[Bug c++/28017] New: lack of guard variables for explicitly instantiated template static data

[Bug c++/28017] New: lack of guard variables for explicitly instantiated template static data

[hhinnant at apple dot com]

I was passed a test case consisting of two translation units:

// test.h

#include <iostream>

class A
{
public:

    A()
    {
        mString = new char[2];
        std::cout << "new: mString has value " << (void*) mString << std::endl;
    }

    ~A()
    {
        std::cout << "delete: mString has value " << (void*) mString <<
std::endl;
        delete [] mString;
    }

    void f()
    {
    }

private:

    char* mString;
};


template<typename T>
class Test
{
public:

    Test()
    {
        sA.f();
    }

private:
    static A sA;
};

template<typename T>
A Test<T>::sA;

// test.cpp

#include "test.h"

template class Test<int>;

int foo()
{
    return 0;
}

// main.cpp

#include "test.h"

int foo();

int main()
{
    Test<int> theTest;
    foo();
    return 0;
}

When compiled with something like:

g++ -o test test.cpp main.cpp

the test shows evidence of the static Test<int>::sA being
constructed/destructed twice:

new: mString has value 0x5002d0
new: mString has value 0x500300
delete: mString has value 0x500300
delete: mString has value 0x500300

The desired output should be more like:

new: mString has value 0x5002d0
delete: mString has value 0x5002d0

Further inspection reveals that the guard variable for this static is not being
generated in test.cpp because of the explicit instantiation there (tested on
Apple's gcc 4.0.1).

I am wondering if an appropriate fix might be in:

gcc/cp/decl2.c, in function start_static_initialization_or_destruction

Change the if statement that looks like:

  if (TREE_PUBLIC (decl) && (DECL_COMMON (decl)
                             || DECL_ONE_ONLY (decl)
                             || DECL_WEAK (decl)))
    {
      tree guard_cond;

to:

  if (TREE_PUBLIC (decl) && (DECL_COMMON (decl)
                             || DECL_ONE_ONLY (decl)
                             || DECL_WEAK (decl)
                             || DECL_EXPLICIT_INSTANTIATION (decl)))
    {
      tree guard_cond;

I looked in the mainline decl2.c and found the same logic in the macro
NEEDS_GUARD_P.  Here I'm suggesting changing:

#define NEEDS_GUARD_P(decl) (TREE_PUBLIC (decl) && (DECL_COMMON (decl)      \
                                                    || DECL_ONE_ONLY (decl) \
                                                    || DECL_WEAK (decl)))

to:

#define NEEDS_GUARD_P(decl) (TREE_PUBLIC (decl) && (DECL_COMMON (decl)      \
                                                    || DECL_ONE_ONLY (decl) \
                                                    || DECL_WEAK (decl) \
                                                    ||
DECL_EXPLICIT_INSTANTIATION (decl)))


-- 
           Summary: lack of guard variables for explicitly instantiated
                    template static data
           Product: gcc
           Version: 4.0.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: hhinnant at apple dot com
  GCC host triplet: darwin ppc


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[pinskia at gcc dot gnu dot org]

------- Comment #1 from pinskia at gcc dot gnu dot org  2006-06-13 19:14 -------
This works on x86-linux-gnu on the mainline.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2006-06-13 19:15 -------
(In reply to comment #1)
> This works on x86-linux-gnu on the mainline.

Oh and in 3.3.3.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[pinskia at gcc dot gnu dot org]

------- Comment #3 from pinskia at gcc dot gnu dot org  2006-06-13 19:18 -------
   || DECL_ONE_ONLY (decl) || DECL_WEAK (decl) \

Actually those looks should include what is defined for Darwin.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   GCC host triplet|darwin ppc                  |
 GCC target triplet|                            |*-darwin


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[pinskia at gcc dot gnu dot org]

------- Comment #4 from pinskia at gcc dot gnu dot org  2006-06-13 19:37 -------

   For Darwin we do not want explicit instantiations to be
   linkonce.  */


This is why this testcase fails on darwin.
We should instead of just adding DECL_EXPLICIT_INSTANTIATION, check
TARGET_WEAK_NOT_IN_ARCHIVE_TOC.

(!TARGET_WEAK_NOT_IN_ARCHIVE_TOC
      || (! DECL_EXPLICIT_INSTANTIATION (decl)
          && ! DECL_TEMPLATE_SPECIALIZATION (decl)))

This is a darwin only issue.


-- 

pinskia 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         |2006-06-13 19:37:25
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[hhinnant at apple dot com]

------- Comment #5 from hhinnant at apple dot com  2006-06-13 21:23 -------
(In reply to comment #4)
>    For Darwin we do not want explicit instantiations to be
>    linkonce.  */
> 
> 
> This is why this testcase fails on darwin.
> We should instead of just adding DECL_EXPLICIT_INSTANTIATION, check
> TARGET_WEAK_NOT_IN_ARCHIVE_TOC.
> 
> (!TARGET_WEAK_NOT_IN_ARCHIVE_TOC
>       || (! DECL_EXPLICIT_INSTANTIATION (decl)
>           && ! DECL_TEMPLATE_SPECIALIZATION (decl)))
> 
> This is a darwin only issue.

