>

int main()
{
    double x = sin(2.0);
    return 0;
}

I am not really sure whether this is a bug with respect to the ISO standard
or whether I made a mistake here. Does anybody know why gcc/libstc++ behaves
like that?

Jens

-- 
10 GB Mailbox, 100 FreeSMS/Monat http://www.gmx.net/de/go/topmail
+++ GMX - die erste Adresse für Mail, Message, More +++

Re: Bug in cmath namespace

[John Love-Jensen]

Hi Jense,

> I am not really sure whether this is a bug with respect to the ISO standard
> or whether I made a mistake here. Does anybody know why gcc/libstc++ behaves
> like that?

Not a bug.

GCC's g++ includes the C <math.h>, #undef's the macros, does a "using ::sin"
from the global namespace, and adds inline versions of sin(float) and
sin(long double).

HTH,
--Eljay

Re: Bug in cmath namespace

[Jens Vollinga]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset="us-ascii", Size: 792 bytes --]

Hi,

> GCC's g++ includes the C <math.h>, #undef's the macros, does a "using
> ::sin"
> >from the global namespace, and adds inline versions of sin(float) and
> sin(long double).

thanks for the fast reply.
My problem was not to understand what gcc does. gcc works like you described
it above. My problem is whether gcc does it correctly according to the ISO
standard. In paragraph 4 of section 17.4.1.2 in that standard it says that
there should be no ::sin if you include cmath, only a std::sin. But since I
am not well trained in legal speak and have only a '98 draft version of the
standard available I am not sure if a am right or wrong.

Jens

-- 
Lust, ein paar Euro nebenbei zu verdienen? Ohne Kosten, ohne Risiko!
Satte Provisionen für GMX Partner: http://www.gmx.net/de/go/partner

Re: Bug in cmath namespace

[John Love-Jensen]

Hi Jens,

Sorry for the typo on your name.  I'm used to typing "Love-Jensen" and my
fingers didn't stop typing in time.  :-)

I *believe* 17.4.1.2.4 for <cmath> requires the std::sin symbol, but does
not preclude global namespace ::sin.

But... I'm not assuredly sure if ::sin is to be expected, or if it is an
implementation detail.  Or if I'm mistaken entirely and ::sin should not be
present at all.

(My druthers would be that ::sin is not present at all.  C'est la vie.)

Sincerely,
--Eljay

Re: Bug in cmath namespace

[random]

Jens Vollinga wrote:
> Hi,
>
>   
>> GCC's g++ includes the C <math.h>, #undef's the macros, does a "using
>> ::sin"
>> >from the global namespace, and adds inline versions of sin(float) and
>> sin(long double).
>>     
>
> thanks for the fast reply.
> My problem was not to understand what gcc does. gcc works like you described
> it above. My problem is whether gcc does it correctly according to the ISO
> standard. In paragraph 4 of section 17.4.1.2 in that standard it says that
> there should be no ::sin if you include cmath, only a std::sin. But since I
> am not well trained in legal speak and have only a '98 draft version of the
> standard available I am not sure if a am right or wrong.
>
>   
g++ is wrong here, but it is unlikely to be fixed, and also in future
versions of the standard it is likely that it will be allowed to put sin
into the global namespace (as nearly everybody does anyway)

Chris

Re: Bug in cmath namespace

[Jens Vollinga]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset="us-ascii", Size: 440 bytes --]

Hi,

> g++ is wrong here, but it is unlikely to be fixed, and also in future
> versions of the standard it is likely that it will be allowed to put sin
> into the global namespace (as nearly everybody does anyway)

Well, then it seems I have to get used to dirty workarounds ... :-(

Thanks!
Jens

-- 
Lust, ein paar Euro nebenbei zu verdienen? Ohne Kosten, ohne Risiko!
Satte Provisionen für GMX Partner: http://www.gmx.net/de/go/partner

Re: Bug in cmath namespace

[Gabriel Dos Reis]

"Jens Vollinga" <vollinga@gmx.de> writes:

| Hi,
| 
| I want to define classes called sin,cos,etc. and include the cmath header at
| the same time.

Although GCC is wrong in the behaviour, it is one that is hard to fix
and probably won't be fixed.  
furthermore, I believe your program would have violated a C++ standard
rule 17.4.3.1 (entire sub-section).

-- Gaby