C++ header files

[Clint Whaley]

Guys,

I got the message below as a support request.  I am not a C++ guy, and I have
never actually read the standard, so I'd like to get some feedback from people
who use it regularly.  This user is requesting that I add:

#ifdef __cplusplus
extern "C"
{
#endif
....

#ifdef __cplusplus
}
#endif

To the main C API header files for C++ users.  ATLAS's present approach is:
   http://math-atlas.sourceforge.net/errata.html#CPPheaders

For the user suggestion, I have several questions:
(1) is the '__cplusplus' macro part of the C++ standard, or a gnu-specific
    extension?
(2) Does the standard say what happens if you have nested extern "C"
    declarations?
(3) What are people's feelings on this proposal?

Thanks,
Clint

>From noreply@...  Sun Jun 22 22:59:13 2008
Return-Path: <noreply@...>
X-Original-To: whaley@...
Delivered-To: whaley@...
To: noreply@...
From: "SourceForge.net" <noreply@...>
Subject: [ math-atlas-Support Requests-2000447 ] Comment on FAQ "How about C++ header files for the C ..."
Content-Type: text/plain; charset="UTF-8"
X-SourceForge-Tracker-unixname: math-atlas
X-SourceForge-Tracker-trackerid: 379483
X-SourceForge-Tracker-itemid: 2000447
X-SourceForge-Tracker-itemstatus: Open
X-SourceForge-Tracker-itemassignee: nobody
X-SourceForge-Tracker-itemupdate-reason: Tracker Item Submitted
X-SourceForge-Tracker-itemupdate-username: Item Submitter
Date: Mon, 23 Jun 2008 03:59:01 +0000
X-Spam-Checker-Version: SpamAssassin 3.0.2 (2004-11-16) on mail0.cs.utsa.edu
X-Spam-Level:
X-Spam-Status: No, score=-2.3 required=5.0 tests=AWL,BAYES_00 autolearn=ham
        version=3.0.2
Status: RO

Support Requests item #2000447, was opened at 2008-06-23 03:59
Message generated for change (Tracker Item Submitted) made by Item Submitter
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=379483&aid=2000447&group_id=23725

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: Problems during install
Group: None
Status: Open
Resolution: None
Priority: 5
Private: No
Submitted By: Nobody/Anonymous (nobody)
Assigned to: Nobody/Anonymous (nobody)
Summary: Comment on FAQ "How about C++ header files for the C ..."

Initial Comment:
Hello,

I installed ATLAS to do some comparisons with
Intel MKL, AMD ACML etc.  The test code is C++ and I am
testing on FreeBSD 7.0.  I ran into the problem of
linking which is covered in the errata entry
"How about C++ header files for the C interfaces?"
I found this after I had worked out the problem
(and yes I did read the docs - just missed this one
as explained below).

A couple of comments:
 1. The error occurs at link time but the errata
    entry does not mention the link symptoms.
    If you are an expert C++ programmer then maybe it
    is obvious but for others I think that a note about
    the symptoms would be useful (esp when searching
    for symptoms on the web).

 2. At the bottom of the entry it says
    "If you are a C++ programmer using ATLAS, and think
     differently, let me know."
    Well I am not a C++ programmer but I cannot see why
    the 'extern (C)' cannot be added to the cblas.h
    header file with appropriate ifdefs e.g.

#ifdef __cplusplus
extern "C"
{
#endif
...
<the rest of cblas.h file>
...
#ifdef  __cplusplus
}
#endif

   This way it is no-op for C programs and will work
   without C++ programmers having to add noise to their
   code.  It is also used in standard system files
   'grep 'extern "C"' /usr/include/*.h'

 3. As for the "Since ATLAS does not provide full OO
    C++ interfaces, I am reluctant to raise the
    expectation that it does by providing C++ specific
    header files."
    If the above ifdefs were added I do not believe
    it will raise expectations.  Since it is an
    explicit 'extern "C"' you know they are C only
    interfaces.
 4. Of course if you do add the 'extern "C"' as
    to the cblas.h file as suggested, existing code
    would have additional enclosing 'extern "C"'.
    I did a quick test and this does not appear to
    cause any problems.