I'm having trouble deciding exactly what you mean.  Is this what you mean:

#define NEEDS_GUARD_P(decl) (!TARGET_WEAK_NOT_IN_ARCHIVE_TOC \
      || (! DECL_EXPLICIT_INSTANTIATION (decl) \
          && ! DECL_TEMPLATE_SPECIALIZATION (decl)))

Or do you mean:

#define NEEDS_GUARD_P(decl) (TREE_PUBLIC (decl) && (DECL_COMMON (decl)      \
                                                    || DECL_ONE_ONLY (decl) \
                                                    || DECL_WEAK (decl) \
                                                    ||
(!TARGET_WEAK_NOT_IN_ARCHIVE_TOC \
      || (! DECL_EXPLICIT_INSTANTIATION (decl) \
          && ! DECL_TEMPLATE_SPECIALIZATION (decl)))))

?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

Re: [Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[Andrew Pinski]

> #define NEEDS_GUARD_P(decl) (TREE_PUBLIC (decl) && (DECL_COMMON (decl)      \
>                                                     || DECL_ONE_ONLY (decl) \
>                                                     || DECL_WEAK (decl) \
>                                                     ||
> (!TARGET_WEAK_NOT_IN_ARCHIVE_TOC \
>       || (! DECL_EXPLICIT_INSTANTIATION (decl) \
>           && ! DECL_TEMPLATE_SPECIALIZATION (decl)))))
> 
> ?

The latter.

-- Pinski

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[pinskia at physics dot uc dot edu]

------- Comment #6 from pinskia at physics dot uc dot edu  2006-06-13 21:24 -------
Subject: Re:  lack of guard variables for explicitly instantiated template
static data

> #define NEEDS_GUARD_P(decl) (TREE_PUBLIC (decl) && (DECL_COMMON (decl)      \
>                                                     || DECL_ONE_ONLY (decl) \
>                                                     || DECL_WEAK (decl) \
>                                                     ||
> (!TARGET_WEAK_NOT_IN_ARCHIVE_TOC \
>       || (! DECL_EXPLICIT_INSTANTIATION (decl) \
>           && ! DECL_TEMPLATE_SPECIALIZATION (decl)))))
> 
> ?

The latter.

-- Pinski


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

Re: [Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[Andrew Pinski]

> 
> 
> 
> ------- Comment #7 from hhinnant at apple dot com  2006-06-13 21:41 -------
> (In reply to comment #6)
> > Subject: Re:  lack of guard variables for explicitly instantiated template
> > static data
> > 
> > > #define NEEDS_GUARD_P(decl) (TREE_PUBLIC (decl) && (DECL_COMMON (decl)      \
> > >                                                     || DECL_ONE_ONLY (decl) \
> > >                                                     || DECL_WEAK (decl) \
> > >                                                     ||
> > > (!TARGET_WEAK_NOT_IN_ARCHIVE_TOC \
> > >       || (! DECL_EXPLICIT_INSTANTIATION (decl) \
> > >           && ! DECL_TEMPLATE_SPECIALIZATION (decl)))))
> > > 
> > > ?
> > 
> > The latter.
> 
> Thanks.  But this doesn't pass the test case on darwin.  I'm not familiar
> enough with the C++ FE to understand TARGET_WEAK_NOT_IN_ARCHIVE_TOC.  Could you
> double check the above.  The "!" in front of DECL_EXPLICIT_INSTANTIATION looks
> especially suspicious to me.

You want the opposite of that like:
(TARGET_WEAK_NOT_IN_ARCHIVE_TOC && (DECL_EXPLICIT_INSTANTIATION (decl) || DECL_TEMPLATE_SPECIALIZATION (decl)))

I was quoting the case when DECL_WEAK would be set on the decl.
TARGET_WEAK_NOT_IN_ARCHIVE_TOC is only defined to 1 for darwin.  

-- Pinski

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[hhinnant at apple dot com]

------- Comment #7 from hhinnant at apple dot com  2006-06-13 21:41 -------
(In reply to comment #6)
> Subject: Re:  lack of guard variables for explicitly instantiated template
> static data
> 
> > #define NEEDS_GUARD_P(decl) (TREE_PUBLIC (decl) && (DECL_COMMON (decl)      \
> >                                                     || DECL_ONE_ONLY (decl) \
> >                                                     || DECL_WEAK (decl) \
> >                                                     ||
> > (!TARGET_WEAK_NOT_IN_ARCHIVE_TOC \
> >       || (! DECL_EXPLICIT_INSTANTIATION (decl) \
> >           && ! DECL_TEMPLATE_SPECIALIZATION (decl)))))
> > 
> > ?
> 
> The latter.

