www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.bugs - [Issue 3555] New: Const function modifies a field when passed a delegate

reply d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=3555

           Summary: Const function modifies a field when passed a delegate
           Product: D
           Version: 2.036
          Platform: x86
        OS/Version: Windows
            Status: NEW
          Severity: normal
          Priority: P2
         Component: DMD
        AssignedTo: nobody puremagic.com
        ReportedBy: tomeksowi gmail.com



PST ---
I believe this should not compile:

class A
{ public:
    int pole;

    void funkcja(void delegate(int) dg) const
    {
        dg(3);
    }
}

void main()
{
    A a = new A;

    void zmien(int w)
    {
        a.pole = w;
    }

    a.funkcja(&zmien);  // pole changed by calling const funkcja
    assert (a.pole == 3);
}

Looks very similar to bug 3534.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Nov 28 2009
next sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=3555


Jonathan M Davis <jmdavisProg gmx.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jmdavisProg gmx.com



PST ---
This is not a bug. And the reason is simple: all marking a function as const
does is make its this reference const. So, funkcja becomes something like

void funkcja(const A this, void delegate(int) dg)
{
    dg(3);
}


The delegate has a separate reference to the object that this points to. Even
if you have a pointer or reference to const, it's perfectly legal to alter what
they refer to through another pointer or reference - const just protects what
is const. If you want to _guarantee_ that a const function cannot in any way
alter any aspect of the class or struct that its a part if, that function must
be both const and strongly pure. So, it'll have to be marked const and pure,
and all of its parameters must either be immutable or implicitly castable to
immutable (so, they must either be primitive types or they must be structs that
don't hold non-immutable pointers or references). Otherwise, it's conceivable
that you'll alter the value of the object that this points to through some
indirection. In most cases, you obviously won't but it _is_ possible.

If the object referred to by this were immutable, then that wouldn't be a
problem, because then _no_ pointer or reference could alter it, but as long as
the item referred to be const is actually mutable, then other pointers or
references which aren't pointers or references to const can alter that item.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Dec 12 2010
prev sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=3555


Walter Bright <bugzilla digitalmars.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
                 CC|                            |bugzilla digitalmars.com
         Resolution|                            |INVALID



13:30:27 PST ---
Jonathan's remarks are correct. Not a bug.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Dec 12 2010