I have not used ATLAS enough to comment fully on its
usefulness but it does appear to be extremely useful
well written software.
(I did read the ATLAS FAQ ;-)

thanks
--
tony maher    tonymaher@...

----------------------------------------------------------------------

You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=379483&aid=2000447&group_id=23725


**************************************************************************
** R. Clint Whaley, PhD ** Assist Prof, UTSA ** www.cs.utsa.edu/~whaley **
**************************************************************************

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Math-atlas-devel mailing list
Math-atlas-devel@...
https://lists.sourceforge.net/lists/listinfo/math-atlas-devel

[smontanaro]

    Clint> I got the message below as a support request.  I am not a C++
    Clint> guy, and I have never actually read the standard, so I'd like to
    Clint> get some feedback from people who use it regularly.  This user is
    Clint> requesting that I add:

    Clint> #ifdef __cplusplus
    Clint> extern "C"
    Clint> {
    Clint> #endif
    Clint> ....

    Clint> #ifdef __cplusplus
    Clint> }
    Clint> #endif

That is the standard idiom.  It will instruct the C++ compiler to not mangle
the names between the braces (C linkage).  C compilers will not see it.

--
Skip Montanaro - skip@... - http://www.webfast.com/~skip/
"Be different, express yourself like everyone else."

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Math-atlas-devel mailing list
Math-atlas-devel@...
https://lists.sourceforge.net/lists/listinfo/math-atlas-devel

[Andrew Reilly-9]

On Tue, Jun 24, 2008 at 05:50:10PM -0500, skip@... wrote:
>     Clint> #ifdef __cplusplus
>     Clint> extern "C"
>     Clint> {
>     Clint> #endif
>     Clint> ....

> That is the standard idiom.  It will instruct the C++ compiler to not mangle
> the names between the braces (C linkage).  C compilers will not see it.

This is probably the wrong forum, but I've never understood why
that was considered to be the appropriate strategy.  Why should
C code have a bunch of spurious C++-related fluff added?  Surely
the correct approach is for C++ code to wrap C includes in
extern "C", as the syntax would seem to indicate?

Every other language that has some way of linking to C code (and
most do) has to do its own compatability shimming, with it's own
foreign-function interface.

My guess is that the process started when it was assumed that
C++ was the future of C, and eventually everything would just be
C++.  That is much less clearly the case, now, IMO.

Cheers,

--
Andrew

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Math-atlas-devel mailing list
Math-atlas-devel@...
https://lists.sourceforge.net/lists/listinfo/math-atlas-devel

[smontanaro]

    Andrew> This is probably the wrong forum, but I've never understood why
    Andrew> that was considered to be the appropriate strategy.  Why should
    Andrew> C code have a bunch of spurious C++-related fluff added?  Surely
    Andrew> the correct approach is for C++ code to wrap C includes in
    Andrew> extern "C", as the syntax would seem to indicate?

I think it's mostly because C++ is used to compile C code.  I know the
Python source code uses this technique.  I believe you can compile it with
g++.

    Andrew> My guess is that the process started when it was assumed that
    Andrew> C++ was the future of C, and eventually everything would just be
    Andrew> C++.  That is much less clearly the case, now, IMO.

Thankfully. :-)

--
Skip Montanaro - skip@... - http://www.webfast.com/~skip/
"Be different, express yourself like everyone else."

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Math-atlas-devel mailing list
Math-atlas-devel@...
https://lists.sourceforge.net/lists/listinfo/math-atlas-devel

[David Chin]

The __cplusplus macro is part of the ISO C++ standard, defined
precisely because the standards committee knew (presumably) that there
would be people who would want to use their old plain C stuff.

Similarly, ISO C has the macro __STDC__ defined.

See here:  http://docs.sun.com/app/docs/doc/819-5267/bkaep?a=view

So, it is more than just idiom. It's required of a standards-compliant compiler.

