[Bug c++/63364] New: GCC optimizer causing memory corruption

[Bug c++/63364] New: GCC optimizer causing memory corruption

[contact at ncomputers dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63364

            Bug ID: 63364
           Summary: GCC optimizer causing memory corruption
           Product: gcc
           Version: 4.8.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: contact at ncomputers dot org

http://ncomputers.org/content/code.php?src=findings/gcc%20optimizer%20memory%20corruption.cpp

We found that probably some optimization flags of the GCC compiler are causing
memory corruption after the reading of a variable's content through a pointer
to a class or a struct.

We've tested and confirmed this bug with these compiler's versions: 4.7.2 and
4.8.2

#include<iostream>
using namespace std;

struct A{
    unsigned int const v;
    A():v(10){}
    void test_A();
};

struct B{
    A const*const&a;
    B(A const*const&aa):a(aa){}
    void test_B(){
        unsigned int it=0;
        cout<<"Value of constant a->v: "<<a->v<<endl;
        do{
            cout<<it<<',';
        }while(++it<a->v);
        cout<<endl<<"Value of constant a->v: "<<a->v<<endl;
        //cout<<endl<<"Value of constant a->v:"<<' '<<a->v<<endl;

        /*
         * GNU Compiler with -O2 or -O3 flag
         * Value of constant a->v changes to zero
         *
         * If this result was shown:
         *
         * Value of constant a->v: 10
         * 0,
         * Value of constant a->v: 0
         *
         * Now comment the previous "cout" line and uncomment the next "cout"
line.
         * Note that this is the only difference between both lines:
         * <<' '
         *
         * The right result should be shown:
         *
         * Value of constant a->v: 10
         * 0,1,2,3,4,5,6,7,8,9,
         * Value of constant a->v: 10
         */
    }
};

void A::test_A(){
    B*b=new B(this);
    b->test_B();
    delete b;
}

int main(){
    A*a=new A();
    a->test_A();
    delete a;
    return 0;
};

[Bug c++/63364] GCC optimizer causing memory corruption

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63364

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |INVALID

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
>    B(A const*const&aa):a(aa){}

>    B*b=new B(this);


There is a temporary variable being created there and it goes out of scope
after the statement is finished.   The reason is because this is not a lvalue,
prvalue.

Here is the quote from the standard:
In the body of a non-static (9.3) member function, the keyword this is a
prvalue expression whose value
is the address of the object for which the function is called.

[Bug c++/63364] GCC optimizer causing memory corruption

[contact at ncomputers dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63364

--- Comment #2 from ncomputers.org <contact at ncomputers dot org> ---
(In reply to Andrew Pinski from comment #1)
> >    B(A const*const&aa):a(aa){}
> 
> >    B*b=new B(this);
> 
> 
> There is a temporary variable being created there and it goes out of scope
> after the statement is finished.   The reason is because this is not a
> lvalue, prvalue.
> 
> Here is the quote from the standard:
> In the body of a non-static (9.3) member function, the keyword this is a
> prvalue expression whose value
> is the address of the object for which the function is called.

Andrew, thank you for your answer!

Yes it is true! We understand now better the reference variables.

Can we write your name in our website?

"Andrew Pinski taught us something new".

[Bug c++/63364] GCC optimizer causing memory corruption

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63364

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to ncomputers.org from comment #2)
> Can we write your name in our website?
> 
> "Andrew Pinski taught us something new".

I guess.