Weak Symbols

[Neelesh Bodas]

Hi all,
does C++ give a way to declare a specific symbol as a "weak symbol"?
In 'C', one can use #pragma directive.
Is this still a legal way in C++ ? Is there a (better?) alternative?

[Mike Wahler]

"Neelesh Bodas" <ne***********@gmail.com> wrote in message
news:11**********************@g47g2000cwa.googlegr oups.com...

Hi all,
does C++ give a way to declare a specific symbol as a "weak symbol"?
What is a 'weak symbol'?
In 'C', one can use #pragma directive.
How so?
Is this still a legal way in C++ ? Is there a (better?) alternative?

Specifically what are you trying to do?

-Mike

[Neelesh Bodas]

Mike Wahler wrote:

What is a 'weak symbol'?
A weak symbol is one which can have legal mutiple definitions across
compilation units.
For example , suppose you have :
//a.c
int fun(int a)
{
return a;
}

//b.c

int fun(int b)
{
return 2*b;
}

Now these files in general cannot be linked together since linker will
complain about multiple definitions of "fun"

In 'C', one can use #pragma directive.
How so?

So if you add the line
#pragma weak fun
to either or both of these files, then it means that the symbol fun
_can_ be multiply defined across compilation units, and the current
definition (which is marked as "weak") will be safely ignored by the
linker. (Of course this is implementation dependent, but each
implementation has its own way. I am talking about gcc in this
example.)

Is this still a legal way in C++ ? Is there a (better?) alternative?

Specifically what are you trying to do?

So specifically what I am trying to ask is whether there is a
language-level support for this concept of "multiple definitions across
compilation units?"
I believe that the "#pragma" issue is compiler dependent. So I was
searching for any such compiler-independent way.

May be this post is a bit off topic. Apologies in that case, I will
post on an approriate group.

[Roy Smith]

In article <11**********************@g47g2000cwa.googlegroups .com>,
"Neelesh Bodas" <ne***********@gmail.com> wrote:

So specifically what I am trying to ask is whether there is a
language-level support for this concept of "multiple definitions across
compilation units?"
I believe that the "#pragma" issue is compiler dependent. So I was
searching for any such compiler-independent way.

Isn't this what namespaces are for?

[Alf P. Steinbach]

* Neelesh Bodas:

Mike Wahler wrote:

What is a 'weak symbol'?

A weak symbol is one which can have legal mutiple definitions across
compilation units.
For example , suppose you have :
//a.c
int fun(int a)
{
return a;
}

//b.c

int fun(int b)
{
return 2*b;
}

Now these files in general cannot be linked together since linker will
complain about multiple definitions of "fun"

In 'C', one can use #pragma directive.

How so?

So if you add the line
#pragma weak fun
to either or both of these files, then it means that the symbol fun
_can_ be multiply defined across compilation units,

That is not standard C (it's a gcc extension), and the explanation is
incorrect: what it does is allow you to create a library function that
can be "overridden" by a definition in user code, a non-weak symbol.

If what you want is to use the same function name and signature for
different functions in different compilation units, then in in standard
C you can use the keyword 'static', and in standard C++ you can
additionally use a namespace, e.g. an anonymous one.

Alternatively, if you guarantee that all definitions are exactly the
same, then in standard C++ you can use the keyword 'inline'.

--
A: Because it messes up the order in which people normally read text.
Q: Why is it such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?

[Jonathan Mcdougall]

Alf P. Steinbach wrote:

* Neelesh Bodas:

Mike Wahler wrote:

What is a 'weak symbol'?
A weak symbol is one which can have legal mutiple definitions across
compilation units.
For example , suppose you have :
//a.c
int fun(int a)
{
return a;
}

//b.c

int fun(int b)
{
return 2*b;
}

Now these files in general cannot be linked together since linker will
complain about multiple definitions of "fun"

> In 'C', one can use #pragma directive.

How so?

So if you add the line
#pragma weak fun
to either or both of these files, then it means that the symbol fun
_can_ be multiply defined across compilation units,

That is not standard C (it's a gcc extension), and the explanation is
incorrect: what it does is allow you to create a library function that
can be "overridden" by a definition in user code, a non-weak symbol.

If what you want is to use the same function name and signature for
different functions in different compilation units, then in in standard
C you can use the keyword 'static', and in standard C++ you can
additionally use a namespace, e.g. an anonymous one.

Well it's not really "additionally". In C++, using static to specify
internal linkage is deprecated. Anonymous namespaces *should* be used
instead.
Alternatively, if you guarantee that all definitions are exactly the
same, then in standard C++ you can use the keyword 'inline'.

Jonathan

[Alf P. Steinbach]

* Jonathan Mcdougall:

* Alf P. Steinbach:

If what you want is to use the same function name and signature for
different functions in different compilation units, then in in standard
C you can use the keyword 'static', and in standard C++ you can
additionally use a namespace, e.g. an anonymous one.
Well it's not really "additionally". In C++, using static to specify
internal linkage is deprecated.

I'm sorry, that's incorrect.

Use of static is deprecated for declaring objects in namespace scope.

Not for functions.

Anonymous namespaces *should* be used instead.

We all have our preferences... ;-)

--
A: Because it messes up the order in which people normally read text.
Q: Why is it such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?

[Jonathan Mcdougall]

Alf P. Steinbach wrote:

* Jonathan Mcdougall:

* Alf P. Steinbach:

If what you want is to use the same function name and signature for
different functions in different compilation units, then in in standard
C you can use the keyword 'static', and in standard C++ you can
additionally use a namespace, e.g. an anonymous one.

Well it's not really "additionally". In C++, using static to specify
internal linkage is deprecated.

I'm sorry, that's incorrect.

Use of static is deprecated for declaring objects in namespace scope.

Not for functions.

A0.1.2 The use of the static keyword is deprecated when declaring
objects in namespace scope (see _basic.scope.namespace_).

Well, I think I stand corrected. Somehow, I knew I couldn't be right.
:)

Jonathan

[Phlip]

Neelesh Bodas wrote:

So specifically what I am trying to ask is whether there is a
language-level support for this concept of "multiple definitions across
compilation units?"

namespace?

Otherwise, you are asking for an extension to C that someone added to their
compiler to deal with that language's lack of namespaces.

--
Phlip
http://www.greencheese.org/ZeekLand <-- NOT a blog!!!

[Branimir Maksimovic]

Neelesh Bodas wrote:

Mike Wahler wrote:

What is a 'weak symbol'?

A weak symbol is one which can have legal mutiple definitions across
compilation units.

Or no definition at all.
#pragma weak fun
void fun(void);
if(fun)......

This is good for simulating inheritance and function overriding in C,
but C++
has that already.

class Base{
public:
virtual void fun(void)=0;
};

........
void (Base* b)
{
b->fun();
//.......
}
................

As a side note some version of libpthread under linux has
pthread_create and
I guess rest of pthread_xxx functions as weak symbols.
One poor programer forgot to link the library and of course linker
does not complains, effectivelly making every call to a pthread_create
into a
(0)(......) (call to pure virtual function :) which produced segfault.
stack trace easilly showed what's going on.

Greetings, Bane.