Thanks.  But this doesn't pass the test case on darwin.  I'm not familiar
enough with the C++ FE to understand TARGET_WEAK_NOT_IN_ARCHIVE_TOC.  Could you
double check the above.  The "!" in front of DECL_EXPLICIT_INSTANTIATION looks
especially suspicious to me.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[pinskia at physics dot uc dot edu]

------- Comment #8 from pinskia at physics dot uc dot edu  2006-06-13 21:47 -------
Subject: Re:  lack of guard variables for explicitly instantiated template
static data

> 
> 
> 
> ------- Comment #7 from hhinnant at apple dot com  2006-06-13 21:41 -------
> (In reply to comment #6)
> > Subject: Re:  lack of guard variables for explicitly instantiated template
> > static data
> > 
> > > #define NEEDS_GUARD_P(decl) (TREE_PUBLIC (decl) && (DECL_COMMON (decl)      \
> > >                                                     || DECL_ONE_ONLY (decl) \
> > >                                                     || DECL_WEAK (decl) \
> > >                                                     ||
> > > (!TARGET_WEAK_NOT_IN_ARCHIVE_TOC \
> > >       || (! DECL_EXPLICIT_INSTANTIATION (decl) \
> > >           && ! DECL_TEMPLATE_SPECIALIZATION (decl)))))
> > > 
> > > ?
> > 
> > The latter.
> 
> Thanks.  But this doesn't pass the test case on darwin.  I'm not familiar
> enough with the C++ FE to understand TARGET_WEAK_NOT_IN_ARCHIVE_TOC.  Could you
> double check the above.  The "!" in front of DECL_EXPLICIT_INSTANTIATION looks
> especially suspicious to me.

You want the opposite of that like:
(TARGET_WEAK_NOT_IN_ARCHIVE_TOC && (DECL_EXPLICIT_INSTANTIATION (decl) ||
DECL_TEMPLATE_SPECIALIZATION (decl)))

I was quoting the case when DECL_WEAK would be set on the decl.
TARGET_WEAK_NOT_IN_ARCHIVE_TOC is only defined to 1 for darwin.  

-- Pinski


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[hhinnant at apple dot com]

------- Comment #9 from hhinnant at apple dot com  2006-06-13 22:02 -------
(In reply to comment #8)

Thanks.  That not only makes sense to me now, but it passes the test. :-)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[hhinnant at apple dot com]

------- Comment #10 from hhinnant at apple dot com  2006-06-19 18:11 -------
It turns out this still isn't quite right.  Looks like we need:

#define NEEDS_GUARD_P(decl) (TREE_PUBLIC (decl) && (DECL_COMMON (decl) \
                                                    || DECL_ONE_ONLY (decl) \
                                                    || DECL_WEAK (decl) \
                                                    ||
(TARGET_WEAK_NOT_IN_ARCHIVE_TOC \
                                                       && DECL_LANG_SPECIFIC
(decl) \
                                                       &&
(DECL_EXPLICIT_INSTANTIATION (decl) \
                                                       || 
DECL_TEMPLATE_SPECIALIZATION (decl)))))

The former solution was dereferencing a null pointer.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[pinskia at gcc dot gnu dot org]

------- Comment #11 from pinskia at gcc dot gnu dot org  2006-06-21 04:50 -------
Actually I take that back, it is also a problem with targets that don't have
weak symbols too.
On x86_64-linux-gnu with the additional flag of -fno-weak:
pc64:~> g++ test.cc test1.cc -fno-weak
pc64:~> ./a.out
new: mString has value 0x502010
new: mString has value 0x502030
delete: mString has value 0x502030
delete: mString has value 0x502030
*** glibc detected *** double free or corruption (fasttop): 0x0000000000502030
***
Abort

So the new part of the test should be just shrunk back to testing
DECL_EXPLICIT_INSTANTIATION instead of the other parts too.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 GCC target triplet|*-darwin                    |non weak
   Target Milestone|---                         |4.0.4


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[pinskia at gcc dot gnu dot org]

------- Comment #12 from pinskia at gcc dot gnu dot org  2006-06-21 05:05 -------
-fno-weak worked in 3.0.4 and below, though I don't know if it was really
working so I am not marking this as a regression.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|                            |3.2.3 3.4.0 4.0.0 4.1.0
                   |                            |4.2.0
   Target Milestone|4.0.4                       |---


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[bkoz at gcc dot gnu dot org]

------- Comment #13 from bkoz at gcc dot gnu dot org  2009-01-27 17:26 -------
*** Bug 25956 has been marked as a duplicate of this bug. ***


-- 

bkoz at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jkp at kirkconsulting dot co
                   |                            |dot uk


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

[Bug c++/28017] lack of guard variables for explicitly instantiated template static data

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=28017

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=80320
      Known to fail|                            |

--- Comment #14 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I think this and PR 80320 both have the same underlying issue.