Cheers,
    Dave

On Tue, Jun 24, 2008 at 7:13 PM, Andrew Reilly
<andrew-atlas@...> wrote:

> On Tue, Jun 24, 2008 at 05:50:10PM -0500, skip@... wrote:
>>     Clint> #ifdef __cplusplus
>>     Clint> extern "C"
>>     Clint> {
>>     Clint> #endif
>>     Clint> ....
>
>> That is the standard idiom.  It will instruct the C++ compiler to not mangle
>> the names between the braces (C linkage).  C compilers will not see it.
>
> This is probably the wrong forum, but I've never understood why
> that was considered to be the appropriate strategy.  Why should
> C code have a bunch of spurious C++-related fluff added?  Surely
> the correct approach is for C++ code to wrap C includes in
> extern "C", as the syntax would seem to indicate?
>
> Every other language that has some way of linking to C code (and
> most do) has to do its own compatability shimming, with it's own
> foreign-function interface.
>
> My guess is that the process started when it was assumed that
> C++ was the future of C, and eventually everything would just be
> C++.  That is much less clearly the case, now, IMO.
>
> Cheers,
>
> --
> Andrew
>
> -------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It's the best place to buy or sell services for
> just about anything Open Source.
> http://sourceforge.net/services/buy/index.php
> _______________________________________________
> Math-atlas-devel mailing list
> Math-atlas-devel@...
> https://lists.sourceforge.net/lists/listinfo/math-atlas-devel
>

--
Email: david.w.h.chin AT gmail.com dwchin AT lroc.harvard.edu
http://gallatin.physics.lsa.umich.edu/~dwchin/Work
Public key: http://david.w.h.chin.googlepages.com/crypto.html
pub 1024D/553CB7B8 2007-08-07 [expires: 2008-08-06]
 Key fingerprint = 0AD6 6DDC 204D AA4A 8957 52D1 0D93 C52A 553C B7B8
uid David Chin <david.w.h.chin AT gmail.com>
uid David Chin <dwchin AT lroc.harvard.edu>
uid David Chin <dwchin.lroc AT gmail.com>

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Math-atlas-devel mailing list
Math-atlas-devel@...
https://lists.sourceforge.net/lists/listinfo/math-atlas-devel

[Andrew Reilly-9]

On Tue, Jun 24, 2008 at 08:27:59PM -0400, David Chin wrote:
> The __cplusplus macro is part of the ISO C++ standard, defined
> precisely because the standards committee knew (presumably) that there
> would be people who would want to use their old plain C stuff.
>
> Similarly, ISO C has the macro __STDC__ defined.

Both true, but irrelevant.  The C++ coders have extern "C" so
that they can use their old C stuff.  The __cplusplus define
is so that you can *pretend* that C code is C++ code.

> See here:  http://docs.sun.com/app/docs/doc/819-5267/bkaep?a=view
>
> So, it is more than just idiom. It's required of a standards-compliant compiler.

No, it's just idiom, because these are C headers that we're
talking about.

Done properly, the C++ code would have:

extern "C" {
#include <atlas.h>
}

and then everyone would be happy, and there would be no need for
spurious #ifdef bracketing in the C header.

The problem is the whole notion that a header file can
meaningfully be both a C header file and a C++ header file.  If
you replace C++ by, say, Eiffel in that sentence, the statement
gets a bit more obvious.  

> Cheers,
>     Dave
>
> On Tue, Jun 24, 2008 at 7:13 PM, Andrew Reilly
> <andrew-atlas@...> wrote:
> > On Tue, Jun 24, 2008 at 05:50:10PM -0500, skip@... wrote:
> >>     Clint> #ifdef __cplusplus
> >>     Clint> extern "C"
> >>     Clint> {
> >>     Clint> #endif
> >>     Clint> ....
> >
> >> That is the standard idiom.  It will instruct the C++ compiler to not mangle
> >> the names between the braces (C linkage).  C compilers will not see it.

C compilers mightn't see it, but C coders will, and they'll
cringe every time.  I know that I do.

Cheers,

--
Andrew

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Math-atlas-devel mailing list
Math-atlas-devel@...
https://lists.sourceforge.net/lists/listinfo/math-atlas-devel