[Bug c++/34480] New: Argument packs treat __null oddly

[Bug c++/34480] New: Argument packs treat __null oddly

[gcc at magfr dot user dot lysator dot liu dot se]

The following program fails to compile in c++0x mode:

template <class ... A> void* f(A... a) { return new (char*)(a...); }

int main() {
    f(__null);
}

Now, this doesn't sound that bad until one notice that this is functionally
equivalent with

#include <vector>

int main() {
    std::vector<char*> v;
    v.push_back(NULL);
}

and that used to and still do work perfectly well under c++98.

Note that if __null is replaced with 0 then it works so this suggests that the
__null case have been forgotten somewhere.


-- 
           Summary: Argument packs treat __null oddly
           Product: gcc
           Version: 4.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: gcc at magfr dot user dot lysator dot liu dot se
 GCC build triplet: i586-pc-linux-gnu
  GCC host triplet: i586-pc-linux-gnu
GCC target triplet: i586-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34480

[Bug c++/34480] Argument packs treat __null oddly

[pinskia at gcc dot gnu dot org]

------- Comment #1 from pinskia at gcc dot gnu dot org  2007-12-15 20:10 -------
Well what type does NULL would have when passed?

And passing 0 also fails.

This might be a defect in the C++0x part of the standard.

Also I think we really do have an ODR violation in libstdc++ if the methods
change with -std=c++0x (which I think the C++0x standard really should
mention).


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pinskia at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34480

[Bug c++/34480] Argument packs treat __null oddly

[sylvain dot pion at sophia dot inria dot fr]

------- Comment #2 from sylvain dot pion at sophia dot inria dot fr  2008-01-02 21:37 -------
In fact, this is not really an issue with variadic templates,
since you get the same problem with:

void* g(int a) { return new (char*)(a); }

int main() {
            g(__null);
}


Nevertheless, I opened an issue for the LWG for the original problem
with std::vector::push_back() which bothers me as well.  The preview
of it can be found here:
http://home.twcny.rr.com/hinnant/cpp_extensions/issues_preview/lwg-active.html#767


I suggest that the component for this bug be changed from c++ to libstdc++,
and the title to something like:

 "[LWG Issue 767] push_back and push_front need overloads for type T"


-- 

sylvain dot pion at sophia dot inria dot fr changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |sylvain dot pion at sophia
                   |                            |dot inria dot fr


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34480

[Bug c++/34480] Argument packs treat __null oddly

[pcarlini at suse dot de]

------- Comment #3 from pcarlini at suse dot de  2008-01-02 21:53 -------
(In reply to comment #1)
> I suggest that the component for this bug be changed from c++ to libstdc++,
> and the title to something like:
> 
>  "[LWG Issue 767] push_back and push_front need overloads for type T"

Please don't. I mean, the PR would be immediately suspended, of course. We
don't want to have PRs for each (thousands) step toward C++0x...


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34480

[Bug c++/34480] Argument packs treat __null oddly

[dgregor at gcc dot gnu dot org]

------- Comment #4 from dgregor at gcc dot gnu dot org  2008-01-18 16:32 -------
Confirmed. This is the right behavior according to the C++0x specification, but
the backward-compatibility issue with push_back is a problem. The C++ committee
is aware is the issue.


-- 

dgregor at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2008-01-18 16:32:16
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34480

[Bug libstdc++/34480] Argument packs treat __null oddly

[dgregor at gcc dot gnu dot org]

------- Comment #5 from dgregor at gcc dot gnu dot org  2008-01-18 16:33 -------
Suspended for now. We'll pick up the library work again once the C++ committee
has resolved this issue. 


-- 

dgregor at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |SUSPENDED
          Component|c++                         |libstdc++


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34480

[Bug libstdc++/34480] Argument packs treat __null oddly

[paolo dot carlini at oracle dot com]

------- Comment #6 from paolo dot carlini at oracle dot com  2008-10-02 13:52 -------
The issue with vector::push_back is fixed in mainline. The issue with std::pair
(per DR 767) is still there, however. Maybe Doug can add something about the
latter...


-- 

paolo dot carlini at oracle dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |paolo at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34480

[Bug libstdc++/34480] Argument packs treat __null oddly

[paolo dot carlini at oracle dot com]

------- Comment #7 from paolo dot carlini at oracle dot com  2008-10-18 23:05 -------
The issue with std::pair is now also fixed by:

2008-10-18  Douglas Gregor  <doug.gregor@gmail.com>

        * include/bits/stl_pair.h (__may_be_null_pointer_init): New.
        (pair::pair): Eliminate the redundant pair(U1&&, U2&&) constructor.
        Add lvalue pair<U1, U2> constructor to handle non-const pair lvalues.
        Remove the old variadic constructor, and instead provide several
        variadic constructors that avoid failing when attempting to
        initialize a pointer from a null pointer constant.
        * testsuite/20_util/pair/moveable.cc (test3): Add new tests with
        initialization of pointers from the null pointer constant.


-- 

paolo dot carlini at oracle dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|SUSPENDED                   |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34480